Why synthetic data is like lucid dreaming (AKA how AI is changing market research forever)

Market research is about understanding people, trends, and behaviours to make better decisions.

It’s a field rooted in curiosity.

But let’s face it, the process is often clunky, expensive, and limited by the messiness of real-world data.

Enter synthetic data. It's created when your data has babies.

So, what is synthetic data, and why should you care?

Synthetic data is like the stand-in actor for real data. It looks, feels, and acts like the real thing, but isn’t tied to actual people or events. Generative AI creates this data by learning patterns from existing datasets and mimicking them in entirely new ones.

Here’s why this is a big deal:

• It protects privacy. We live in a world where data privacy isn’t negotiable. Traditional market research often flirts with privacy risks, but synthetic data makes those risks obsolete. It's not real data - it just acts like it. This keeps regulators happy and consumers reassured.
• It’s faster and cheaper. Market research can be painfully slow and expensive. Need data on a niche audience in a remote region? Generative AI can create a dataset faster than you can say 'field study.' Also - want to test a product launch in South America without waiting for months of research? Done.
• It can simulate data for scenarios you couldn’t otherwise access - like what Gen Z parents might still want in 2030 (do they even know?). Some questions are tricky to ask in the real world. Synthetic data lets you explore these areas without stepping on ethical landmines.

The Catch (Because there’s always a catch)

Have you heard of lucid dreams? Apparently, you can do pretty much anything in a lucid dream, apart from: seeing the time or your own hands.

Using synthetic data is a bit like lucid dreaming. They both exist in a space where reality is convincingly imitated, but certain details reveal their constructed nature.

Try this: Google an image of a watch.

You'll notice that the majority of watches follow the 10 past 10 phenomenon.

The position of the hands at 10:10 creates a symmetrical 'V' shape that frames the logo or brand name, which is often placed near the centre of the watch face. The symmetry looks clean, balanced, and visually appealing (which is why this time was chosen by watch marketers to optimise visuals for psychology and impact).

This also means that the data used to create training sets for images of watches are massively biased.

So, here’s what to watch out for when we're dealing with synthetic data:

• Bias replication. If the data you train AI on is flawed, your synthetic data will inherit those flaws. As we all know 'Garbage in, garbage out'.
• Validation hurdles. You’ll still need to cross-check synthetic data with real-world trends to ensure it’s accurate.

Why this matters

Market research isn’t going away, but how we do it? That’s evolving.

Synthetic data is undeniably one of the coolest tools we have at our disposal. It can mimic human behaviour, fill in data gaps, and even dream up entirely new datasets - all while keeping privacy intact. It’s efficient, scalable, and bursting with potential.

But, here’s the thing: synthetic data is only as good as the data it learns from. Feed it biased, incomplete, or skewed information, and it’ll happily churn out results that look convincing but quietly inherit those flaws.

Basically, synthetic data can do wonders, but when it comes to watches, it insists on staying fashionably stuck at 10:10.

Maybe artificial intelligence is just another morning person...