The Analytics Times

The Ultimate Guide to Embedded Analytics

Establishing a trusted data foundation for AI

Introduction

Artificial intelligence (AI) is expected to greatly improve industries like healthcare, manufacturing, and customer service, leading to higher-quality experiences for customers and employees alike. Indeed, AI technologies like machine learning (ML) have already helped data practitioners produce mathematical predictions, generate insights, and improve decision-making.

Furthermore, emerging AI technologies like generative AI (GenAI) can create strikingly realistic content that has the potential to enhance productivity in virtually every aspect of business. However, AI can’t succeed without good data, and this paper describes six principles for
ensuring your data is AI-ready

The six principles for AI-ready data

It would be foolish to believe that you could just throw data at various AI initiatives and expect magic to happen, but that’s what many practitioners do. While this approach might seem to work for the first few AI projects, data scientists increasingly spend more time correcting
and preparing the data as projects mature.

Additionally, data used for AI has to be high-quality and precisely prepared for these intelligent applications. This means spending many hours manually cleaning and enhancing the data to ensure accuracy and completeness, and organizing it in a way that machines can easily understand. Also, this data often requires extra information — like definitions and labels — to enrich semantic meaning for automated learning and to help AI perform tasks more effectively.

Therefore, the sooner data can be prepared for downstream AI processes, the greater the benefit. Using prepped, AI-ready data is like giving a chef pre-washed and chopped vegetables instead of a whole sack of groceries — it saves effort and time and helps ensure that the final dish is promptly delivered. The diagram below defines six critical principles for ensuring the “readiness” of data and its suitability for AI use.

The remaining sections of this paper discuss each principle in detail.

Data has to be diverse.

Bias in AI systems, also known as machine learning or algorithm bias, occurs when AI applications produce results reflecting human biases, such as social inequality. This can happen when the algorithm development process includes prejudicial assumptions or, more commonly, when the training data has bias. For example, a credit score algorithm may deny a loan if it consistently uses a narrow band of financial attributes.

Consequently, our first principle focuses on providing a wide variety of data to AI models, which increases data diversity and reduces bias, helping to ensure that AI applications are less likely to make unfair decisions.

Diverse data means you don’t build your AI models on narrow and siloed datasets. Instead, you draw from a wide range of data sources spanning different patterns, perspectives, variations, and scenarios relevant to the problem domain. This data could be well-structured
and live in the cloud or on-premises. It could also exist on a mainframe, database, SAP system, or software as a service (SaaS) application. Conversely, the source data could be unstructured and live as files or documents on a corporate drive.

It’s essential to acquire diverse data in various forms for integration into your ML and GenAI applications.

Data has to be timely.

While it’s true that ML and GenAI applications thrive on diverse data, the freshness of that data is also crucial. Just as a weather forecast based on yesterday’s conditions isn’t conducive for a trip you plan to take today, AI models trained on outdated information can produce inaccurate or irrelevant results. Moreover, fresh data allows AI models to stay current with trends, adapt to changing circumstances, and deliver the best possible outcomes. Therefore, the second principle of AI-ready data is timeliness.

It’s critical that you build and deploy low-latency, real-time data pipelines for your AI initiatives to ensure timely data. Change data capture (CDC) is often used to deliver timely data from relational database systems, and stream capture is used for data originating from IoT devices that require low-latency processing. Once the data is captured, target repositories are updated and changes continuously applied in near-real time for the freshest possible data.

Remember, timely data enables more accurate and informed predictions.

Data has to be accurate.

The success of any ML or GenAI initiative hinges on one key ingredient: correct data. This is because AI models act like sophisticated sponges that soak up information to learn and perform tasks. If the information is inaccurate, it’s like the sponge is soaking up dirty water, leading to
biased outputs, nonsensical creations, and, ultimately, a malfunctioning AI system. Therefore, data accuracy is the third principle and a fundamental tenet for building reliable and trustworthy AI applications.

Data accuracy has three aspects. The first is profiling source data to understand its characteristics, completeness, distribution, redundancy, and shape. Profiling is also commonly known as exploratory data analysis, or EDA.

The second aspect is operationalizing remediation strategies by building, deploying, and continually monitoring the efficacy of data quality rules. Your data stewards may need to be involved here to aid with data deduplication and merging. Alternatively, AI can help automate and accelerate the process through machine-recommended data quality suggestions.

The final aspect is enabling data lineage and impact analysis — with tools for data engineers and scientists that highlight the impact of potential data changes and trace the origin of data to prevent accidental modification of the data used by AI models.

High-quality, accurate data ensures that models can identify relevant patterns and

relationships, leading to more precise decisions, generation, and predictions.

Data has to be secure.

AI systems often use sensitive data — including personally identifiable information (PII), financial records, or proprietary business information — and use of this data requires responsibility. Leaving data unsecured in AI applications is like leaving a vault door wide open. Malicious actors could steal sensitive information, manipulate training data to bias outcomes, or even disrupt entire GenAI systems. Securing data is paramount to protecting privacy, maintaining model integrity, and ensuring the responsible development of powerful AI applications. Therefore, data security is the fourth AI-ready principle.

Again, three tactics can help you automate data security at scale, since it’s nearly impossible to do it manually. Data classification detects, categorizes, and labels data that feeds the next stage. Data protection defines policies like masking, tokenization, and encryption to obfuscate the data. Finally, data security defines policies that describe access control, i.e., who can access the data. The three concepts work together as follows: first, privacy tiers should be defined and data tagged with a security designation of sensitive, confidential, or restricted. Next, a protection policy should be applied to mask restricted data. Finally, an access control policy should be used to limit access rights.

These three tactics protect your data and are crucial for improving

the overall trust in your AI system and safeguarding its reputational value.

Data has to be discoverable.

The principles we’ve discussed so far have primarily focused on promptly delivering the right data, in the correct format, to the right people, systems, or AI applications. But stockpiling data isn’t enough. AI-ready data has to be discoverable and readily accessible within the system. Imagine a library with all the books locked away — the knowledge is there but unusable. Discoverable data unlocks the true potential of ML and GenAI, allowing these workloads to find the information they need to learn, adapt, and produce ground-breaking results. Therefore,

discoverability is the fifth principle of AI-ready data.

Unsurprisingly, good metadata practices lie at the center of discoverability. Aside from the technical metadata associated with AI datasets, business metadata and semantic typing must also be defined. Semantic typing provides extra meaning for automated systems, while additional business terms deliver extra context to aid human understanding. A best practice is to create a business glossary that maps business terms to technical items in the datasets, ensuring a common understanding of concepts. AI-assisted augmentation can also be used to automatically generate documentation and add business semantics from the glossary. Finally, all the metadata is indexed and made searchable via a data catalog.

This approach ensures that the data is directly discoverable, applicable,

practical, and significant to the AI task at hand.

Data has to be easily consumable by MLs or LLMs.

We’ve already mentioned that ML and GenAI applications are mighty tools, but their potential rests on the ability to readily consume data. Unlike humans, who can decipher handwritten notes or navigate messy spreadsheets, these technologies require information to be represented in specific formats. Imagine feeding a picky eater — if they won’t eat what you’re serving, they’ll go hungry. Similarly, AI initiatives won’t be successful if the data is not in the right format for ML experiments or LLM applications. Making data easily consumable helps unlock the potential of these AI systems, allowing them to ingest information smoothly and translate it into intelligent actions for creative outputs. Consequently, making data readily consumable is the final principle of AI-ready data.

Making Data Consumable for Machine Learning

Data transformation is the unsung hero of consumable data for ML. While algorithms like linear regression grab the spotlight, the quality and shape of the data they’re trained on are just as critical. Moreover, the effort invested in cleaning, organizing, and making data consumable by ML models reaps significant rewards. Prepared data empowers models to learn effectively, leading to accurate predictions, reliable outputs, and, ultimately, the success of the entire ML project.

However, the training data formats depend highly on the underlying ML infrastructure. Traditional ML systems are disk-based, and much of the data scientist workflow focuses on establishing best practices and manual coding procedures for handling large volumes of files. More recently, lakehouse-based ML systems have used a database-like feature store, and the data scientist workflow has transitioned to SQL as a first-class language. As a result, well-formed, high-quality, tabular data structures are the most consumable and convenient data format
for ML systems.

Making Data Consumable for Generative AI

Large language models (LLMs) like OpenAI’s GPT-4, Anthropic’s Claude, and Google AI’s LaMDA and Gemini have been pre-trained on masses of text data and lie at the heart of GenAI. OpenAI’s GPT-3 model was estimated to be trained with approximately 45 TB of
data, exceeding 300 billion tokens. Despite this wealth of inputs, LLMs can’t answer specific questions about your business, because they don’t have access to your company’s data. The solution is to augment these models with your own information, resulting in more correct, relevant, and trustworthy AI applications.

