Key Takeaways
- Synthetic data in marketing is artificially generated data that mimics real data. A model trained on patterns from real datasets creates new records instead of surveying or observing actual consumers.
- A digital twin is related but different: it’s a model of an individual, built on real data, that predicts behavior and responses and produces synthetic data as its output.
- The main advantages over real data are faster project deployment without recruiting respondents, privacy protection, and scalability without added cost for each new record.
- A general AI tool isn’t a substitute for synthetic data. Its answers draw on population-level data anyone can access, and it can carry bias forward.
- Synthetic data is only as accurate as the real data it’s trained on. It should be validated against real-world benchmarks.
- Practical use cases include testing a product concept with a small, hard-to-reach customer segment and generating several distinct audience options for an agency pitch.
Synthetic data has gone from a niche research tool to something brand and agency professionals are actively trying to implement. Just think about some of the recent vendor pitches and conference sessions you’ve attended: Six months ago, “using synthetic data in marketing” was likely rarely mentioned, if at all.
Now, it’s everywhere: 69% of market research professionals report using synthetic data in the past year, and 71% believe most research will use synthetic responses within three years, according to Greenbook GRIT. The same research also shows that since last year, synthetic data adoption among the largest service-led suppliers has nearly doubled.
One of the reasons synthetic data is gaining rapid visibility is that marketers need to close their knowledge gap fast. They have to learn what it is, what it’s not, what the benefits are, and how to use it, all while executives are asking a question that’s several stages down the line: How do we get measurable outcomes from this?
If you feel like you’re behind on the learning curve, you’re absolutely not alone. This is case where everybody is starting from zero, and we all have to learn about something new. This blog will act as a primer for those brand and agency marketers who are getting themselves up to speed on synthetic data by answering basic questions, with the goal of giving you a platform you can then use to figure out how to apply it to your particular business.
Essential Definitions and Answers to Common Questions
What is synthetic data?
Gartner provides a simple, user-friendly definition: “Synthetic data refers to generated data that mimics real data.” In other words, it’s artificially generated data that’s built to resemble real data without being collected directly from real people.
So, instead of surveying or observing actual consumers, a model is trained on patterns from real datasets and then used to generate new, synthetic records that reflect those same patterns and relationships.
What is a digital twin?
When you’ve encountered synthetic data, you may have heard about a related topic, the digital twin. This is an AI representation or simulation of individuals used to predict and anticipate their behavior and responses.
While these two concepts overlap, they aren’t the same. A digital twin is a model that can be built on real data or a real customer and then outputs synthetic data.
Why Would a Company Use Synthetic Data in Marketing Instead of Real Data?
Synthetic data offers several major advantages over real data, including:
- Enhanced privacy protection: When you use real data, there are all kinds of risks to people’s privacy. This creates ongoing liability for you. If your systems experience a breach, real people are exposed to identity theft, financial harm, or even embarrassment if sensitive person details are released.
With synthetic data, there’s no direct relationship between the data points and real people, so you can share your insights across teams, within organizations, and with external partners more freely. This is particularly useful if you work in an industry that has a lot of regulatory compliance, like healthcare or financial services. You can collaborate more quickly with synthetic data because partners who might normally be restricted from accessing real customer data will have fewer hoops to jump through with synthetic. - Faster project deployment: With traditional research, you have to recruit respondents, field a survey, wait for responses to come in, clean the data, and analyze it. All of that takes a long time! Even if you purchase data from somebody else and skip the first few steps, you still only get who’s in the panel, meaning you’d only be able to target those specific people. The research isn’t modeled and projected to the full population. And you still have to integrate the data with your martech stack and analyze it. If you’re dealing with a niche or hard-to-reach population, the process can be even slower.
With synthetic data, the timeline is shortened. “Respondents” don’t need to be recruited. Once a model is trained on real, validated data, generating a new dataset for a new population or a different question is simply a matter of running that model again. There’s no need for a multi-month custom research cycle anymore. - Unlimited scalability: Anytime you want to expand your collection of real data, it costs you more time and money. Scaling synthetic data doesn’t come with such a restraint. Once the underlying model is built and validated, generating additional synthetic records, whether it’s 1,000 or 100,000, carries no meaningful incremental cost or recruitment burden since there’s no real person being found or compensated for each new record.
