Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Digital Marketing › Marketing Strategy

Audience Research

How to learn who customers are and what they think, from data you already have and research you run, and how to turn findings into segments and messages.
Guideintermediate8 min readUpdated

What research is for

Every later decision in a marketing strategy is a bet on the audience: which segments to serve, how to position, which channels to use, what to say. Without research those bets are made on the team's assumptions, which are usually the assumptions of people who are not the customer. Research does not remove the bet; it improves the odds and, just as valuable, it gives the team a shared picture to argue from.

The research needs to answer four questions, and a programme that answers only the first — who they are, in demographic terms — has done the easy part and skipped the useful part.

  1. Who are they? Demographics, firmographics, location, device, where they already spend attention.
  2. What are they trying to do? The job behind the purchase: the situation, the goal, the obstacles. The jobs-to-be-done framing is useful here because it describes the customer's problem rather than the product.
  3. How do they decide? What triggers the search, what alternatives they consider, who else is involved, what proof they need, what stops them.
  4. What words do they use? For the problem, the category, the product and the benefits. These become keywords, ad copy and page headings.

Sources you already have

Most businesses are sitting on more audience data than they have read.

  • Web analytics (GA4 or similar): who visits, from where, on what devices, which pages hold attention, where people leave, which paths lead to conversion. Segmenting by source and by converting versus non-converting visitors is the first cut.
  • Search data. Search Console shows the queries that already bring people to the site and the ones where the site appears but is not clicked. Keyword tools show the wider demand around the category and the questions people ask. The query list is the audience's vocabulary, written by them.
  • CRM and sales records. Who bought, how often, at what value, through which channel, with what objections in the notes. Customer value by segment (LTV) tells you who is worth researching most.
  • Support and service conversations. The questions people ask after buying reveal what they expected and did not get; the questions before buying reveal what the site failed to answer.
  • Reviews, of your own product and of competitors'. Reviews are unprompted, specific and in the customer's own words; the complaints in competitors' reviews are positioning opportunities.
  • Video platform audience reports. A business with a YouTube channel can read who returns, when viewers are on the platform and which other channels they watch; the guide to reading YouTube audience reports covers what each report can and cannot tell you.
  • On-site search and form fields. What people type into the site's own search box, and what they write in free-text fields.

Reading these takes time and no budget. A useful discipline is to collect verbatim phrases into one document, tagged by theme, before trying to interpret anything.

Research you run

Primary research fills the gaps the existing data cannot: why, rather than what.

Customer interviews. Short conversations with recent buyers and with people who considered and did not buy. Ask about the situation that led to the search, the alternatives considered, what nearly stopped them, and what they would tell a friend. Avoid asking people to predict their behaviour; ask about what they did. A small number of interviews per segment usually reaches the point where new conversations repeat earlier ones.

Surveys. Useful for quantifying what interviews found — how common each job, trigger or objection is — and for reaching people who will not take a call. Keep them short, ask about past behaviour rather than intentions, and leave at least one open question.

Social listening. What people say about the category and competitors in public: forums, groups, comments, creator content. This is where the unguarded language and the running jokes about the category live.

Observation and testing. Watching people use the site or the product (moderated sessions, session recordings with consent) shows the gap between what they say and what they do. The CRO cluster covers this in depth; for strategy, a handful of sessions is often enough to overturn an assumption.

Sales and support staff. The people who talk to customers all day hold most of the answers to the four questions and are rarely asked. An hour with them is often the highest-return research step available. When a study needs an outside researcher or a budget, the research brief template fixes the decision, questions, sample and method before fieldwork starts.

Turning findings into something usable

Research that ends in a slide deck changes nothing. The outputs that the rest of the strategy can act on are:

  • Segments — groups that differ in a way that changes what the business should do; the segmentation guide covers how to choose the dividing lines.
  • A journey map per priority segment — the stages, questions, touchpoints and obstacles, built by the method in the journey mapping tutorial.
  • A vocabulary — the customer's words for the problem, the category and the benefits, organised by journey stage, which feeds keyword research, ad copy and page structure.
  • Objections and proof — the reasons people do not buy, paired with the evidence that answers each one.
  • Personas, if the team finds them useful — one-page summaries of each segment written from the research, with the jobs, triggers, objections and vocabulary, and nothing invented to fill gaps. The comparison of buyer personas and jobs-to-be-done covers when a persona helps and when the job is the more useful output. Personas generated by an AI model without research behind them are hypotheses, not findings, as the guide to synthetic personas and simulated respondents explains.
Illustrative example — not a real result

Research for a home-insulation installer

Analytics shows most enquiries come from search, on mobile, in the evenings, and that the pricing page is the last page before most form submissions. Search Console shows queries about cost dominate, followed by queries about grants and about a specific installation method. Reviews of competitors complain about mess, delays and unclear quotes. Six interviews with recent customers reveal the trigger is almost always a winter energy bill, the main alternative considered is doing nothing, and the decision maker wants a fixed price before anyone visits. The strategy that follows targets cost and grant queries, leads with a fixed-quote promise, uses the competitors' complaints as proof points, and times the heaviest activity to the first cold weeks of the year.

An interview guide that works

The questions below produce usable answers because they ask about what happened, not about preferences.

  1. "Tell me about the moment you realised you needed something like this. What was going on?"
  2. "What did you do first? What did you search for or who did you ask?"
  3. "What else did you consider, including doing nothing? What put you off each one?"
  4. "What nearly stopped you choosing us? What would have made it easier?"
  5. "When you describe this to a friend, what words do you use?"
  6. "What has it been like since? What did you expect that did not happen?"

Record the conversation, with permission, and transcribe the phrases verbatim: the exact words are the material for keywords, headlines and page copy, and paraphrasing loses them.

Research plan checklist

0/8

Keeping research current

Audiences drift: new competitors change the alternatives, platform shifts change where attention sits, price changes change who buys. Research is refreshed on a cycle — a light annual pass over the existing-data sources and a few interviews, with a deeper pass when something changes (a new product, a new market, a result that nobody can explain). The annual planning guide includes it in the cycle; the customer analytics expertise page shows how the author uses behavioural data for the ongoing part. Each refresh starts better from a store of past findings, and the tutorial on building a research repository sets one up with consistent tags, review dates and personal data kept out.

Common mistakes

  • Demographics as the whole answer. Age and location rarely explain why anyone buys.
  • Asking people what they want. They describe the product they already know. Ask what they did and why.
  • Research by the loudest customers. Reviews and forums over-represent the delighted and the furious; interviews with ordinary buyers balance them.
  • Inventing to fill gaps. A persona with a made-up hobby is less useful than a half-page of verbatim quotes.
  • Researching once. The audience the strategy was built on is not the audience three years later.
  • Keeping the findings in a report. The value is in the segments, the map and the vocabulary that the rest of the plan uses.

The next step in the strategy sequence is segmentation, which decides which of the groups the research found the business will actually serve; competitor analysis covers the alternatives those groups consider.

Frequently asked questions

How many customer interviews are enough?

Enough that new interviews stop producing new themes, which for a single segment is often somewhere in the first dozen or so. The aim is understanding, not statistical significance; a survey can quantify what interviews reveal.

Are personas still useful?

As a summary of research, yes; as a substitute for it, no. A persona built from interviews, analytics and search data is a useful shorthand for a team. One invented in a workshop with a stock photo and a name is fiction.

What is the cheapest research a small business can do?

Read what customers already wrote: reviews of your product and competitors', support conversations, sales call notes, the search terms that bring people to the site. Then talk to five recent customers. Both cost time, not money.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Part of these learning paths

Also related

Related