The method for integrating your corporate data into an LLM-based application is called retrieval augmented generation, or RAG. The technique generally uses text information derived from unstructured, file-based sources such as presentations, mail archives, text documents, PDFs, transcripts, etc. The text is then split into manageable chunks and converted into a numerical representation used by the LLM in a process known as embedding. These embeddings are then stored in a vector database like Chroma, Pinecone, and Weviate. Interestingly, many traditional database vendors — such as PostgreSQL, Redis, and SingleStoreDB — also support vectors. Moreover, cloud platforms like Databricks, Snowflake, and Google BigQuery have recently added support for vectors, too.

Whether your source data is structured or unstructured, Qlik’s approach ensures that
quality data is readily consumable for your GenAI, RAG, or LLM-based applications.

The AI Trust Score

Having defined the six core principles of data readiness and suitability, the questions remain: can the principles be codified and easily translated for everyday use? And how can the readiness for AI be quickly discerned? One possibility is to use Qlik’s AI Trust Score as a global and understandable readiness indicator.

The AI Trust Score assigns a separate dimension for each principle and then aggregates each value to create a composite score, a quick, reliable shortcut to assessing your data’s AI readiness. Additionally, because enterprise data continually changes, the trust score is regularly checked and frequently recalibrated to track data readiness trends.

Your AI Trust Score aggregates multiple metrics into a single, easy to-understand readiness Score.

Qlik Talend data foundation for AI

The need for high-quality, real-time data that drives more thoughtful decisions, operational efficiency, and business innovation has never been greater. That’s why successful organizations seek market-leading data integration and quality solutions from Qlik Talend to efficiently deliver trusted data to warehouses, lakes, and other enterprise data platforms. Our comprehensive, best-in-class offerings use automated pipelines, intelligent transformations, and reliable Datasets quality to provide the agility data professionals crave with the governance and
compliance organizations expect.

So, whether you’re creating warehouses or lakes for insightful analytics, modernizing operational data infrastructures for business efficiency, or using multi-cloud data for artificial intelligence initiatives, Qlik Talend can show you the way.

Conclusion

Despite machine learning’s transformative power and generative AI’s explosive growth potential, data readiness is still the cornerstone
of any successful AI implementation. This paper described six key principles for establishing a robust and trusted data foundation that
combine to help your organization unlock AI’s true potential.

Next Steps

For more information or enquiries about Qlik products and services, feel free to contact us below.

Harness the Full Value of Your SAP Data for Cloud Analytics

Powered by Qlik and Snowflake

SAP is the Life Blood of Your Enterprise

Unlock your SAP data for faster time to insight and value

Today’s digital environment demands instant access to information for real-time decision-making, improved business agility, amazing customer service, and competitive advantage. Every person in your business, no matter what their role, requires easy access to the most accurate data set to make informed decisions – especially when it comes to SAP.

Industries like Manufacturing, Retail, CPG, Oil and Gas, and many others rely on actionable data from SAP to make strategic decisions and maintain competitive advantage

Today's Business Needs for Analytics

Faster Time to value

• Reduce time to insight with increased speed and agility
• More self-service analytics

Freedom

• Use SAP data anywhere
• Utilize 3rd party software tools most suited for the job at hand

Modern Analytics

• Combine SAP + non-SAP data
• Real-time, predictive (AI/ML)

Control Costs

• Facilitate scale with better cost control
• Reduce SAP Analytics TCO

Fully Leveraging SAP Data is Difficult

So why doesn’t every company embrace the opportunity to use new analytic platforms — by streaming live SAP data for real time analytics or combining it with other data sources in data lake and data warehouse platforms in the cloud?

SAP systems come with a number of unique challenges that are inherent to the platform. They are challenging to integrate because they are structurally complex with tens of thousands of tables that have intricate relationships as well as proprietary data formats making data inaccessible outside of SAP applications.

This complexity means that integrating data for analytics can be cumbersome, time consuming and costly.

SAP datasets are full of value, but they won’t do your organization any good unless you can easily use them in a cost-effective, secure way

The Fast Path to Extracting Value from SAP Data in the Cloud

Realize the value of all your SAP data with Qlik® and Snowflake

Qlik helps you efficiently capture large volumes of changed data from your source SAP systems, as well as other enterprise data sources, and deliver analytics ready data in real-time to Snowflake. Data sets are then easily cataloged, provisioned, and secured for all your analytic initiatives such as AI, machine learning, data science, BI
and operational reporting.

Qlik and Snowflake remove the business and technical challenges of complexity, time and cost, while gaining flexibility and agility for your SAP data. This unlocks your SAP data for different analytics use cases, and allows you to combine it with other enterprise data for even deeper insights and increased value.

SOLUTION USE CASE

Data Warehousing
Quickly load a new SAP data into the Snowflake Data Cloud in real-time gaining new and valuable insights.

Data Lake
Easily ingest or migrate SAP data into the Snowflake Data Cloud to support a variety of workloads unifying and securing all your organizations data in one place.

Data Science & AI/ML
Speed up workflows and transform SAP data into ML-powered insights using your language of choice with Snowflake.

The Benefits of Offloading SAP Data into Snowflake

The Snowflake platform provides a simple and elastic alternative for SAP customers that simultaneously ensures an organization’s critical information is protected. Snowflake was built from scratch to run in the cloud and frees companies from cloud provider lock-in.

Architectural Simplicity

Snowflake improves the ease of use of SAP with greater architectural simplicity. Snowflake can ingest data from SAP operations systems (both on-premises and cloud), third-party systems, and signal data – whether its in structured or semi-structured formats. This provides a foundation for customers to build 360-degree views of customers, products, and the supply chain enabling trusted business content to be accessible to all users.

Convenient Workload Elasticity

Snowflake’s multi-cluster shared data architecture separates compute from storage, enabling customers to elastically scale, up and down, automatically or on the fly. Users can apply dedicated compute clusters to each workload in near-unlimited quantities for virtually unlimited concurrency without contention. Once
a workload is completed, compute is dialed back down in seconds so customers don’t have to deal with the frustrations of throttling SAP BW queries.

Reliable Data Security

With Snowflake, simplicity also means data security. All data is always encrypted—in storage or in transit—as a built-in feature of the platform. Data is landed
once and views are shared out. This means one copy, one security model, and hundreds of elastic compute clusters with monitors on each one of them.

Accelerate the Delivery of Analytics-Ready SAP Data to Snowflake

Breaking the barriers to modernizing SAP data analytics

Ingest and deliver SAP data to Snowflake in real time.

Qlik accelerates the availability of SAP data to Snowflake in real-time with its scalable change data capture technology. Qlik Data Integration supports all core SAP and industry modules, including ECC, BW, CRM, GTS, and MDG, and continuously delivers incremental data updates with metadata to Snowflake in real time.

Automate the data warehouse lifecycle and pipeline.
Once your SAP data has landed in the Snowflake Data Cloud, Qlik automates the data pipeline without the hand coding associated with ETL approaches. You can efficiently model, refine and automate data warehouse or lake lifecycles to increase agility and productivity.

SAP Solution Accelerators
Qlik also provides a variety of Business Solutions such as Order to Cash. Unique SAP accelerators with preconfigured integrations and process flows leverage Qlik’s deep technical mastery of SAP structures and complexities. They enable real-time ingestion (CDC) and rapid integration of SAP data into Snowflake. The automated mapping and data model generation eliminate the need for expensive, highly technical and risky manual mapping and migration processes.

Discover bolder business insights.
Use business analytics tools such as Qlik Sense or any other BI tool to explore the data, create interactive dashboards, and carry out a variety of BI use cases for data-driven insights.

A Proven and Optimized Solution for Unlocking Insights from SAP

Real-time data ingestion (CDC) from SAP to Snowflake

Automated mapping and data model generation for analytics (data marts)

Prepackaged solution accelerators for common business use cases (order-to-cash, financials,
inventory management)

Decode SAP proprietary source structures (pool/cluster tables, HANA/CDS views)

Supports all core and industry-specific SAP modules

World-class SAP expertise to support presales and customer success

A Track Record of Collaboration and Success

Qlik is a Snowflake Elite partner with Snowflake Ready validated Solutions for Data Integration and Analytics. We accelerate time-to-insight with our end-to-end data integration and analytics solution taking you from your unrealized data value in your SAP landscape to informed action in your Snowflake Data Cloud. Qlik’s solution for Snowflake users automates the design, implementation and updates of data models while minimizing the manual, error-prone design processes of data modeling, ETL coding and scripting. As a result, you can accelerate
analytics projects, achieve greater agility and reduce risk – all while fully realizing the instant elasticity and cost advantages of Snowflake’s Data Cloud.

Next Steps

For more information or enquiries about Qlik products and services, feel free to contact us below.

Artificial Intelligence:
Our Strategy

How we infuse AI across our business,
from our products to how we operate

Introduction

A long-time leader and innovator in the data, analytics, and AI space, Qlik is perfectly positioned to fully embrace AI — not only in our products, but also in the way we conduct business — and do so responsibly. As the rise of generative AI accelerated the requirement for organizations to modernize their data fabric, it created new opportunities for Qlik to innovate in support of our customers’ efforts in developing and implementing their AI strategies. Over the past year, we have continued to lead through new acquisitions, product
innovation, talent development, technology investments, and by establishing new systems and processes.

