How to understand your audience and make sure your message resonates

Audience research goes wrong when it starts with a method. Someone commissions a survey, or runs a persona workshop, before anyone has written down the question the work is meant to answer. What comes out is a document nobody consults, because it was never attached to a decision.

Start at the other end. Name the decision, choose a method that can answer it, and record the evidence in a form someone else can check.

Start from the decision

A question is worth researching when 2 plausible answers would lead to different work. If you would write the same page either way, the research is decoration.

Write the question as a sentence with a decision attached. "Do buyers compare us against agencies or against hiring in-house, and which comparison should the pricing page answer?" is researchable. "Who is our audience?" is not, because no answer to it changes anything you do next.

Demographic attributes are worth collecting when the decision is about reach: which publications to approach, which regions to sell into, how large a market is. They do not carry the words a person uses for their own problem, and most messaging decisions need those words. Extracting that language is a job of its own, covered in how to align your brand messaging with your target audience.

The questions worth answering

6 question families cover most of the decisions a small team faces:

• what problem they believe they have, and the words they use for it
• what they do about it now, including doing nothing
• what happened immediately before they started looking
• who else is involved in the decision, and what each of those people needs to agree
• what would stop them buying, and what evidence would remove that objection
• what a good outcome looks like to them, in their terms

The last one is where research most often stops too early. A stated goal such as "more leads" is usually shorthand for something more specific, and the specific version is the part a message can address. The distinction between what people ask for and what they need is worked through in user needs versus what users ask for.

Match the method to the question

No method answers every question. Each is good at a particular kind of evidence and blind to the rest.

Question	Method	What it gives you	Its limit
What words they use for the problem	Support tickets, live chat and pre-sale emails	The problem stated under pressure, unprompted	Covers only people who already found you
What triggered the search	Interviews with recent buyers, within weeks of the purchase	Sequence and cause, in their order	Memory degrades quickly, so timing matters
What they do instead	Interviews, plus reviews of the alternatives	The real competitor, which is often a spreadsheet or nothing	People understate the effort of their current workaround
Who else decides	Interviews with the buyer, asking who they had to convince	The approvers a persona omits	The buyer may not name the person who blocked it
What stops them	Interviews with people who did not buy	The objection, and whether evidence would answer it	Hardest group to recruit, and the most useful
How widely a view is held	A survey, written after the interviews	Rough proportions across a larger group	Can only ask questions you already know how to frame
What they search for	Site search logs and the queries bringing visitors in	Verbatim phrasing, at scale, with no recall bias	No context, and no way to ask why

Sequence matters more than coverage. Qualitative work comes first, because a survey can only ask what you already thought to ask. Run the interviews and read the tickets, then use a survey to size what you found.

Stop interviewing when new conversations stop producing new problem descriptions. That point arrives sooner than most people expect within a single role, and it resets for every distinct role you sell to.

What to record

Most audience documents record conclusions: a persona, a summary, a set of adjectives. When the market moves, nobody can say which parts are still true, because the trail back to the source is gone.

Record the evidence instead, one row per observation:

Column	What goes in it
Date	When the observation was made, because evidence expires
Source	Interview, ticket, review, search log or survey
Who	The role and the situation, not the person's name
Observation	The quote or the behaviour, verbatim and untidied
Question it answers	Which of your research questions this bears on
Decision it feeds	The page, claim or choice it should change

The last column keeps the record honest. An observation feeding no decision is worth keeping as background, and it does not justify commissioning more research.

A conclusion still gets written. It sits above the rows, and every line of it points at the rows underneath, so a colleague can disagree with your reading without redoing the fieldwork.

When to do it again

Repeat the work when something specific changes, rather than on a schedule:

• you start selling to a role you have not interviewed
• the product changes which problem it solves
• support language shifts, and tickets begin describing something you do not recognise
• a claim that used to land stops being repeated back to you in enquiries

The last signal is the cheapest to watch, and it is a measurement question as much as a research one. Turning it into something you can track is covered in how to set KPIs and marketing objectives.

Write the question before you pick the method

Take the next content or messaging decision on your list. Write the question whose answer would change it, as a full sentence. Find that question in the table above and run the one method next to it. Record what you find as rows, with the decision named in the last column.