Quantitative market research, read by social anthropologists

Running the statistics is the part software has made easy. Knowing what a number means isn’t. A figure is never a neutral fact, it’s a social fact: someone decided who to ask, how to ask, and what would count as an answer, and every one of those choices is human. We’re trained to read the behaviour behind the percentage, to see the two segments hiding inside one average, to notice when a result is technically significant and commercially irrelevant. The maths tells you what moved, but only the reading tells you why, and the why is the part you can act on. That’s the discipline we bring to your numbers.

Because the tool isn’t where studies go wrong. A platform will happily field a questionnaire in an afternoon, including a bad one. Most of what decides whether a number is trustworthy happens before a single response arrives, in how the instrument is built: a leading word, a scale that nudges, an option nobody admits to out loud, and the data comes back clean and still wrong. We design the instrument so the question doesn’t write the answer, then read the result as the behaviour of real people. The software is the easy part. The craft is knowing what to ask, of whom, and what the answer actually means.

This is the right thing to worry about, because a growing share of online panel data isn’t what it claims to be: bots, professional survey-takers, speeders clicking through for the incentive. We never go straight to a full launch: a soft launch checks the data is behaving before we scale, then bots, speeders, straight-liners and duplicates are screened out before a single number is read. We choose panels for quality, not just price, and partner with providers who incentivise honestly rather than reward volume. Clean data isn’t a formality at the end, it’s a discipline that runs through the whole fieldwork. A number built on contaminated data is worse than no number, because it looks just as convincing.

You can outsource the fieldwork cheaply, but you can’t outsource the thinking cheaply and still expect a decision you can defend. The low-cost route hands you an automated report: toplines, a stack of bar charts, and a significance test presented as if it settled the matter. What it skips is the part that earns its fee: breaking every figure by who said it, testing what actually moves the outcome, reading each result as a group of people with a motive. The analysis is where a study becomes a decision, and it’s exactly the part the cheap version leaves out. On a decision worth real money, the saving on research is often the most expensive line in the budget.

Scale is what a large agency sells, and it comes with a cost you feel later. Your study is often designed by senior names and then handed to a rotating junior team, run through a standard process built for throughput, and handed back as a deck someone else has to turn into a decision. We stay small on purpose: the people who design your study are the same ones who read it and sit with you to make sense of it, and we take on a deliberately small number of high-stakes projects rather than everything that comes through the door. You aren’t buying capacity, you’re buying senior judgement on the decision that keeps you up at night.

Each answers a different kind of question, and knowing which is which is half of doing this well. Quantitative is for the how many: size, share, proportion, the ranking or the price that has to hold up at scale, the proof you can defend in the room. Qualitative is for the why: the motivation and the meaning behind the number, the reason a person acts. Social listening is for the live conversation: what is being said right now, the culture and the early signal you can still act on. Most high-stakes decisions need more than one, which is the whole reason the H3™ method exists. Rather than sell you the lens we happen to run, we help you see which one your question really needs, then fuse them into one answer.

By being as careful about what a number can’t carry as about what it can. Precision is easy to manufacture: two decimal places make any figure look certain, whether or not the sample earns it. We’re honest about the range around every estimate, we flag when a result is directional rather than conclusive, and we refuse to dress up a small base as a hard finding. For specialist audiences we read the signal qualitatively rather than force a confidence interval that would mislead. A number you can trust is one that tells you how sure to be, not just what the answer is. That honesty is what makes it defensible when the board pushes back.

Senior specialists, from the first conversation to the final reading. We’re a boutique on purpose: the people who design your questionnaire are the same ones who screen the data, run the analysis and sit with you to make sense of it. Your study isn’t scoped by senior names and then handed to a junior team working to a template. When we tell you what a number means and why it matters for your decision, a researcher with real expertise in both statistics and social science is standing behind that reading.

A decision you can defend, not a dashboard you have to interpret alone. Depending on the study, that can mean segmentation maps, key driver analysis, crosstabs read by subgroup, confidence and margins stated plainly, and interactive tools you can question in the room. But the centre of it is always a clear account of what the numbers mean for your specific decision, and a recommendation you can put in front of the people who have to sign it off. A number tells you where you stand, and we hand you the reading that tells you what to do next, because that’s the part we won’t leave you to work out alone.