Pillar 1

AI Foundation

AI can’t succeed without good data: It is fully dependent on an organization’s ability to establish a trusted data foundation. This was already the case with predictive AI, but the rise of generative AI — which relies on data to function — has accelerated the need for companies to modernize their data fabric. Our point of view is that there are six principles to follow for creating AI-ready data and our
product strategy for our data integration and quality portfolio fully aligns to them:

1. Data should be diverse (coming from a wide range of resources) to remove bias in AI systems

2. Data should be timely to make accurate and informed predictions

3. Data should be accurate to ensure reliability and trustworthiness in AI

4. Data should be secure to safeguard the reputation of your AI

5. Data should be discoverable to enable use of relevant and contextual data

6. Data should be consumable for ML training and LLM integration

Our Portfolio

Our data integration portfolio has always been designed to move data from any source to any target, in real time. As these destinations will often use AI on this data, this is data integration operating in the service of AI, including generative AI. Qlik’s differentiation is our ability to take the best-in-class capabilities that we are known for (real-time data integration and transformation at scale) and make
them available for generative AI use cases.

In July 2024, we launched Qlik Talend Cloud®. This new flagship offering combines the best functionality of legacy solutions Qlik Cloud® Data Integration, Talend® Cloud, and Stitch Data, and is designed to help our customers implement a trusted data foundation for AI.

Qlik Talend Cloud is built on Qlik’s cloud infrastructure platform, with the focus on managing the data integrity of our customers’ AI, analytics, and business operational projects. It offers a unified package of data integration and quality capabilities that enable data engineers and scientists to deploy AI-augmented data pipelines that deliver trusted data wherever it’s needed. This includes:

Support for vector databases and multiple LLMs that help build data pipelines to support Retrieval Augmented Generation (RAG) applications
Ability to use custom SQL to transform datasets for training machine learning models
Address the trust and compliance needs of our customers in their use of AI through data lineage, impact analysis, and the ability to assess the trustworthiness of AI datasets (providing a trust score) We have provided productivity-enhancing tools (like a co-pilot) for data engineers (prompt to SQL), with more coming later this year.

We have provided productivity-enhancing tools (like a co-pilot) for data engineers (prompt to SQL), with more coming later this year.

What’s Next

For the latter part of 2024, we plan to introduce a range of dedicated components to support RAG implementations with the leading vector databases, embedding models, and LLMs. This will offer data engineers implementing AI workloads the same reliability and scalability they expect when operationalizing all their other workloads.

Looking ahead, our 2025 plan includes further enhancements through generative AI to further improve data engineer productivity, including data pipeline design tasks, dataset auto-classifications, automated workflows, and AI-assisted record deduplication.

WHO’S IT FOR

Data Engineers and Data Architects

These professionals need to ensure that data that will be used for downstream AI processes is of high quality and trustworthy. They also want to be able to deliver that data throughout their organization using AI-augmented, no-code pipelines.

Pillar 2

AI-Powered Analytics

Enriching analytical applications and workflows with AI-powered capabilities promotes enhanced, data-centric decision making and accelerates insights. While there has been much hype around generative AI over the last year, our point of view is that it isn’t the solution to everything. Instead, we believe that both predictive AI (i.e. traditional AI), which processes and returns expected results such as analyses and predictions, and generative AI, which produces newly synthesized content based on training from existing data, hold huge potential.

Therefore, our product strategy for our analytics portfolio encompasses both predictive and generative AI.

Our Portfolio

AI has always been foundational to Qlik Cloud Analytics, our flagship analytics offering. From analytics creation and data prep to data exploration — with natural language search, conversational analytics, and natural language generation — Qlik Cloud® is designed to enhance everything users do with AI.

Today, we offer a full range of AI-powered, augmented analytics capabilities that deepen insight, broaden access, and drive efficiency. This includes:

Automated insights: auto-generate a broad range of analyses in a few clicks
Natural language analytics (Insight Advisor): get answers to questions with relevant text and visualizations in ten languages
Proactive insights: proactively notifies users when AI detects important changes

What’s Next

Our product roadmap for Analytics AI is about enhancing outcomes through automation and integrated intelligence, spanning the following tenets of AI-powered analytics:

AI-assisted analytics, which provide improved ways to author and engage with business-ready content such as sheets, analysis types, reports, etc.
Generating and communicating insights, which provide a range of diagnostic, predictive, and prescriptive insights automatically through annotations
Natural language assistance, which provide users assistance to engage with their data, platform, and operations through natural language

WHO’S IT FOR

Application Creators and Users

These professionals are looking to build and use AI-infused applications for more powerful data analysis to support decision making — and do it in a way that is intelligent, automated, embedded, and intuitive (hence easier to adopt).

Pillar 3

AI Deployment (Self-Service AI)

Companies today are looking to create value with AI by building and deploying AI models. But following the hype of generative AI in 2023, this year there has been a shift in focus1 from large language models, which necessitate significant investments, to smaller models that are more cost efficient, easier, and faster to build and deploy.

Qlik’s product strategy is perfectly aligned to this shift. We offer self-service AI solutions that enable companies to deliver an AI experience for advanced, specific use cases in a way that is efficient and affordable with fast time to value.

Our Portfolio

In July 2024, we launched Qlik Answers™, a plug and-play, generative AI-powered knowledge assistant. Qlik Answers is a self-service AI solution that can operate independently from other Qlik products and is sold separately.

This tool allows organizations to deploy an AI model that can deliver answers from a variety of unstructured data sources. The ability to analyze unstructured data enables Qlik to deliver unique value to our customers, as it’s commonly believed that 80% of the world’s data is unstructured2. A study that the firm ETR conducted on our behalf in April 2024 also found that while companies understood the value potential of being able to deliver insights from unstructured data, less than one-third felt their organization was well equipped to do so.

With Qlik Answers, organizations can now take advantage of an out-of-the-box, consolidated self-service solution that allows users to get personalized, relevant answers to their questions in real time with full visibility of source materials. As with all Qlik products, our customers can also be assured that their data stays private. Moreover, with Qlik Answers, users will only have access to data that is curated for a specific
use case. With multiple, domain-specific knowledge bases being accessible to assistants, organizations stay in control of what content users can access.

To help ensure a successful implementation, our pricing and packaging for Qlik Answers includes starter services delivered by our customer success organization.

Since 2021, Qlik has been offering another self-service AI solution for predictive analytics, Qlik AutoML®. Like Qlik Answers, Qlik AutoML can

be purchased separately.

Qlik AutoML provides a guided, no-code machine learning experience that empowers analytics teams to perform predictive analytics without the support of data science teams. With AutoML, users can:

Auto-generate predictive models with unlimited tuning and refinement
Select and deploy the best-performing models based on scoring and ranking
Make predictions with full explainability

Note: While AutoML runs inside of Qlik Cloud, it can also be used independent of Qlik Cloud Analytics. We have customers who use a real-time API to return predictions back to their own systems without having to access Qlik Cloud.

Finally, Qlik also offers connectors to enable its customers to integrate third-party generative AI models in their analytics apps, load scripts, and automations. Qlik Cloud customers have the option to leverage our AI Accelerator program to integrate large language models into their applications.

What’s Next

In September 2024, we introduced new enhancements to Qlik AutoML’s capabilities,
including augmented MLOps, model optimization, and analytics views, with plans
for additional upgrades through the end of the year and into 2025. Future improvements are focused on the ability to create time-aware models and the introduction of a full, end-to-end MLOps lifecycle for models developed on the platform to ensure they can be adequately monitored and governed.

Although Qlik Answers is a new product, we’ve already augmented its knowledge base and assistant capabilities, with more enhancements planned.

WHO’S IT FOR

Decision-Makers and End Users

These professionals want to leverage AI in a self-service way to get insights and answers that will help them make the best predictions and decisions for their area(s) of responsibility.

AI Advistory and Governance

In order to continue to develop innovative AI products and capabilities — and to ensure we do so with ethical integrity — we have put in place a rich ecosystem of AI expertise to help steer our strategy and direction. Above all, we are deeply committed to the responsible development and deployment of our technology in ways that earn and maintain people’s trust.

Principles for Responsible AI

We have created a set of principles guiding the responsible development and deployment of our technology, available publicly at qlik.com/Trust/AI. These principles are:

Reliability: We design our products for high performance and availability so customers can safely and securely integrate and analyze data and use it to make informed decisions.

Customer control: We believe customers should always remain in control of their data and how their data is used so we design our products with finegrain security controls, including down to the row (data) and object level.

Transparency and explainability: We design our products to make it clear when customers engage with AI. We strive to make clear the data, analysis, limitations, and/or model used to generate AI-driven answers so our customers can make informed decisions on how they use our technology

Observability: We design our products so customers can understand lineage, access, and governance of data, analytics, and AI models used to inform answers and automate tasks.

Inclusive: We believe diversity, equity, inclusion, and belonging drive innovation and will continue to foster these beliefs through our product design and development.

Qlik has a process and staff in place to monitor for any upcoming legislation that would impact our business, such as new AI laws. As legislative changes occur, we assess these laws and adjust our AI compliance program accordingly.

