How to Write a Discovery Questionnaire Clients Actually Complete

Every other document on this list succeeds by being read. A discovery questionnaire only succeeds if something comes back, which makes it the only client document whose quality is measured entirely by somebody else’s willingness to do work.

That changes the design problem completely. A brilliant, thorough questionnaire that gets a quarter of the way through and is abandoned is worse than a blunt one that gets returned.

The failure: an internal checklist, made public

Most questionnaires are the list of things the sender needs to know, written in the sender’s vocabulary, in the order the sender’s process consumes them. Forty questions, several of which are really three questions, several of which the client has to translate out of your terminology before they can even begin.

The client opens it, sees the length, decides it is a two-hour job, and puts it on the pile for a quiet afternoon that never arrives. Nothing about your process was wrong. The document was.

Four rules before you write a single question

Do not ask what you can find out yourself. Anything on their website, in their public accounts, or visible in the tools they have already given you access to. Asking is not diligence; it is a tax on their time that also signals you have not looked.

Do not ask what you will not act on. Go through your last questionnaire and mark every answer that changed a decision. The unmarked ones are the ones costing you completions.

One question per question. “Who is your audience and what do they currently use?” gets you half an answer to whichever half was easier.

Put the hardest question after three easy ones. Not first, where it stops them, and not last, where they are out of patience. Third or fourth, when they are invested and warm.

The five sections

Why you are asking, and how long it takes. Two sentences and an honest number of minutes. Time it yourself by answering it as a client would. If your estimate is wrong the reader finds out at question six and stops.

The five questions you would ask if you only had five. Ordered so that if they abandon it here, you still have enough to start. This section is a hedge against the most likely outcome.

The specific questions, grouped by who can answer them. Not by topic. See below — this is the trick that raises completion.

Anything that needs a decision rather than information, flagged as such. A question that requires two people to agree is a different kind of ask, and marking it lets your contact route it instead of stalling on it.

What happens next, and by when. What you will do with the answers, when they get something back, and the date you need it by. An ask without a date is a wish.

Group by who can answer, not by topic

This is the highest-leverage change available and almost nobody makes it. A questionnaire grouped by topic — brand, technical, commercial — requires your contact to read every question, work out which ones are not theirs, and either guess or start a chain of emails.

Grouped by role — “for whoever runs the website”, “for whoever owns the budget”, “for whoever handles the content” — the document becomes routable. Your contact forwards three short sections to three people instead of forwarding one long document to nobody.

It also makes the length feel different. Thirty questions read as an ordeal; three sets of ten, each addressed to a different person, read as reasonable.

Ask questions that can be answered

The craft is almost entirely in the phrasing. Some rewrites that reliably work:

“What is your budget?” → “Which of these ranges are we in, so I do not design something you cannot buy?” with three ranges to pick from. People who will not name a number will happily tick a box.

“Who is your target audience?” → “Describe the last customer you were pleased to win. What did they do before they found you?” Abstract questions get abstract answers.

“What are your goals?” → “A year from now, what would have to be different for this to have been worth doing?”

“Any brands you like?” → “Name a competitor you would be annoyed to lose to, and one thing they do better than you.” The second half is where the useful answer is.

And offer an exit on every open question: “skip this if it is not relevant” at the top of the document raises completion, because the real reason people abandon a form is a question they cannot answer, not a question they do not want to.

Reading the signal on a document that asks for work

Because success here is a returned document rather than a read one, the reading signal is diagnostic rather than encouraging — it tells you what to change.

Opened, read properly, nothing returned. The document asked for more than the reader was willing to do. Shorten it, do not chase it. The follow-up that works is not “did you get a chance” but a shorter version with four questions in it.

Opened repeatedly by what looks like several readers. It is being routed internally, which is what the role-grouped structure was for. Give it a week before you nudge anybody.

Never opened at all. Wrong channel, wrong person, or wrong week — the three explanations, and which to test first.

One structural point worth knowing if you are sending this to several people at once: sending each recipient their own link is what lets you tell which of them the document travelled through, rather than seeing one anonymous crowd. Quixli grades reader identity honestly rather than guessing at it, and what each kind of share can tell you is set out on the features page.