Customer Research

Your customers have the answers. Are you asking the right questions?

Most business leaders know they should do more customer research. Fewer know they're doing it badly — and a bad survey is worse than no survey at all.

Matthew Gaunt  |  April 2025  |  Customer Research

I've been saying this for years to every client I work with: ask your customers. They will tell you more in ten minutes than six months of internal debate. The problem is not that businesses don't want to do research. The problem is that most surveys are written by people who already know the answer they want — and every question in the survey quietly points the respondent towards it.

Leading questions. Loaded language. Double-barrelled questions that muddle the data. Scales that change halfway through. Questions written to validate a decision that's already been made. It happens everywhere — and not because people are dishonest. It happens because it is genuinely hard to write a clean, unbiased survey if nobody has taught you how.

So I did something about it. I'm not a research methodology expert — but I know the principles, and I know Claude well enough to build something useful. I put together a structured review and improvement prompt that applies established frameworks — Total Survey Error (TSE), Dillman's Tailored Design Method, and MRS best practice — to any survey draft you share with it.

The prompt is free to use. I've shared the full text below. Here's what it does — and why each step matters.

Step One

Before you look at individual questions, look at the whole shape of the survey. Is the order right? Standard practice runs screener questions first, then behavioural questions, then attitudinal, then demographic. Most surveys get this backwards — they open with "how old are you?" when what you actually need to establish is whether this person is in your target audience at all.

The structural step also checks length. Respondent fatigue is real. A survey that takes more than eight minutes to complete will haemorrhage completions — and the people who drop out are rarely random. You lose the busiest, most time-poor respondents first: often exactly the people whose opinion you most need.

Step Two

This is where the detail lives. The prompt reviews every question against ten specific failure modes — the most common ways a survey question stops producing honest data and starts producing comfortable data.

The ten failure modes it checks for:

• Double-barrelled questions — asking two things at once, so you can't tell which one the answer refers to.
• Leading questions — language that nudges the respondent towards the answer you'd prefer.
• Loaded language — emotional or assumptive wording that skews responses before the respondent has started thinking.
• Ambiguous wording — words like "recently", "often", or "satisfied" that different respondents will read completely differently.
• Acquiescence bias — the well-documented tendency for respondents to agree with statements regardless of their true view.
• Social desirability bias — questions where the "acceptable" answer and the honest answer are different, and respondents know it.
• Scale inconsistencies — mixing 1–5 and 1–10 scales so results can't be compared across questions.
• Response option problems — overlapping ranges, missing "none of the above", or forcing a choice when neutrality is the honest answer.
• Jargon — technical language that lay respondents will guess at rather than admit they don't understand.
• Built-in assumptions — questions that assume something about the respondent that may not be true.

Step Three

Consumer customers and commercial customers are not the same audience. A survey written for a residential homeowner should feel warm, plain, and easy. The same survey sent to a facilities manager or procurement lead should be professional, concise, and to the point. Most surveys get written in a single corporate register that is slightly too formal for consumers and slightly too vague for commercial buyers.

This step also flags anything that sounds like marketing — surveys should not be brand exercises. Respondents who feel they are being sold to while filling in a survey give worse data, and lower completion rates.

Step Four

This is where the prompt shifts from critique to construction. It advises on question types — when to use a Likert scale versus NPS versus a simple ranking versus open text. These are not interchangeable. NPS is useful for tracking loyalty over time. Open text is useful for surfacing themes you haven't predicted. Ranking is useful when you need to know relative priority, not absolute satisfaction.

It also recommends where to use skip logic — so respondents who haven't used a particular service aren't asked to rate it, and you don't have to filter that noise out of your data after the fact.

Step Five

The final output is a complete, professionally reformatted survey — ready to load directly into HubSpot Forms. Every question is formatted consistently: question text, question type, response options, and any skip logic. It closes with a warm, brief thank-you that respects the respondent's time.

The intent of every original question is preserved throughout. This is not a rewrite exercise. It is a bias-removal exercise. The form changes; the function doesn't.

The Bigger Point

Reinforcement thinking is the enemy of good research

The hardest thing about customer research is not writing the questions. It is staying genuinely open to an answer you didn't expect — or didn't want. Most survey drafts I see are written by people who have already formed a view. The survey becomes a mechanism to confirm it rather than test it.

That is not research. It is expensive guesswork dressed up as evidence.

The prompt below helps you break that pattern. Use it. And if you've built something better — share it.

Free to Use

The full prompt — copy and use it

Copy the prompt below into Claude (or your AI tool of choice). Paste your survey draft at the bottom where indicated. Run it. Work through the output section by section.

I work with consumer-facing boards and founder-led businesses on the commercial decisions that follow the data — pricing, positioning, channel strategy, customer acquisition. If you're using research to inform a bigger strategic question, I'm happy to have a conversation. No pitch. Just a conversation.