Moderated Sessions in English and French, Synthesised

Five moderated sessions per audience segment settle most design arguments, and the sixth rarely changes the answer. What decides whether a study was worth booking is not the count. It is whether the tasks put the participant somewhere they could fail. Studies that come back inconclusive were almost always scripted so failure was impossible, and I will rewrite a protocol rather than run it. Writing the tasks: Every task traces back to a decision you are about to make, set in the participant's own context — their own dog, their own room and its measurements, their own phone with their own notifications arriving. No hypothetical shopping. No feature tour. Prototypes get tested only when the flow they encode is real enough to defend under pressure; a clickable file with three working paths teaches you about three paths, and I would rather say that before we recruit anyone than explain it in the debrief. Moderating, in English and in French: Sessions run in the participant's own language, moderated in it rather than through a script translated once and read aloud, which is the standard failure of pan-Canadian research and it costs you exactly the participants who were already reluctant to speak. I moderate with a light hand: think-aloud prompts, nothing leading, silences held past the point of comfort. I do not defend the product when a participant criticises it, including when the person who built it is watching. Stakeholders are welcome to observe. They may not speak. Recruiting: I write the screener and review who comes through it. Panels that supply professional participants get turned down, and the tell is a screener pass rate above roughly 40 percent — real incidence for a specific buyer is far lower than that, and a panel hitting it is coaching. Where your customer base is reachable, recruiting from your own list beats any panel for anything involving purchase intent, and I will push for that even though it fills more slowly and makes the schedule worse. Synthesis: The write-up is organised by decision, not by participant, because a deck of eight participant summaries pushes the reader into doing the analysis themselves and they will do it badly. Each decision carries what was observed, how many participants it applied to, what contradicted it, and a recommendation with my confidence stated in words rather than a rating. Where sessions disagreed, the disagreement stays in the document — a study that resolved every open question was probably scripted to resolve them. Clips are cut to the moment and cited, so a stakeholder disputing a conclusion can watch the evidence in under a minute instead of relitigating it in a meeting. Not included: No survey design, no quantitative validation, no unmoderated panel testing, no experiment work off the back of the findings. This ends at a documented set of decisions with the evidence attached to each. Spend the money elsewhere when: The question is which of two headlines converts better and you have the traffic. That is an experiment, and it is cheaper than this. The question is how many people do something. Eight participants cannot answer it and I will not let a deliverable imply otherwise. The team has already decided and wants footage of customers agreeing. Moderated work earns its cost in three situations: the behaviour is known and the reason is not, the flow does not exist yet, or the argument has run for a month and nobody in it has watched a person use the thing.