Next Steps

For more information or enquiries about Qlik products and services, feel free to contact us below.

Customer Story (Data Integration) — Vale

Vale achieves yearly benefit of $600m

“Everybody’s in the same place. They can talk to each other and see the same information on different dashboards updated in near real time. That’s the kind of interaction Qlik is enabling.”

Jordana Reis, Enterprise Integration Architect, Vale S.A.

Solution Overview

Customer Name

Vale S.A.

Industry

Mining

Geography

Brazil

Function

Sales, Supply Chain Management

Business Value Driver

New Business Opportunities, Reimagined Processes

Challenges

Improve visibility across previously manual and disconnected processes
Deliver near real-time access to critical
business information
Enable staff across different functions to carry out integrated planning

Solution

Using Qlik Data Integration to handle and automate ETL processes, Vale developed the Integrated Operations Center to provide a clear overview of the supply chain.

Results

Qlik Data Integration enables low latency ETL processes and ease of use
Business benefits topped $300 million after just one month of operation
Staff can now build their own custom dashboards in minutes

An end-to-end industry giant

Brazil’s primary economic sector comprises critical industries such as agriculture, forestry and mining, all of which act as key sources of food, fuel and raw materials. Business units range in size from subsistence smallholdings to global giants with worldwide operations. And at the apex of the mining industry sits Vale S.A.

Founded 80 years ago, the Brazil-based metals and mining corporation is the world’s largest producer of iron ore and nickel. Vale is also the most valuable business in Latin America, with an estimated market value of $111 billion and rising, and a presence in 30 countries.

While mining remains the core of its business, Vale’s operations also encompass logistics, including an extensive network of railroads, ports and blending terminals, and shipping which distributes the company’s products across the world. Also supporting its operations are Vale’s own power plants and iron pelletizing facilities.

Vale’s dry bulk supply chain is also a large-scale service, and one of the biggest transport and distribution operations in Brazil. Vale owns around 490 locomotives and more than 29,500 rail freight cars, and ships much of its iron ore and pellet output from Brazil, around the African coast, to China and Malaysia, often in its own or chartered vessels, including Very Large Ore Carriers (VLOCs).

Long distances and complex processes

Managing Vale’s global operation involves a series of complex and resource-intensive distribution processes. These were placed into sharp focus in 2015 when the business faced falling commodities prices and an increasingly competitive market.

“The geographic distances we cover, from the extraction of iron ore to delivery to customers, are very long,” says Jordana Reis, Enterprise Integration Architect at Vale. “That becomes an even bigger issue when our main competitors are closer to our buyers than we are.”

Vale’s operations were managed by a series of manual and largely disconnected processes, with different departments handling their own functions and using their own methodologies, often with legacy systems. “There were people looking at the mining aspect, people looking at ports, people looking at sales, but we didn’t have an integrated view of these operations,” explains
Richardson Nascimento, Data and AI Architect at Vale. “That was the process we needed to fix.”

This lack of an integrated view of the business was causing a range of challenges, including mismatches between production and transport capacity, logistical inefficiency and product quality management issues. “We were also missing out on valuable sales opportunities, simply because we didn’t know if we could fulfill them,” recalls Reis.

New ETL processes accelerate insight

Vale developed the Centro de Operações Integradas (Integrated Operations Center, or COI) as an operating model. One of its pillars is to provide a means of aggregating and processing the vast amounts of data it was generating but only partially using. The COI would then act as a central framework, updated in near real time, on which Vale could base decisions, better manage its production and supply chain and support its people and processes.

“When we realized how much data we would need to move to really enable COI, we started thinking about how we could automate the process,” says Nascimento. “The main driver was low latency replication. We had a target to move all this information in less than 15 minutes, and Qlik Data Integration was clearly the best option.”

Vale collaborated closely with both Microsoft and Qlik teams during the purchase process. “Both teams were very active and interested in making COI happen,” says Reis. “They gave us honest opinions and helped us to achieve our goals.”

COI uses Qlik Replicate IaaS with Microsoft Azure in tandem with a range of data repositories such as Azure SQL Database and Azure Synapse, with Qlik Replicate acting as the principal enabler of the process. Another key factor in the choice of Qlik Data Integration was agentless operation, and its efficiency in reading application databases and transaction logs without impacting their activity.

COI’s main data sources are Vale’s in-house Manufacturing Execution Systems (MES), responsible for each stage of the value chain (Mining, Rail, Ports and Pelletizing), all based on Oracle databases; the chartering system Softmar and VesselOps,
based on SQL Server; and Vale’s in-house value chain optimization systems, also based on Oracle databases.

Nascimento also points to Qlik Data Integration’s importance in supporting tools such as Azure Databricks as part of Vale’s strategy to use machine learning and artificial intelligence to augment human decisions. Vale is using several tools for big data processing, such as Azure Machine Learning. “That’s one of the tools that we’re trying to leverage more,” he notes. “Azure Machine Learning is simple to use and easy to teach.”

Importantly, Reis highlights Qlik’s ease of use and speed of implementation and operation. “It changed our extract, transform and load (ETL) process and how we make data available,” she notes. “We reduced the effort to make data available to build less complex dashboards, for instance, from four weeks to just four hours.”

Velocity and visibility of information

COI began to deliver benefits almost immediately on its launch in 2017. It enabled a new integrated planning process, giving staff across the business full visibility into the supply chain improving the ability to manage their respective operations
in a collaborative environment.

“Everything related to operations is now under COI’s umbrella,” says Nascimento. “It covers the mines, the ports, railroads, shipping and sales and freight negotiations. COI enables planning and optimization across the supply chain.”

Users can now define and build their own dashboards, while corporate dashboards also enable insights and support decisions at board level. COI’s value is neatly encapsulated in Vale’s videowalls, giant room-sized panels featuring custom dashboards that enable cross-functional collaboration. “Everybody’s in the same place,” says Reis.

“They can talk to each other and see the same information on different dashboards updated in near real time. That’s the kind of interaction Qlik is enabling.”

Nascimento also highlights Vale’s asset monitoring center, which uses a similar and connected operating model to COI that combines with other tools to provide insights into asset lifecycles, enabling preventive maintenance and extending the efficiency and working lives of machinery, plant, vehicles and more.

“It’s not just about the speed of the decisions, but that we can make different types of decisions,” Nascimento explains. “We can now adjust production in line with logistical capacities, for example. And that’s transformational.”

Multi-million dollar savings

The initial launch of COI in 2017 delivered staggering results almost immediately, enabling business benefits in terms of sales won, costs saved and efficiencies gained totaling $300 million after just one month of operation and $600 million annual savings.

This, however, is just the start. COI is what Reis describes as “a lighthouse project”, with the data architecture implemented by the Integrated Operations Center and enabled by Qlik now used across multiple other projects covering areas such as safety,
geotechnical methods and autonomous machinery.

“Our long-term strategy is based on Qlik and Microsoft Azure. Once we saw the benefits on COI, we set Qlik Data Integration as our target information integration architecture for the whole enterprise,” concludes Reis. “We also have a program to migrate as many systems as possible to Microsoft Azure, including our data repositories for analytics. And of course, we will use Qlik Data Integration and Qlik Compose there too.”

Next Steps

For more information or enquiries about Qlik products and services, feel free to contact us below.

Speed Your Data Lake ROI

Five Principles for Effectively Managing Your Data Lake Pipeline

Introduction

Being able to analyze high-volume, varied datasets is essential in nearly all industries. From fraud detection and real-time customer offers to market trend and pricing analysis, analytics use cases are boosting competitive advantage. In addition, the advent of the Internet of Things (IoT) and Artificial Intelligence (AI) are also driving up the volume and variety of data that organizations like yours want and need to analyze. The challenge: as the speed of business accelerates, data has increasingly perishable value. The solution: real-time data analysis.

Data lakes have emerged as an efficient and scalable platform for IT organizations to harness all types of data and enable analytics for data
scientists, analysts, and decision makers. But challenges remain. It’s been too hard to realize the expected returns on data lake investments, due to several key challenges in the data integration process ranging from traditional processes that are unable to adapt to changing platforms and data transfer bottlenecks to cumbersome manual scripting, lack of scalability, and the inability to quickly and easily extract source data.

Qlik®, which includes the former Attunity data integration portfolio, helps your enterprise overcome these obstacles with fully automated, high-performance, scalable, and universal data integration software.

Evolution of the Data Lake

Combining efficient distributed processing with cost-effective storage for mixed data sets analysis forever redefined the economics and possibilities of analytics. Data lakes were initially built on three pillars: the Hadoop foundation of MapReduce batch processing, the Hadoop Distributed File System (HDFS), and a “schema on read” approach that does not structure data until it’s analyzed. These pillars are evolving:

The Apache ecosystem now includes new real-time processing engines such as Spark to complement MapReduce.
The cloud is fast becoming the preferred platform for data lakes. For example, the Amazon S3 distributed object-based file store is being widely adopted as a more elastic, manageable, and cost-effective alternative to HDFS. It integrates with most other components of the Apache Hadoop stack, including MapReduce and Spark. The Azure Data Lake Store (ADLS) is also gaining traction as a cloud- based data lake option based on HDFS.
Enterprises are adopting SQL-like technologies on top of data stores to support historical or near- real time analytics. This replaces the initial “schema on read” approach of Hadoop with the “schema on write” approach typically applied to traditional data warehouses.

