Market signals are clearer than ever.
What The Market Is Telling Us
Over the past several months, I’ve had substantive conversations with marketing leaders at FTSE 100 advertisers across FMCG, retail, beverages and entertainment. The pattern is consistent: The appetite for digital twins is not just positive; in many cases, it is more enthusiastic than for other new methodologies.
What I’m hearing from brand leaders maps to a specific pain point: Decisions are being made too slowly, on too little evidence and not with the scale needed. One FMCG brand told me their urgent calls are often made on gut instinct simply due to time constraints. A beverage brand described flying blind in noncore markets with no signal and no confidence. This is the problem digital twins are built to solve: moving decision intelligence upstream, before the budget commitment, before the global roll-out, before the expensive misstep.
The Test That Matters: When Twins Respond Like Humans
The key question is, what does “similar to human” mean in this context? The framing matters enormously. Directional similarity is not the same as individual prediction or even the same answers. The validation framework we tend to use at Kantar operates across four dimensions:
• Accuracy: individual twin answers against what real people actually said, for the same question.
• Consistency: asking the same question multiple times; the answer should mean the same thing each time.
• Signal Strength: twin outputs, compared against real survey samples to confirm the group-level pattern holds.
• Say-Do Gap: twins don't apply social norms the way humans do. Ask a human, "Would you drink alcohol in the morning?" and most will instinctively say no, even if technically they might at a wedding. A twin considers all plausible scenarios and may say yes—not because it's wrong, but because it lacks the social filter.
All four need to pass and be accounted for before outputs can be treated as actionable intelligence rather than generative noise. This is not yet an industry standard. Most players in this space operate without this level of traceability. The market is going to learn this the hard way: A persuasive-looking twin response with no audit trail back to source data is not a research insight; it’s a hallucination with a professional finish.
Why Use Cases And Skills Matter More Than Platforms
One of the most important lessons from two years in this space is that digital twins are not a universal replacement for research. They are strongest for exploration, iteration and early decision intelligence. They are not suited (yet!) for final KPI scoring, claims testing or anything that feeds investor reporting.
What’s emerging as the right model is the development of specialized intelligence skills. Preconfigured analytical workflows that define the question type, the data layers queried and the output format. For example, a skill for stress-testing and refining ideas before final screening. A proposition optimizer that shows how different audiences react to a price change, specific tensions or pack redesign. These aren’t just product features; they are the mechanism by which generative research becomes repeatable, scalable and auditable.
This is where the real enterprise value lies. A bespoke pilot is an interesting proof point. A skill that runs the same workflow for 50 clients across 20 markets, with consistent outputs and a traceable methodology, is a new operating model. We need to do the harder work of carefully scoping those skills with marketers, applying craft, testing them against specific use cases and building the governance layer that makes the outputs defensible to boards and legal teams.
What’s Being Built And Why It’s Different
I can share early signals on where rigorous digital twin approaches are heading. The architecture that matters is not a single model making clever predictions; it is a layered signal intelligence engine. At the base: proprietary panel data from real, consenting individuals, continuously refreshed and fully permissioned across 55 markets. Above that: open-source behavioral, demographic and market signals reconciled with the panel. Then, the brand's proprietary assets. All of these turn a generic twin population into one that understands your category, your brand and your competitive context.
The differentiator isn’t the interface or the generative layer. It’s whether every output can be traced back to a real source panelist and whether the methodology is certified against responsible AI frameworks and standards outlined by professional associations such as ESOMAR and MRS. These credentials are the foundation on which enterprise and brands can begin to act on synthetic decision intelligence rather than simply being impressed by it.
What This Means For The Consumer Intelligence Industry
We are at an inflection point. Digital twins are credible enough to be taken seriously by FTSE 100 marketing departments, but still immature enough that poor implementations will generate high-profile failures that set the capability back. The industry’s job right now is to be honest about the boundaries: Twins complement humans; they don’t replace them. They inform primary research; they don’t eliminate it. One of the companies I spoke to during my industry research did something refreshingly bold: They focused on just two pre-validated use cases and stopped there. No overreach, just real clarity and conviction.
For those of us who have been working in synthetic research long enough to have seen the early skepticism give way to genuine interest, the opportunity is real. But it will be won by the organizations willing to do the unglamorous work: rigorous validation, use-case specificity, skills development and client education. The technology is ready to deliver decision intelligence at scale. The question is whether the industry is disciplined enough to deploy it well. The best synthetic data engines are those designed with a clear understanding of their own limits.
Watch this space for my next article.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
