What consumer insights are
A consumer insight is an explanation of behaviour that is specific enough to change a decision. It has three parts: something people did, the reason they did it, and a consequence you can act on. Drop any one of the three and you are left with a fact, an opinion, or a slide.
The word gets used loosely enough that it has almost stopped meaning anything — every supplier of every kind of data now sells “insight”. So it is worth being strict about it, because the strict version is the one that earns a seat at a stage gate. If nobody does anything differently on Monday, it was not an insight.
That test is unforgiving on purpose. It rules out most dashboards, most tracking studies and a good share of what arrives in a debrief deck. It also explains why the job title exists: somebody has to be accountable for the difference between knowing something about consumers and doing something about it.
Data, research, insight — three different things
These get used interchangeably, and the confusion is expensive, because each one answers a different question and only one of them settles an argument.
Illustrative. The point is the level of the statement, not the figures.
Data is what happened. Research is what you asked. An insight is the thing that changes a decision. Data and research are inputs; the insight is the output, and it only exists once somebody can act on it. This is also why volume does not help: a hundred more rows of the first kind never turn into the third.
The gap between the second and the third is the hard one, and it is not a matter of effort. People frequently cannot report the reason for a food choice, because most eating is habitual and largely automatic — so they supply a plausible reason instead of the real one[1]. Ask harder and you get a better-worded guess. The reason has to be observed.
Read the finding out loud and finish this sentence: “so we will…”. If the sentence completes itself, it is an insight. If it stalls, or if it completes with “…keep monitoring”, you have data that has been formatted to look like a conclusion.
The three jobs a consumer insights function does
Strip away the deliverables and there are three, and only three. Each one is a different moment in a launch, and each one needs different evidence. They also map onto why Eatpol has three products rather than one: Nova to shape, Studio to clarify, Vox and Domus to protect.
Shape the decision while it is still open
Evidence belongs in the room before the brief is written and the tooling is ordered. After that point, everything you learn is either a confirmation or a delay.
This is the part most often skipped, and the part that separates the teams whose work lands from the teams whose work gets filed. In Insights2020 — an industry study of over 350 CEOs, CMOs and insights leaders across 60 markets — 79% of insights functions at over-performing companies took part in strategic decision-making at all levels, against 47% at under-performers[2]. The same study found the function reporting directly to the CEO at 33% of over-performers versus 13% of under-performers. Read it as a description rather than a cause: the functions that are in the room early are the ones attached to the companies that are winning.
Being early also means challenging the assumption in the brief while challenging it is still free. The most useful thing you can do at that stage is widen the option set: not “is this idea good?” but “what else should have been on this list?”
Eatpol Nova exists for exactly that moment. It starts you from 821 products indexed across 10 award and innovation programmes, plus market-trend signals from 86 countries, instead of a blank page and a hunch. Its Observatory puts the 529 with label imagery in front of you at once, so you can see what a category has actually been rewarding — formats, claims, positionings — before anyone commits to a direction.
Make the trade-offs explicit
Not every idea deserves a launch slot. Most of the value here is in making the choice between two options a finding rather than an argument.
Prioritisation is where insight work either earns its place or turns into diplomacy. Two concepts, one slot, and a room in which the most senior opinion tends to win. The only thing that reliably breaks that tie is evidence about behaviour: which one people actually used, kept in the cupboard and went back to.
Asking them to rate the options will not settle it, and this is the single worst case for stated intent. Purchase intentions predict sales better for existing products than new ones, better for durables than non-durables, and better over short horizons than long[3]. A new food product is on the wrong side of all three. Intentions in general translate into action roughly half the time[4] — fine for a directional read, far too soft to kill somebody's project with.
Eatpol Studio is where the comparison lives. Every study lands in one place: the footage, the transcripts, the coded behaviour and the recommendations, side by side across concepts, so a trade-off can be shown rather than asserted. Then the conversation moves from whose judgement to trust to which cost you would rather carry — which is the conversation you wanted in the first place.
Protect the growth you already have
Short-term pressure is exactly when brands ship the untested thing. One week of real behaviour costs less than a delisting.
The third job is defensive, unglamorous, and the one that quietly pays for the function. When a quarter is under pressure, the reformulation goes out without a home test, the pack change ships to hit a cost target, and the line extension is approved because the slot was already booked. Each of those is a bet against a product that currently works.
The downside is not confined to the new item. Peer-reviewed work on brand dilution found that an extension carrying attribute information inconsistent with what people already believe about the parent brand can weaken those beliefs — and that dilution was less likely when consumers saw the extension as atypical of the brand[5]. In other words the risky launch is not the odd one out on the shelf. It is the one that sits closest to your core.
Nor is an established product a fixed quantity. When 105 people ate the same meat sauce once a week for ten weeks at home, boredom rose and acceptance fell over the run, least of all where there was variety and choice[6]. What you sell is being re-judged every week by people who will not tell you when they stop.
Against that, the cost of checking is small. Vox interviews your target consumers on video and comes back in about two days; Domus puts the product in their kitchen and shows you what happens over a week of real use. Set that against 50–75% of new food products failing within two years (peer-reviewed estimates; industry panels put it as high as 85%)[7] — a figure that is itself contested, with a later empirical study of food launches putting category success between 58% and 88%[8]. Either way, enough launches miss to make one week of evidence the cheap side of the trade.
Why insights fail to land
Good work gets ignored for boringly consistent reasons. Almost always one of these three, and none of them is about the quality of the analysis.
That third one has a name. The distance between what consumers say in a survey and what they do with the product at home is the say-do gap, and it is what makes reassuring research so easy to produce and so expensive to believe. A concept test can hand you a top-box score and a launch that fails anyway, without either result being a mistake — they simply measured different things.
The fix for all three is the same, and it is structural rather than methodological: move the work earlier, scope it to the decision rather than the questionnaire, and make the primary evidence behaviour rather than agreement. That is harder to schedule and much harder to argue with.
There is no credible peer-reviewed figure for how much of corporate research goes unused, or for how far a food concept test overstates real trial. Both numbers circulate widely; every version we could trace led back to vendor marketing with no method attached. So they are not on this page. Ask anyone who quotes one for the paper.
What this looks like as work
Concretely, on a food or drink launch: consumers recruited to match your target audience use your product in their own kitchen and record it on their own phone. You get the footage and the analysis, typically one week from kickoff — fast enough to sit inside a stage-gate cycle rather than pushing it.
What comes back is not a score. It is a ranked list of specific, fixable things: the step where four of ten people got the preparation wrong, the moment the texture changed someone's face, the point in the pack where they gave up and reached for scissors. You still ask questions — you just stop treating the answer as the primary evidence, because now you can check it against what the same person did thirty seconds earlier.
That is the whole discipline, really. Be early, be specific, and bring something a sceptic can watch. See how a study runs, or book a call and we will walk through what it would look like in your category.