While the pillars are evolving, the fundamental premise of the data lake remains the same:

organizations can benefit from collecting, managing, and analyzing multi-sourced data on distributed commodity storage and processing resources.

Requirements and Challenges

As deployments proceed at enterprises across the globe, IT departments face consistent challenges when it comes to data integration. According to the TDWI survey (Data Lakes: Purposes, Practices, Patterns and Platforms), close to one third (32%) of respondents were concerned about their lack of data integration tools and related Hadoop programming skills.

Traditional data integration software tools are challenging, too, because they were designed last century for databases and data warehouses. They weren’t architected to meet the high-volume, real-time ingestion requirements of data lake, streaming, and cloud platforms. Many of these tools also use intrusive replication methods to capture transactional data, impacting production source workloads.

Often, these limitations lead to rollouts being delayed and analysts forced to work with stale and/or insufficient datasets. Organizations struggle to realize a return on their data lake investment. Join the most successful IT organizations in addressing these common data lake challenges by adopting the following five core architectural principles.

Five Principles of Data Lake Pipeline Management

1. Plan on Changing Plans

Your architecture, which likely will include more than one data lake, must adapt to changing requirements. For example, a data lake might start out on premises and then be moved to the cloud or a hybrid environment. Alternatively, the data lake might need to run on Amazon Web Services, Microsoft Azure, or Google platforms to complement on-premises components.

To best handle constantly changing architectural options, you and your IT staff need platform flexibility. You need to be able to change sources and targets without a major retrofit of replication processes.

Qlik Replicate™ (formerly Attunity Replicate) meets these requirements with a 100% automated process for ingesting data from any major source (e.g., database, data warehouse, legacy/mainframe, etc.) into any major data lake based on HDFS or S3. Your DBAs and data architects can easily configure, manage, and monitor bulk or real-time data flows across all these environments.

You and your team also can publish live database transactions to messaging platforms such as Kafka, which often serves as a channel to data lakes and other Big Data targets. Whatever your source or target, our Qlik Replicate solution provides the same drag-and-drop configuration
process for data movement, with no need for ETL programming expertise.

Two Potential Data Pipelines — One CDC Solution

2. Architect for Data in Motion

For data lakes to support real- time analytics, your data ingestion capability must be designed to recognize different data types and multiple service-level agreements (SLAs). Some data might only require batch or microbatch processing, while other data requires stream processing tools or frameworks (i.e., to analyze data in motion). To support the complete range, your system must be designed to support technologies such as Apache Kafka, Amazon Kinesis, Azure Event Hubs, and Google Cloud Pub/Sub as needed.

Additionally, you’ll need a system that ensures all replicated data can be moved securely, especially when sensitive data is being moved to a cloud-based data lake. Robust encryption and security controls are critical to meet regulatory compliance, company policy, and end-user
security requirements.

Qlik Replicate CDC technology non-disruptively copies source transactions and sends them at near-zero latency to any of the real- time/messaging platforms listed above. Using log reader technology, it copies source updates from database transaction logs – minimizing impact on production workloads – and publishes them as a continuous message stream. Source DDL/schema changes are injected into this stream to ensure analytics workloads are fully aligned with source structures. Authorized people also can transfer data securely and at high speed across the wide-area network (WAN) to cloud-based data lakes, leveraging AES-256 encryption and dynamic multipathing.

As an example, a US private equity and venture capital firm built a data lake to consolidate and analyze operational metrics from its portfolio companies. This firm opted to host its data lake in the Microsoft Azure cloud rather than taking on the administrative burden of an on-premises infrastructure. Qlik Replicate CDC captures updates and DDL changes from source databases (Oracle, SQL Server, MySQL, and DB2) at four locations in the US. Qlik Replicate then sends that data through an encrypted File Channel connection over a WAN to a virtual machine–based instance of Qlik Replicate in Azure cloud.

This Qlik Replicate instance publishes the data updates to a Kafka message broker that relays those messages in the JSON format to Spark. The Spark platform prepares the data in microbatches to be consumed by the HDInsight data lake, SQL data warehouse, and various other
internal and external subscribers. These targets subscribe to topics that are categorized by source tables. With the CDC-based architecture, this firm is now efficiently supporting real-time analysis without affecting production operations.

3. Architect for Data in Motion

Your data lake runs the risk of becoming a muddy swamp if there is no easy way for your users to access and analyze its contents. Applying technologies like Hive on top of Hadoop helps to provide an SQL-like query language supported by virtually all analytics tools. Organizations like yours often need both an operational data store (ODS) for up-to-date business intelligence (BI) and reporting as well as a comprehensive historical data store (HDS) for advanced analytics. This requires thinking about the best approach to building and managing these stores to deliver the agility the business needs.

This is more easily said than done. Once data is ingested and landed in Hadoop, often IT still struggles to create usable analytics data stores. Traditional methods require Hadoop-savvy ETL programmers to manually code the various steps – including data transformation, the creation of Hive SQL structures, and reconciliation of data insertions, updates, and deletions to avoid locking and disrupting users. The administrative burden of ensuring data is accurate and consistent can delay and even kill analytics projects.

Qlik Compose™ for Data Lakes (formerly Attunity Compose for Data Lakes) solves these problems by automating the creation and loading of Hadoop data structures, as well as updating and transforming enterprise data within the data store. You, your architects, or DBAs can automate the pipeline of BI ready data into Hadoop, creating both an ODS and HDS. Because our solution leverages the latest innovations in Hadoop such as the new ACID Merge SQL capabilities, available today in Apache Hive you can automatically and efficiently process data insertions, updates, and deletions. Qlik Replicate integrates with Qlik Compose for Data Lakes to simplify and accelerate your data ingestion, data landing, SQL schema creation, data transformation, and ODS and HDS creation/updates.

As an example of effective data structuring, Qlik works with a major provider of services to the automotive industry to more efficiently feed and transform data in a multi-zone data lake pipeline. The firm’s data is extracted from DB2 iSeries and then landed as raw deltas in an Amazon S3-based data lake. In the next S3 zone, tables are assembled (i.e., cleansed and merged) with a full persisted history available to identify potential errors and/or rewind, if necessary. Next these tables are provisioned/presented via point-in-time snapshots, ODS, and comprehensive change histories. Finally, analysts consume the data through an Amazon Redshift data warehouse. In this case, the data lake pipeline transformed the data while structured data warehouses perform the actual analysis. The firm is automating each step in the process.

A key takeaway here is that the most successful enterprises automate the deployment and continuous updates of multiple data zones to reduce time, labor, and costs. Consider the skill sets of your IT team, estimate the resources required, and develop a plan to either fully staff your project or use a technology that can reduce anticipated skill and resource requirements without compromising your ability to deliver.

Automating the Data Lake Pipeline

4. Architect for Data in Motion

Your data management processes should minimize production impact and increase efficiency as your data volumes and supporting infrastructure grow. Quantities of hundreds or thousands of data sources affect implementation time, development resources, ingestion patterns (e.g., full data sets versus incremental updates), the IT environment, maintainability, operations, management, governance, and control.

Here again organizations find automation reduces time and staff requirements, enabling staff to efficiently manage ever- growing environments. Best practices include implementing an efficient ingestion process, eliminating the need for software agents on each source system, and centralizing management of sources, targets, and replication tasks across the enterprise.

With Qlik Replicate, your organization can scale to efficiently manage data flows across the world’s largest enterprise environments. Our zero-footprint architecture eliminates the need to install, manage, and update disruptive agents on sources or targets. In addition, Qlik Enterprise
Manager™ (formerly Attunity Enterprise Manager) is an intuitive and fully automated, single console to configure, execute, monitor, and optimize thousands of replication tasks across hundreds of end points. You can track key performance indicators (KPIs) in real time and over
time to troubleshoot issues, smooth performance, and plan the capacity of Qlik Replicate servers. The result: the highest levels of efficiency and scale.

5. Depth matters

Whenever possible, your organization should consider adopting specialized technologies to integrate data from mainframe, SAP, cloud, and other complex environments. Here’s why:

Enabling analytics with SAP-sourced data on external platforms requires decoding data from SAP pooled and clustered tables and enabling business use on a common data model. Cloud migrations require advanced performance and data encryption over WANs.

And deep integration with mainframe sources is needed to offload data and queries with sufficient performance. Data architects have to take these and other platform complexities into account when planning data lake integration projects.

Qlik Replicate provides comprehensive and deep integration with all traditional and legacy production systems, including Oracle, SAP, DB2 z/ OS, DB2 iSeries, IMS, and VSAM. Our company has invested decades of engineering to be able to easily and non-disruptively extract and decode transactional data, either in bulk or real time, for analytics on any major external platform.