Can’t I just use AI instead of synthetic data?
Many marketing teams are already leaning on AI heavily to perform a variety of functions, and some may wonder why they can’t just use what they’ve already invested in instead of synthetic data. AI’s biggest benefit is that it gives you answers quickly, but that’s where the benefits end. AI cannot act as an effective stand-in for synthetic data because:
- Its answers are based on population-level data, which anyone can get, resulting in audiences that are generic rather than differentiated and therefore competitive.
- It wasn’t built for the work synthetic data does, so its responses are just guesses.
- AI tools tend to carry bias forward, which means there’s a good chance your results will contradict those of traditional research and you won’t be aware of it.
Is synthetic data accurate?
Once marketers learn what synthetic data is, the next question they ask is something along the lines of, “Okay, but can I trust it?” The answer: It depends.
Specifically, whether you can trust synthetic data in marketing depends on the data it’s ground in. Synthetic data, just like the agentic AI you’re accustomed to using, is only as good as the real data used to train the model generating it. If your synthetic dataset was trained on biased or outdated real data, for instance, it will just reproduce those same flaws at scale.
Synthetic data is a model of reality, not reality itself. It can be validated against known outcomes and used with real confidence in many use cases, but it shouldn’t be treated as a perfect substitute for real, individual level data on a population you could just study directly.
What can’t synthetic data do?
Synthetic data is powerful when the question fits. Here are a few things it isn’t good for:
- Responding to breaking news, when real-time events are still shaping opinions
- Giving insights into new or emerging topics, which don’t have enough history to accurately ground a prediction
- Awareness measurement that lets you know whether someone has heard or seen something
- Messaging and creative testing, in which you get reactions to a finished, viewable execution
Practical Use Cases for Synthetic Data in Marketing
In this section, we’ll address two use cases: one for a brand, and one for an agency. Let’s start with the brand.
Brand Use Case: Testing a new product concept across underrepresented customer segments
A skincare brand is developing a new product line that’s specifically for customers with a certain skin condition. Unfortunately, this condition isn’t common, and the number of customers in the brand’s own data who live with the issue is pretty small. If the brand goes the route of a traditional survey, it’ll be slow and expensive. They may even have to spend time and money recruiting more participants with the skin condition if their own segment is too small to produce a statistically reliable result.
With synthetic data, the skincare brand can easily generate a statistically representative synthetic sample of this small segment that’s grounded in real data about this population and closely related ones. The team can get feedback not just on their product concept, but also on messaging and purchase intent. If the synthetic results suggest strong interest, the brand can decide what to do next: It can either go ahead with a traditional study, having used the synthetic data as a low-cost test, or it can proceed with a marketing campaign to launch the new product to the real world.
Agency Use Case: Generating multiple audience options for a new client pitch
An agency is preparing a pitch for a prospective client. As always, there’s fierce competition to win the account, and the team needs to show up with audiences that are genuinely distinct from the angles the client has already considered. However, it’s hard to justify commissioning real research when the pitch hasn’t even been won yet and the deadlines make squeezing in custom data totally unrealistic.
Using synthetic data, the agency can quickly generate several audience options, each reflecting different underlying purchase motivations or behavioral patterns. That way, the team can walk into the room with multiple data-backed angles to discuss with the client without having spent time and money on commissioning primary research for a pitch with no guaranteed return.
Want to Learn More about Synthetic Data?
As synthetic data is used more in marketing, more use cases will be established, as will the outcomes brands and agencies can expect from this kind of data. In the meantime, it’s important to understand what synthetic data is and what its strengths and limitations are before you decide if it’s right for you.
If you’re ready to learn more beyond this blog, schedule a consultation with a Resonate data expert today. Our knowledge is grounded in nearly 20 years of experience working with all kinds of data, including synthetic, and we’ll be able to help you decide if synthetic data is the right move for your business.
