Most discovery phases open with the wrong question. What do customers want from the website? It sounds rigorous. It produces a persona deck, a survey with a 4% response rate, and a set of findings that could describe any retailer in the category.
Here is the better question, and it costs one afternoon:
What did customers phone up to ask?
Why That Question Is Better
Three reasons, and they compound.
The answer already exists. It is held, in unusual detail, in the heads of the people who answered the phone. They have heard the same question four hundred times. They can tell you what it was without a workshop, a whiteboard or a research budget.
It is behavioural, not hypothetical. A survey asks people to predict their own future behaviour, which they are bad at. A recurring phone question is a record of friction that was real enough to make somebody pick up a telephone — the highest-effort action a customer will take before abandoning.
It is almost never what the catalogue prioritises. This is the part that makes the exercise worth doing. A merchandiser will optimise the hero image and the category taxonomy. The recurring call might be about whether the jar is pasteurised, whether it ships frozen, or what the actual weight is. Nobody’s roadmap has “confirm the weight” on it.
Prominence is a finite resource, and most teams allocate it by internal consensus. A recurring phone question is the only cheap signal you will ever get about where it should actually go.
What Happened When We Asked
On the Gutta på Haugen project — an Oslo delicatessen, trading since 1994, whose website was a form and a callback until we replaced it — we asked exactly this question. The answer narrowed the problem usefully in two directions at once.
“Their shoppers were well informed. However offline most people would ask about allergens.” — Designer, Kilowott
Read that carefully, because there are two findings in one sentence.
First: the audience was not the problem. These were people who already knew good food. They did not need the category explained, benefits sold, or hand-holding through a decision. A great deal of standard ecommerce copywriting advice — educate, reassure, build desire — was simply not the job here. That single observation ruled out an entire register of design.
Second: there was exactly one recurring blocker, and it was allergens. Not price, not provenance, not delivery. One question, asked consistently, by informed shoppers who otherwise knew what they wanted.
What We Built, and What We Would Weight Differently
The product page had to absorb the phone call, so it carries the producer name, size variants, a delivery estimate, ratings with review count, loyalty points against the price, and four collapsible panels: product information, nutritional information, about the producer, and allergens — with allergens broken out as its own panel rather than folded into nutrition.
That is the right structural decision. Allergens earned a dedicated home because they were the recurring question.
Now the honest part. That panel sits fourth in a collapsed list, roughly a screen and a half below the buy button. Asked directly why it was not surfaced higher, the designer’s answer was straightforward:
“It was not extremely important.”
We published that in the whitepaper under “what we would weight differently,” and it is worth repeating here.
To be precise about the legal position: under EU Regulation 1169/2011, which applies in Norway through the EEA, mandatory food information including allergens must be available to the customer before the purchase is concluded in distance selling, and provided without additional cost. The information is on the product page before purchase, so the design meets that floor. This is not a compliance finding.
It is a weighting question, and the weighting is arguable. When your own research surfaces exactly one recurring blocker, that is the strongest signal you will get about where to spend prominence. Treating it as ordinary catalogue metadata — fourth panel, collapsed, below the fold — is a defensible call, but it is a weaker one than it first appears. Knowing the floor is not the same as deciding where to sit above it, and the second decision deserves to be made deliberately rather than inherited from a component library.
How To Run the Exercise
It takes an afternoon and produces a ranked list.
- Interview the people who answered the phone. Not the marketing team. The staff who took the calls, worked the counter, or answered the inbox. Two or three of them, separately.
- Ask for the top five recurring questions, in frequency order, in their own words. Do not offer categories — categories contaminate the answer.
- Ask which questions stopped a sale and which were merely asked. This is the distinction that matters most, and it is the one that survey data cannot give you.
- Ask what customers got wrong. Wrong assumptions about shipping, sizes, storage or substitutions are the questions people did not know to ask.
- Map each answer to a place on a page, and give the sale-stoppers the prominence. Anything that recurs but does not block gets depth, not position.
Step three is where most of the value sits. A frequent question is not automatically a purchase blocker, and the two deserve different treatments — position for blockers, depth for the merely curious. Our own project is a reasonable example of getting the structure right while leaving the position debatable.
The Wider Point
This question is one row in a larger inventory. In an expertise-led shop, the counter staff were doing eight commercial jobs, of which answering questions is only the first — the rest are set out in the counter problem.
Two of those jobs have their own research shortcut. Recommending is best solved not by asking what people want but by finding where they have already declared a plan, which is why the recipe page turns out to be a storefront. And pointing out today’s offer is a question of when emphasis is true rather than how loud it is, which we cover in emphasis that is always on is not emphasis.
One last thing about timing. The people who answered the phone are only available until the phone stops ringing. Institutional memory of a manual process evaporates within about a year of replacing it — staff move on, and nobody writes down what they were doing because it was never a feature. That is the same argument we make about baselines in measure the process you are about to delete, and it applies with equal force to knowledge.
The research is free. It is just perishable.