When decision makers at an international food industry leader needed a current view and continuous integration of production-capacity data, customer orders, and purchase orders to efficiently process, distribute, and sell tens of millions of chickens each week, they turned to Qlik. The company had struggled to bring together its large datasets, which were distributed across several acquisition-related silos within SAP Enterprise Resource Planning (ERP) applications. The company relied on slow data extraction and decoding processes that were unable to match orders and production line-item data fast enough, snarling plant operational scheduling and preventing sales teams from filing accurate daily reports.

The global food company converted to a new Hadoop Data Lake based on the Hortonworks Data Platform and Qlik Replicate. It now uses our SAP-certified software to efficiently copy SAP record changes every five seconds, decoding that data from complex source SAP pool and cluster tables. Qlik Replicate injects this data stream – along with any changes to the source metadata and DDL changes – to a Kafka message queue that feeds HDFS and HBase consumers subscribing to the relevant message topics (one topic per source table).

Once the data arrives in HDFS and HBase, Spark in-memory processing helps match orders to production on a real-time basis and maintain referential integrity for purchase order tables within HBase and Hive. The company has accelerated sales and product delivery with accurate real-time operational reporting. Now, it operates more efficiently and more profitably because it unlocked data from complex SAP source structures.

Streaming Data to a Cloud-based Data Lake and Data Warehouse

How Qlik Automates the Data Lake Pipeline

By adhering to these five principles, your enterprise IT organization can strategically build an architecture on premises or in the cloud to meet historical and real-time analytics requirements. Our solution, which includes Qlik Replicate and Qlik Compose for Data Lakes, addresses key challenges and moves you closer to achieving your business objectives.

The featured case studies and this sample architecture and description show how Qlik manages data flows at each stage of a data lake pipeline.

Your Data Lake Pipeline

Take a closer look, starting with the Landing Zone. First, Qlik Replicate copies data – often from traditional sources such as Oracle, SAP, and mainframe – then lands it in raw form in the Hadoop File System. This process illuminates all the advantages of Qlik Replicate, including full load/CDC capabilities, time-based partitioning for transactional consistency and auto-propagation of source DDL changes. Now, data is
ingested and available as full snapshots or change tables, but not yet ready for analytics.

In the Assemble Zone, Qlik Compose for Data Lakes standardizes and combines change streams into a single transformation-ready data store. It automatically merges the multi-table and/or multi-sourced data into a flexible format and structure, retaining full history to rewind and identify/remediate bugs, if needed. The resulting persisted history provides consumers with rapid access to trusted data, without having to understand or execute the structuring that has taken place. Meanwhile, you, your data managers, and architects maintain central control of the entire process.

In the Provision Zone, your data managers and architects provision an enriched data subset to a target, potentially a structured data warehouse, for consumption (curation, preparation, visualization, modeling, and analytics) by your data scientists and analysts. Data can be continuously updated to these targets to maintain fresh data.

Our Qlik software also provides automated metadata management capabilities to help your enterprise users better understand, utilize, and trust their data as it flows into and is transformed within their data lake pipeline. With our Qlik Replicate and Qlik Compose solutions you can add, view, and edit entities (e.g., tables) and attributes (i.e., columns). Qlik Enterprise Manager centralizes all this technical metadata so anyone can track the lineage of any piece of data from source to target, and assess the potential impact of table/column changes across data zones. In addition, Qlik Enterprise Manager collects and shares operational metadata from Qlik Replicate with third-party reporting tools for enterprise-wide discovery and reporting. And our company continues to enrich our metadata management capabilities and contribute to open-source industry initiatives such as ODPi to help simplify and standardize Big Data ecosystems with common reference specifications.

Conclusion

You improve the odds of data lake success by planning and designing for platform flexibility, data in motion, automation, scalability, and deep source integration. Most important, each of these principles hinge on effective data integration capabilities.

Our Qlik technology portfolio accelerates and automates data flows across your data lake pipeline, reducing your time to analytics readiness. It provides efficient and automated management of data flows and metadata. Using our software, you and your organization can improve SLAs, eliminate data and resource bottlenecks, and more efficiently manage higher-scale data lake initiatives. Get your analytics project back on track and help your business realize more value faster from your data with Qlik.

Next Steps

For more information or enquiries about Qlik products and services, feel free to contact us below.

Data Drives Business

Data Integration Considerations for ISVs and Data Providers

Real-Time Data and AI Drive Businesses Today

Data is an extremely valuable asset to almost every organization, and it informs nearly every decision an enterprise makes. It can be used to make better decisions at almost every level of the enterprise—and to make them more quickly. But to take full advantage of the data and to
do so quickly requires artificial intelligence (AI). So, it is no surprise that nearly all participants in our research (87%) report that they have enabled or piloted AI features in analytics and business intelligence applications. Today, data is collected in more ways and from more
devices and more frequently than ever before. It can enable new methods of doing business and can even create new sources of revenue. In fact, the data and analyses themselves can be a new source of revenue.

Independent software vendors (ISVs) and data providers understand the importance of data in AI-based processes, and they are designing products and services to help enterprises step in and harness all this data and AI-generated business energy. To maximize the opportunities,
ISVs and data providers need to recognize that enterprises use various types of data, including data from both internal and external
sources. In fact, our research shows that the majority of enterprises (56%) are working with 11 or more sources of data. Governing the various data sources becomes critical because poor quality data leads to poor AI models. Our research shows the top benefit of investing in data governance, reported by three-quarters of participants (77%), is improved data quality.

Real-Time Data and AI Drive Businesses Today

The most common types of collected data include transactional, financial, customer, IT systems, employee, call center, and supply
chain. But there are other sources as well, many external to the enterprise. Nine in 10 enterprises (90%) are working with at least one
source of external data, which could mean location data, economic data, social media, market data, consumer demographics government data, and weather data. To be useful, all of that must be integrated.

“Data integration” is the process of bringing together information from various sources across an enterprise to provide a complete, accurate, and real-time set of data that can support
operational processes and decision-making. But nearly one-third of enterprises (31%) report that it is hard to access their data sources, and more than two-thirds (69%) report that preparing their data is the activity where they spend the most time in their analytics
processes. The process of data integration often places a burden on the operational systems upon which enterprises rely.

At the same time, enterprises also need to be able to integrate applications into their data processes. ISVs and data providers must bring data together with applications so it is easier for enterprises to access and use the very data they provide.

Data Integration Is Not Easy

Simple linkages such as open database connectivity and Java database connectivity (ODBC/JDBC), or even custom-coded scripts, are not sufficient for data integration. While ODBC/JDBC can provide the necessary “plumbing” to access many different data sources, it offers little assistance to application developers in creating agile data pipelines. Simple connectivity also does nothing to assist with consolidating or transforming data to make it ready for analytics, for instance, in a star schema. Nor does simple connectivity provide any assistance in dealing with slowly changing dimensions which must be tracked for many types of AI analyses.

Simple connectivity does little to help enterprises transform the data to ensure its standardization or quality. Data from various sources often contains inconsistencies, for instance in customer reference numbers or product codes. Accurate analyses require that these inconsistencies be resolved as the data is integrated. Similarly, data quality is an issue that must be addressed as the data is integrated. Our research shows these two issues of data quality and consistency are the second most common time sinks in the analytics process.

Nor does simple database connectivity help enterprises effectively integrate data from files, applications or application programming interfaces (APIs). With the proliferation of cloudbased applications, many of which only provide API access, ODBC/JDBC connectivity may not be an option. And many enterprises still need to process flat files of data, as our research shows that these types of files are the second most common source of data for analytics.

Data integration is not a one-time activity, either. It requires the establishment of data pipelines that regularly collect and consolidate
updated data. A greater infrastructure is needed around these pipelines to ensure that they run properly and to completion. ISVs and data providers that rely only on simple connectors must create and maintain this extra infrastructure themselves.

Those data pipelines also need to be agile enough to support a variety of styles of integration. Batch updates are still useful for bulk transfers of data, but other more frequent styles of updating are needed as well. Our research shows that nearly one-quarter of enterprises (22%) need to analyze data in real time. Since the most common sources of information are transactional and operational applications, it is important to create pipelines that can access this data as it is generated. Incremental updates and change data capture (CDC) technology can solve this problem and these are becoming competitive necessities.

Real-time requirements are even more demanding when we consider event data, where nearly one-half (47%) of enterprises process it within seconds. Then, as applications and organizational requirements change, the data pipelines must reflect those changes. Therefore, the tools used to support such a wide variety of ever-changing sources need to be open enough to be easily incorporated into a wide variety of processes.

But if ISVs and data providers focus their energies on maintaining data pipelines, it distracts resources from the core business. Creating data pipeline infrastructure that is highly performant and efficient requires years of engineering. Simple bulk movement of entire data sets is slow and inefficient, even though it may be necessary for initial data transfers. Subsequent data transfers, however, should use a data replication scheme or CDC approach, creating much smaller data transfers and much faster processes.

