The Problem with Synthetics and “Customer Research”

Inevitably, the arrival of synthetic data into marketing, and then into customering, is upon us. But sometimes, just because someone turns up at the front door, it doesn’t mean they’re welcome. We must choose who we let in, and there’s a gigantic difference between the use of such data for market research, and for what many call “customer research”.

‍Arise Synthetics

Probably the most well-known company to advance the case – and commercial offerings – of synthetic data is, Evidenza. The company, which was founded by Jon Lombardo and Peter Weinberg, formerly of LinkedIn fame, is growing rapidly and enjoys global customers like EY, Salesforce, and Mars to name a few. At Cannes in 2025, it announced a global partnership with Dentsu, integrating its audience research into the firm’s media planning processes. 

The real power, or promise, of synthetic research in the service of market diagnostics, is found in its touted ability to automate not just reliable market segmentation, but other valuable information like excess share of voice, funnels, pricing analysis, and category entry points.

Good stuff.

But that’s market diagnostics. It’s fine, exciting even, and a space to watch. By contrast, those now claiming to use synthetic data within the customer base, are doubling down on one of the industries big errors and supercharging the calamity.

The big error of course, is the engrained fable that we can manage a customer base via “feedback”. The concept was largely made popular by a customer experience association, an organisation that behaves like it was formed by a survey software analyst, because it was.

Around the world, many have been influenced by this error, never stopping to wonder why no field of disciplinary management and no regulated profession, uses feedback as a central tenet of its work. If they had, they might have found that much of the data generated by all this feedback is objectively among the least reliable forms of data that can be produced. 

Instead, they march on, oblivious to both the dishevelled nature of the data, and the more robust, evidence-based management practice available to them. Once synthetic data came along, their “progression” from manual surveys to synthetic customer research was inevitable. 

Different tool. Same rabbit hole.

One of the benefits of synthetic data for market diagnostics, is its ‘speed to value’. Compared to primary research which takes months to plan and execute, results can be available in literally hours, or with some analysis and adjunct service, days.

But customer management is not an annual cycle comprised of long-term brand building and shorter-term campaign motions. It is entirely a longitudinal discipline, and when someone in the customer research camp comes along to spruik their new AI solution, claiming these rapid results, they mis-understand the very foundation of the field. Damage quickly follows or worsens.

So here we are. One use case, market diagnostics, finds real value in synthetic data while another, the misplaced idea of ‘customer research’, most certainly does not. 

‍Understanding the Difference

The reason for that deviation, is because markets are heterogenous in nature and are divisible, while the economics of a customer base are founded on service, a paradigm that requires interaction at the level of each, individual.

For that we must operate with a level of ethnographic stealth, informed by nuanced individual journeys, created and owned by the customer, which can change constantly. We might enrich some aspects of that observation, particularly in added context from conversational data points or other forms, but we do so to inform real time interactions – a discipline based on significant literature. 

The broadly applied catchall of “customer research” to justify survey addiction (or to profit from it) is more of an industrial money-making machine and has little to do with effective management of the customer asset. Equally, the term “behavioural”, used liberally and without understanding to describe the same, or its newer synthetic and AI versions, is mostly product marketing. Word salad. Sizzle. No sausage.

It’s a truism lost on many, that mass marketing concepts rarely translate to the customer base. Most commonly, it’s not unusual for companies to over-index on sales messaging, which the literature shows can result in a lowering of purchase propensity. Case in point, martech practitioners often find themselves harming a customer base if they don’t know how to regulate based on the relevant critical theory and controls.

Well, the same issue presents itself here. Those that have come to understand that market research is important, have erroneously assumed that same methods or ideas apply in the customer base.