Determine When You Should Use Synthetic Data
Clarify the value that synthetic data can offer your organization and data program.
Data is the indispensable foundation of decision-making, but what can organizations do when usable data is lacking? With the aid of AI, synthetic data is fast becoming a more viable alternative. Our comprehensive primer on synthetic data helps you determine if it would be useful for your organization and how to communicate its value to decision-makers.
Synthetic data has a number of use cases, such as training AI models or testing software, where it can be a genuinely valuable, rapid, and inexpensive stand-in for real data. However, it should be relied on only when real-world data is unavailable, inadequate, or cannot be used due to confidentiality or privacy concerns. Organizations must have a clear understanding of the problem it is meant to solve and be able to directly relate it to the wider business strategy, in order to unlock its potential.
1. "Fake" data has real value.
Though created artificially rather than from real-world experience, synthetic data can provide decision-makers with actionable insights that benefit the organization as a whole. It is especially useful in cases where obtaining particular kinds of real-world data would be impractical or unethical, such as in the financial or healthcare sectors.
2. Plan to mitigate synthetic data’s risks.
Synthetic data carries its own inherent risks and sometimes falls short of representing the real world. Organizations must develop a plan to mitigate those risks and be aware of synthetic data's limits.
3. Make the business benefits obvious.
Like real data, synthetic data is not used for its own sake, but to achieve specific organizational outcomes. IT leaders must articulate the benefits of using synthetic data for particular use cases and link those benefits to overall organizational strategy to convince stakeholders of its value.
Use this step-by-step research to determine if synthetic data is right for your organization
Our research includes four-step guidance and a comprehensive template to help you decide whether synthetic data is right for you, and features highlights of an interview with NVIDIA Vice President of AI Research Sanja Fidler detailing the AI heavyweight's Cosmos platform, which creates synthetic data for robotics and self-driving cars. Use our comprehensive framework to clarify synthetic data’s specific value to your organization while outlining the business case to decision-makers.
- Articulate the business use case by engaging stakeholders, linking the use case to strategic objectives, and setting out the problem synthetic data is meant to solve.
- Identify the data gap to address by examining your data challenges, current data set, aims, and use case readiness.
- Assess your ability to execute by determining who should be involved and how, reviewing your data governance policies, and documenting your data generation plan.
- Make the case for synthetic data use, including monitored KPIs for expected benefits, and a risk monitoring plan.
On Demand
Webinar
Use Synthetic Data to Unlock New Data Value Streams
Play WebinarDetermine When You Should Use Synthetic Data
Clarify the value that synthetic data can offer your organization and data program.
Executive Summary
Your Challenge
- Your data team is working to respond to a business use case and you're trying to determine if synthetic data could provide value.
- It's not clear what use cases are viable for synthetic data solutions or when synthetic data would prove to be a better option than collecting or purchasing real data.
- The value of using synthetic data as a solution needs to be determined and explained to the business by connecting it to the strategic objectives, explaining the choices made, and detailing the operational considerations.
Common Obstacles
- Synthetic data is only viable for specific use cases given certain context.
- Synthetic data can only provide value in those use cases if specific challenges need to be addressed relating to data scarcity, privacy/security, bias, simulation, or cost considerations.
- The number of options for how to generate synthetic data can quicky add complexity to the decision that it could provide value.
- Integrating synthetic data initiatives into an operational environment involves planning with people, consideration of data governance policies, and clear communication with decision-makers.
Info-Tech's Approach
- Articulate your strategic objective for the use case. Clearly identify how the synthetic data use case relates to your data value streams within your larger data platform architecture.
- Determine if you have a suitable use case. Synthetic data is typically useful for training AI models, testing software, ensuring privacy or security while data sharing, or research.
- Determine if synthetic data can address your challenges with the use case. Synthetic data is used to close the gap between the data you have and the data required for your use case. But first you have to understand the gap that needs to be filled.
- Operationalize your initiative. Complete a RACI chart and integrate the initiative to your data governance policy.
Info-Tech Insight
While synthetic data should be viewed as something that's used only when using real data isn't possible, there are many use cases where synthetic data holds great value and should be deployed. Acknowledging the shortcomings while highlighting the expected benefits for a specific use case can help a data lead negotiate their way through a corporate governance process.
Blueprint deliverables
Key deliverable:
Determine When You Should Use Synthetic Data – Template
This template contains everything required to evaluate your synthetic data initiative and explain its value to stakeholders.
Each step of this blueprint is accompanied by supporting deliverables to help you accomplish your goals:
Evaluation Checklist – Template
Within the key deliverable, find a checklist to help you evaluate the use of synthetic data with a series of yes/no questions.
Insight summary
Synthetic data provides value in the same way real data can
Just as real data can be employed in data programs to create value connected to strategic business objectives, synthetic data can be generated to serve the same interests.
New synthetic data generation methods foster a growing market
More use cases are possible as AI models can generate synthetic data at scale with high precision and at low cost.
Synthetic data use cases address five types of core data challenge
Data challenges relating to privacy/security, scarcity, bias, simulation, and cost are the drivers behind generating synthetic data.
Synthetic data closes the gap between your real data and your use case
Synthetic data makes data initiatives possible where they weren't before by filling in the data that wasn't available previously.
Technical benefits translate into business befits
The benefits related to a synthetic data use case are often spoken about in technical, data-oriented terms, but they can be translated into business objectives that focus on expected benefits and mitigated risks.
Synthetic data should only be used when there's no alternative
Using real data is always the best option. Using synthetic data should therefore come with a plan to mitigate inherent risks and a plan to collect the real data required to replace the synthetic data if possible.
Data's stock is going up…
Over the last few years, the value of organizational data has been driven up to all-time highs as several market trends converge to make high-quality data a competitive advantage.