Advantages of a Modern Data Fabric

A modern data fabric is based on a cloud-native architecture and includes orchestration and automation capabilities that enhance the design and execution of data pipelines that consolidate information from across the enterprise. As data becomes a new source of revenue, sometimes referred to as “data as a product,” a modern data fabric must also enable easy access to, and consumption of, data. A key component to delivering data in this fashion is strong data catalog capabilities. AI assisted search, automated profiling and tagging of data sources, and tracking the lineage of that data through its entire life cycle make it easier to find and understand the data needed for particular operations and analyses. Collecting and sharing this metadata in a data catalog not only provides better understanding and access to the data, but also improves data governance. Our research shows that enterprises that have adequate data catalog technology are three times more likely to be satisfied with their analytics and have achieved greater rates of self-service analytics.

Orchestration and access via APIs are also critical to ISVs and data providers as these allow the remote invocation of data pipelines needed for the coordination and synchronization of various interrelated application processes, even when they are distributed across different cloud applications and services. These APIs need to span all aspects from provisioning to core functionality for orchestration to be effective. Automation of these orchestration tasks can enhance many aspects of data pipelines to make them both more efficient and more agile.
Automated data mapping, automated meta data creation and management, schema evolution, automated data mart creation, and data warehouse and data lake automation can quickly and efficiently create analytics-ready data. When combined with orchestration, automation can also provide “reverse integration” to update data in source systems when necessary and appropriate.

Modern data integration platforms employ AI/ML to streamline and improve data processing. AI/ML can be used to automatically detect anomalies in data pipelines, such as whether the pipelines suddenly processed an unusually small number of records. Such an anomaly could indicate a problem somewhere else in the pipeline. AI/ML can also be used to automatically deal with errors in pipelines and routine changes, such as those in the sources or targets. AI/ML can also determine the optimal execution of pipelines, including the number of instances to create or where different portions of the pipeline should be processed. AI/ML can be used to enrich data with predictions, scoring or classifications that help support more accurate decision-making. We assert that by 2027, three-quarters of all data processes will use AI and ML to accelerate the realization
of value from the data.

Modern data integration platforms must also incorporate all appropriate capabilities for data governance. Data sovereignty issues may require that data pipelines be executed only within certain geographies. Compliance with internal or regulatory policies may require single sign-on or the use of additional credentials to appropriately track and govern data access and use. Therefore, a platform with built-in capabilities for governance can help identify personally identifiable information and other sensitive or regulated data. But implementing any of these modern data integration platform requirements can impose a significant burden on ISVs and data providers.

Illustrative Use Cases

Product Distributors

For organizations with hundreds of thousands of SKUs and hundreds of thousands of customers, managing orders and inventories can be a time consuming process. Using a modern data-as-a-product approach with standardized data governance and a centralized data catalog can reduce costs dramatically and enable self-service online ordering. This approach also creates more agility to meet customer needs and provides better, more timely visibility into operations.

Insurance Industry

Insurance technology data providers can use data integration to help their customers be more competitive by providing access to up-to-date information that enables online quotes. Data is the key to the accurate pricing of insurance liabilities, and many of the sources and targets exist in the cloud, but they require support for a variety of endpoints. By using CDC-based replication, however, both claims and market data can be collected, consolidated, and distributed within minutes. As a result, millions of quotes can be generated each day where each incorporates real-time analysis of vast volumes of data.

Other Applications

Data integration can be the key to many other ISVs and data providers. Mobile application providers can integrate location data with transaction data to provide broader market data on consumer behavior. Talent management ISVs can integrate data relating to internal performance and compensation with external market data to improve employee acquisition and retention. Foreclosure data can be collected, consolidated, and distributed to support loan origination and servicing operations. Vendor data can be collected and provided to improve procurement processes augmenting supplier performance analyses with risk, diversity, sustainability and credit scores. And regardless of the vertical industry or line-of-business function, faster access to more data generally produces better results.

Other Considerations

Once data is integrated, it can provide the basis for a broad range of analytics and AI. By supporting these analyses and data science, ISVs and data providers can extend the value of their capabilities and therefore increase their revenue opportunities. Choosing a data integration platform that also supports analytics and AI will make it easier for enterprises to capture this revenue. In fact, our research shows that reports and dashboards are the most common types of analytics used by more than 80% of enterprises. However, when considering analytics providers, look at those that support other newer techniques as well, such as AI/ML and natural language processing, which are projected to be required by 80% of enterprises in the future.

Enterprises need to use data to help drive actions. Data can help them understand what has happened and why, but they ultimately need to process what they have learned and then take action. In many situations, however, there is simply no time to review data to determine what
course of action to take. ISVs and data providers can help their customers derive more value from data by using real-time information to trigger the appropriate actions.

ISVs and data providers are using technology to add value to business processes. While all business processes typically require data, data integration itself is merely a means to the end. If the process is not done properly, it can detract from the overall approach, so it requires careful design and development. Enterprises should ideally spend their time on core competencies, not on developing data integration technology. By using a full-featured, purpose-built data integration platform, they can ensure that the data needed by ISVs and data providers is robust and available in a timely manner.

Next Steps

Explore all available data sources, along with their accessibility, that can boost the value of your services.
Recognize the value of data catalog and data governance in enabling data-as-a-product.
Consider platforms that go beyond simple connections to data sources and that minimize the amount of development and maintenance work required.
To maximize performance and minimize the impact on production systems, create repeatable and agile pipelines that operate efficiently.
Look for platforms with significant automation capabilities to maximize productivity and responsiveness.
Ensure that your architecture provides a modern, cloud-native approach.

Move From Worksheets to Workflows with Automation

How 5 Companies Automate Data Prep and Analytics

For business analysts who spend their days mired in tedious data tasks, the struggle is real. Manually cleansing, blending and analyzing a growing volume of complex data is taking more time than ever. While spreadsheets are useful for basic tasks, when data gets messy or large, they quickly become slow and cumbersome to work with — not to mention prone to errors. That’s why more organizations are arming their analysts with easy-to-use automated solutions that enable them to deliver fast, accurate, data-driven insights and eliminate the burden of time consuming manual processes.

In the following real-world customer stories, you’ll learn how analysts and business professionals use Alteryx to simplify and automate complex analytical processes with an intuitive drag and-drop interface and built-in AI-guidance. Read how they are saving time, achieving bottom line results, and adding more value using their business expertise combined with advanced analysis. Plus, learn about an opportunity to simulate AI-powered solutions with your own use case.

DoorDash is the largest food delivery service in the U.S., supporting hundreds of thousands of merchants and millions of customers in more than 500 cities across North America.

Business Challenge

The accounting team at DoorDash was dealing with a growing volume of complex data with mounting pressure to speed up processes. They also had to meet the rigorous standards of SOX compliance.

Analysts relied on manual processes and spreadsheets to collect, reconcile, and analyze massive amounts of data — a time-consuming process that was prone to errors.

Alteryx Solution

DoorDash uses Alteryx to automate and streamline operational processes, data acquisition, and in-depth financial analysis. By replacing manual processes with easy to-build automated workflows, the finance team now saves 25,000 hours annually.

The end-to-end automation solution also frees up financial analysts to focus on value-added, strategic tasks that drive more accurate accounting.

Results

Mayborn Group reduces manual processes by 90% and optimizes product promotion offers using Alteryx.

Baker Tilly is a top ten accounting firm that offers specialized federal tax compliance and planning expertise to help businesses optimize value while
minimizing their tax burden.

Business Challenge

The tax team responsible for unclaimed property reporting had to collect and process hundreds of thousands of files with up to a million lines or records of data — all coming from multiple disparate sources.

The process of collecting, cleaning, and analyzing the massive files took anywhere from several days to weeks to complete. There was also a risk of exposure for the client if any amount of unclaimed property was missed due to human error.

Alteryx Solution

Baker Tilly now uses Alteryx to automate data prep, processing, and reporting. Non-technical teams in the unclaimed property department were able to build their own workflows to consolidate files and apply analytics, reducing the time spent preparing deliverables by 50%.

By upskilling domain subject matter experts in easy-to-use analytics, they
eliminated the need to rely on technical experts for dashboard and report building.

Results

Automating data prep and analytics with Alteryx reduced reporting processing time by 50% and decreased regulatory errors by 70%.

Bank of America a multinational investment bank and financial services company serving approximately 56 million U.S. consumer and small business relationships.

Business Challenge

The enterprise testing team at Bank of America must ensure that all regulators are notified of any applicable transactions. The team was manually prepping and cleansing tens of millions of transactions for quality assurance every day.

The entire process, from the time of the transactions to the moment the
regulators were notified, took about two months. The delayed response left the organization susceptible to costly regulatory fines.

Alteryx Solution

Bank of America added Alteryx to its data stack of Tableau, Qlik, and MicroStrategy to create a streamlined workflow that alerts the testing team when they need to take corrective action on any regulatory measures.

Using Alteryx, the quality assurance process has transformed from reactive to proactive, with the ability to address issues as they occur, rather than waiting two months for the results. The testing team can also easily share the reports for regulatory transparency

Results

Automated, real-time data prep with Alteryx reduces quality assurance processing time by 60 days at Bank of America.

Located in 90 countries, Siemens Energy operates across the energy landscape, from conventional and renewable power to grid technology and electrifying complex industrial processes

Business Challenge

The transmission unit at Siemens struggled to efficiently collect and analyze production, logistics and financial data from 36 factories worldwide. Analysts were tied up in spreadsheets for hours, consolidating, validating, and prepping data.

The time-consuming manual processes prevented the organization from
achieving full value from its data and kept analysts from spending time on higher impact, strategic initiatives.

Alteryx Solution

Siemens adopted Alteryx to automate and scale large and complex data projects. With intuitive, drag-and-drop features, domain experts could build their own workflows and share insights with ease.

In less than six months, the team created 350 automated workflows in Alteryx and saved thousands of hours eliminating manual processes. Alteryx users at Siemens are also helping drive a wider culture of analytics across the organization

Results

Using Alteryx, analysts at Siemens built 350 automated processes and saved thousands of hours.

Mayborn Group is an award-winning retail brand that produces a broad range of baby products available in 60 countries worldwide.

Business Challenge

As a global brand with hundreds of products and retailers, Mayborn had data coming in from more than 100 internal and external sources. Consolidating the data for a holistic view to understand market and customer behavior at a regional and retailer level was a significant challenge.

Analysts had to manually merge and process data by individual retailers, region by region, with very specific views in isolation of one another. The analytics team found it impossible to scale the process efficiently with manual processes.

Alteryx Solution

Mayborn uses Alteryx to automate the process of blending and analyzing disparate data sets. Now, they have a centralized, global view of point-of-sale data that allows them to better focus on product quality and competitive strategies.

The time saved using Alteryx allows the analytics team to focus on strategic initiatives including product promotions. They realized significant ROI by using Alteryx to analyze and optimize promotional offers that increased sales.

Results

Mayborn Group reduces manual processes by 90% and optimizes product promotion offers using Alteryx.

Suite of Analytics Solutions for Active Ageing Centres

Embrace Transformation for Enhanced Care

“To strengthen support for seniors in the community, we will need to raise the capabilities of our health and social ecosystems. Digitalization of the Community Care sector will be a key pillar in this effort” – Mr. Ng How Yue, Permanent Secretary, Ministry of Health

Active Aging Centres (AACs) across Singapore play a crucial role in supporting our senior community. However, resource and manpower challenges have long been an issue, and these will become more pressing as Singapore’s population rapidly ages by 2030.

This is where technologies can empower AACs to become data-driven-ready, ensuring that operations and resources are always optimized. By leveraging AI and other advanced technologies, AACs can make data-driven decisions to enhance their services.

SIFT has been assisting in transforming the social service sector, and we invite AAC leaders to connect with us for ideas on how to integrate digitalization into their organizations. You will gain insights from real use cases, and we can help you develop a strategic digitalization roadmap that provides clarity and aligns with your organizational goals.

Client Challenges

SIFT Analytics Group has designed a customized analytics solution for a non-profit social service organization seeking to optimize manpower management and streamline daily activities. This solution consolidates operational data with real-time activity tracking, offering a centralized platform for efficient management of both staff and volunteers. In addition to improving operational efficiency, it monitors costs, revenue, and other key metrics, empowering the organization to make data-driven decisions that enhance overall service delivery.

Demographic Insights

Age Distribution: Analyzing the age range of participants to tailor programs accordingly (e.g., physical activities for 60-70, cognitive programs for 70+).

Health Status: Tracking common health conditions and mobility levels to design appropriate fitness, wellness, or mental health programs.

Program Participation

Attendance Tracking: Analyzing which activities (e.g., fitness classes, social events, workshops) are most popular.

Engagement Trends: Identifying times of day or year when attendance peaks, helping to plan more effective schedules.

Program Outcomes: Evaluating the effectiveness of health, social, or learning programs by measuring improvements in participants’ well-being.

Volunteer and Staff Contributions

Volunteer Hours and Activities: Tracking the number of hours worked by volunteers and the areas in which they contribute.

Staff-to-Participant Ratio: Ensuring sufficient staff or volunteer resources are available for high-demand activities.

Health & Wellness Outcomes

Physical Health Metrics: Monitoring fitness assessments, mobility, or vital health indicators (e.g., heart rate, blood pressure) before and after participation.

Mental Health & Cognitive Function: Surveying participants’ cognitive function or emotional well-being over time.

Social Impact

Social Engagement: Measuring increases in social interactions and community integration through surveys or participation in group activities.

Resource Allocation

Cost per Participant: Analyzing the cost of delivering each program or service to ensure resources are being used effectively.

Funding Sources & Impact: Connecting donor funding or grants to specific initiatives and their outcomes.

Satisfaction and Feedback

Participant Feedback: Regularly surveying participants to gauge satisfaction with programs, facilities, and services.

Program Improvement: Using feedback to make data-driven decisions on improving existing offerings or introducing new activities.

What is Your Data Strategy?

Is your data strategy ready to drive growth and innovation?

What is Your Data Strategy:

A data strategy is more than just a roadmap—it’s a strategic framework for navigating the complexities of data management, governance, and analytics. It outlines how your organization collects, stores, processes, and leverages data to achieve key business objectives.

In an era where advanced analytics and AI are reshaping industries, a robust data strategy is essential to staying ahead of the competition.

Building a Comprehensive and Advanced Data Strategy

The foundation of an advanced data strategy starts with aligning your data initiatives to business outcomes. Whether you aim to enhance customer engagement, drive innovation, or optimize efficiencies, your strategy must be tailored to those specific goals. It involves evaluating the current data landscape, identifying gaps such as data silos, poor data integration, or suboptimal quality, and establishing a clear path to address these challenges.

At the core of a high-performing data strategy are advanced components like data governance, ensuring data quality, security, and regulatory compliance, and data architecture that supports scalability, agility, and real-time analytics. The integration of cutting-edge technologies such as AI, machine learning, and predictive analytics can propel your data strategy to new heights, enabling deeper insights and more informed decision-making.

The Role of Advanced Technologies

As the demand for real-time insights grows, modern data strategies must incorporate advanced tools and technologies that facilitate data processing at scale. SIFT Analytics leverages state-of-the-art platforms and solutions, enabling organizations to integrate machine learning and AI to extract powerful insights from large datasets, predict trends, and drive data-driven innovation.

Consult SIFT for Your Data Strategy

SIFT Analytics provides end-to-end expertise in designing advanced data strategies, ensuring your data architecture is optimized for both current needs and future growth. With their scalable, flexible solutions, SIFT empowers organizations to move beyond traditional analytics and unlock the full potential of their data ecosystem.

👉 Consult SIFT

The Analytics Times

The Ultimate Guide to Embedded Analytics

Establishing a trusted data foundation for AI

Introduction

The six principles for AI-ready data

Data has to be diverse.

Data has to be timely.

Data has to be accurate.

Data has to be secure.

Data has to be discoverable.

Data has to be easily consumable by MLs or LLMs.

The AI Trust Score

Qlik Talend data foundation for AI

Conclusion

More Data-Related Topics That Might Interest You

The Analytics Times

Harness the Full Value of Your SAP Data for Cloud Analytics

Powered by Qlik and Snowflake

SAP is the Life Blood of Your Enterprise

Today's Business Needs for Analytics

Faster Time to value

Freedom

Modern Analytics

Control Costs

Fully Leveraging SAP Data is Difficult

The Fast Path to Extracting Value from SAP Data in the Cloud

The Benefits of Offloading SAP Data into Snowflake

Accelerate the Delivery of Analytics-Ready SAP Data to Snowflake

A Proven and Optimized Solution for Unlocking Insights from SAP

A Track Record of Collaboration and Success

More Data-Related Topics That Might Interest You

The Analytics Times

Artificial Intelligence: Our Strategy

How we infuse AI across our business, from our products to how we operate

Introduction

AI Foundation

AI-Powered Analytics

AI Deployment (Self-Service AI)

AI Advistory and Governance

More Data-Related Topics That Might Interest You

The Analytics Times

Customer Story (Data Integration) — Vale

Vale achieves yearly benefit of $600m

Solution Overview

An end-to-end industry giant

Long distances and complex processes

New ETL processes accelerate insight

Velocity and visibility of information

Multi-million dollar savings

More Data-Related Topics That Might Interest You

The Analytics Times

Speed Your Data Lake ROI

Five Principles for Effectively Managing Your Data Lake Pipeline

Introduction

Evolution of the Data Lake

Requirements and Challenges

Five Principles of Data Lake Pipeline Management

4. Architect for Data in Motion

5. Depth matters

How Qlik Automates the Data Lake Pipeline

Conclusion

More Data-Related Topics That Might Interest You

The Analytics Times

Data Drives Business

Data Integration Considerations for ISVs and Data Providers

Real-Time Data and AI Drive Businesses Today

Real-Time Data and AI Drive Businesses Today

Data Integration Is Not Easy

Advantages of a Modern Data Fabric

Illustrative Use Cases

Other Considerations

Next Steps

More Data-Related Topics That Might Interest You

The Analytics Times

Move From Worksheets to Workflows with Automation

How 5 Companies Automate Data Prep and Analytics

More Data-Related Topics That Might Interest You

Artificial Intelligence:
Our Strategy

How we infuse AI across our business,
from our products to how we operate