- AI and ML require large amounts of data for training, and even when pretrained models can be used, organizational data is beneficial for use in customization and fine-tuning.
- Business intelligence and data-driven decision-making are helping organizations optimize operations and identify new sources of revenue.
- Data privacy regulations have emerged to make it more challenging to collect and use real-world data, especially when it contains people's personal information.
- The volume and variety of data collected is growing exponentially with more means of capture and collection, as well as storage and analysis, opening up new opportunities for organizations that can manage the complexity.
"Data is the lifeblood of modern healthcare." (NPJ Digital Medicine, 2023)
"Data is vital for AI technical improvements." (HAI, 2024)
"Access to … data will be a key determinant of success for enterprises." ("Data Strategy for an AI Future," CIO, 2024)
…but organizations struggle with data gaps
Data quality sees the largest gap between perceived importance and satisfaction among business stakeholders, compared to other core IT services delivered to the business. It's IT's greatest area of underperformance in the eyes of the business.
Core challenges
- Legacy systems isolate data in silos that are hard to share.
- Privacy and security requirements further limit the amount of data that can be stored or shared within the organization.
- Targeted data collection projects can be complex, requiring time-consuming and expensive initiatives.
Average gap between importance and satisfaction
Synthetic data helps address business challenges when real-world data is lacking
Synthetic data is always created for a purpose. Our world is awash in data, and more is created every day. Yet at the same time we see a burgeoning market to create synthetic data. Why? Because specific business challenges bring specific requirements for the data needed. We may have limited access to that variety of real-world data, and creating it or recording it from the real world may be time-consuming or costly. Synthetic generation offers the chance to rapidly create the data that data scientists and analysts require to address the challenges their organizations face.
Market analysts estimate the synthetic data generation market was worth between $218 and $288 million in 2022-23 and project it to grow to between $1.8 billion and $2.4 billion by 2030, implying a compound annual growth rate of between 31-35% (Fortune Business Insights; Grandview).
Top industries using synthetic data by market share (2022)
Source: Fortune Business Insights
Data facts
- 90% of the world's data was generated in the past two years.
- It's estimated that 181 zettabytes of data will be generated in 2025 – 90 times the amount generated in 2010.
- In 2024, 403 million terabytes of data are created every day.
- Video is responsible for over half of global data traffic (54%).
- The US has more than 5,300 data centers – more than 10 times more than any other country
(Exploding Topics, 2023).
Phase 1
Determine When You Should Use Synthetic Data
Phase 1 |
|---|
|
1.1 Articulate the business use case 1.2 Identify the data gap to address 1.3 Assess ability to execute 1.4 Make the case |
This phase will walk you through the following activities:
- Articulate the business use case
- Identify the data gap
to address - Assess ability to execute
- Make the case
This phase involves the following participants:
- Data lead
- CTO or other executive supervisor
- Data scientists (optional)
Step 1.1
Articulate the business use case
Activities
- 1.1.1 Articulate the business use case
- This step involves the following participants:
- Data lead
- CTO or other executive supervisor
- Data scientists (optional)
Outcomes of this step
- Visualization use case to strategic objectives
- Problem statement
Synthetic data is used for several typical use cases
What is the synthetic data for?
|
Training AI models |
Testing Software |
Scenario Planning |
Research |
|---|---|---|---|
|
Training an AI model to make accurate predictions, from cognitive vision to statistical analysis to large language models. |
Testing software for its performance in edge case scenarios, or at greater scale and volume. Often to test if software is ready to exit the development environment and enter the operating environment. |
Simulation of diverse scenarios including rare conditions or conditions that have not yet been encountered. Sometimes accomplished with the creation of digital twins. |
Data sharing for research purposes that involves third parties that could provide insights in terms of business analytics or for academic consideration. Sharing can be facilitated through payment, creating monetization opportunity. |
On Demand
Webinar
Use Synthetic Data to Unlock New Data Value Streams
Play Webinar
Brian Jackson
Principal Research Director
Steve Willis
Practice Research Director, Data, Analytics, Enterprise Architecture & AI
Ryan Brunet
Principal Research Director
About Info-Tech
Info-Tech Research Group is the world’s fastest-growing information technology research and advisory company, proudly serving over 30,000 IT professionals.
We produce unbiased and highly relevant research to help CIOs and IT leaders make strategic, timely, and well-informed decisions. We partner closely with IT teams to provide everything they need, from actionable tools to analyst guidance, ensuring they deliver measurable results for their organizations.
What Is a Blueprint?
A blueprint is designed to be a roadmap, containing a methodology and the tools and templates you need to solve your IT problems.
Each blueprint can be accompanied by a Guided Implementation that provides you access to our world-class analysts to help you get through the project.
- Determine When You Should Use Synthetic Data – Template
Talk to an Analyst
Our analyst calls are focused on helping our members use the research we produce, and our experts will guide you to successful project completion.
Book an Analyst Call on This Topic
You can start as early as tomorrow morning. Our analysts will explain the process during your first call.
Get Advice From a Subject Matter Expert
Each call will focus on explaining the material and helping you to plan your project, interpret and analyze the results of each project step, and set the direction for your next project step.
Unlock Sample ResearchContributors
- Simon (Haoyu) Sun, Manager, Customer Advisory, SAS
- Ash Aly, chief data scientist, QuantCreative
- Grace Marshall, Sr Specialist, Generative AI, Amazon Web Services
- Sanja Fidler, VP of AI Research, NVIDIA and Associate Professor, University of Toronto
Determine When You Should Use Synthetic Data
Build a Data Warehouse
Leverage Big Data by Starting Small
Architect Your Big Data Environment
Beyond Survival
Prepare for AI Regulation
