What content research must answer
A content programme can be planned from four findings. First, what people ask at each stage of the FunnelModel of the customer journey stages.: the symptoms they describe before they know the name of their problem, the comparisons they make once they do, the doubts they raise before acting, and the problems they hit after buying. Second, the words they use, because a piece written in the brand's vocabulary will not be found by someone searching in their own, and will not feel written for them when they arrive. Third, what a satisfying answer looks like: how much depth, which format, which evidence, which objections it has to address. Fourth, where they look — which channels, communities and people they already trust on the subject — because that decides distribution as much as production.
Google's guidance on helpful content frames the same requirement from the other side: content that serves people's needs, shows experience and expertise, and leaves the reader feeling they got what they came for. The E-E-A-TExperience, Expertise, Authoritativeness, Trust. signals search quality guidance describes are, in practice, what an audience that has been properly researched recognises as "written by someone who understands my situation".
Segmentation, personas and the journey itself are not this guide's job. The audience research guide in the fundamentals section defines the segments; journey mapping defines the stages and the moments within them. Content research sits on top of both and asks the narrower question: what do these people need to read, watch or hear at each moment?
Sources
No single source gives the whole picture. Interviews explain reasoning but are expensive; search data is cheap and broad but only shows people who already know the words; support logs are honest but skew towards existing customers and problems. A good research sprint uses several and triangulates.
| Source | What it reveals | Blind spot | Effort |
|---|---|---|---|
| Customer interviews | Reasoning, language, the moment the search began, what nearly stopped the purchase | Small sample; recent buyers remember selectively | High per interview, low to arrange |
| Sales call notes and recordings | Objections, comparisons, questions asked before buying, the words prospects use | Skews to people who reached sales; sales framing | Low if calls are already recorded |
| Support tickets and chat logs | Post-purchase problems, confusion the product or content creates, recurring how-to questions | Mostly existing customers; symptom language | Low to export, medium to tag |
| Site search logs | What visitors expected to find and could not | Only visitors who used the search box | Low |
| Search query data | Demand volume, phrasing, related questions, what the results page rewards | Misses the pre-vocabulary stage; no reasoning | Low to medium |
| Communities, forums, review sites, social threads | Unprompted questions and complaints, the vocabulary of the undecided, who people trust | Noisy; loud minorities; needs manual reading | Medium |
| Surveys and polls | Priorities across a larger sample; validation of interview themes | Respondents answer what is asked, not what they think about | Medium |
| Competitor and adjacent content | Which questions are already answered well, and which are answered badly or not at all | Shows supply, not demand | Low |
Start with whatever is already recorded — calls, tickets, search logs — because reading existing material costs nothing to collect, then spend the interview time on the gaps it exposes.
Interviews, sales calls and support logs
Interviews are the only source that explains why. A content research interview is short and structured around the decision rather than the product: what was happening when the person first looked for help, what they typed or asked, what they read and whether it helped, what they compared, what nearly stopped them, and what they wish someone had told them earlier. Ask for stories rather than opinions ("tell me about the last time…") and record the exact phrases, because the phrases become titles, headings and search terms. Include people who did not buy; their questions are the ones the current content fails to answer. A model asked to role-play the audience is no substitute for these conversations, as the guide to synthetic personas and simulated respondents explains, though it can help rehearse the interview guide beforehand.
Sales calls carry the same material at scale. Where calls are recorded or notes are kept, read a batch and tag every question and objection by stage. Patterns appear quickly: the comparison every prospect asks for, the pricing question that the site avoids, the integration doubt that loses deals. Each recurring item is a brief waiting to be written, and the sales team becomes an ally in research if the resulting content is sent back to them to use.
Support logs are the honest record of what customers struggle with after buying. Export a period of tickets or chat transcripts, tag the recurring topics, and separate product problems (which go to the product team) from understanding problems (which content can fix). The second group feeds retention content directly and often reveals gaps earlier in the funnel too: a question that appears in support usually appeared, unanswered, during evaluation.
Search and community data
Search data turns questions into demand. For each theme the interviews surfaced, the query data shows how many people type it, in which phrasings, which related questions appear alongside it and what kind of page the results reward. That last point matters for content research as much as for SEOImproving visibility in unpaid search results.: the results page is a record of what has satisfied people, and reading it tells you the format and depth an answer needs. The keyword research section covers the mechanics of gathering, grouping and classifying queries; the content researcher uses its output as a map of demand by stage, and treats low-volume queries with strong intent as seriously as high-volume ones with none. Where the tools show no volume at all, as in niche business-to-business markets and new categories, the guide to keyword research when volumes are low explains how to infer demand from these sources and confirm it after publishing.
Communities fill the gap search data leaves. Forums, review sites, professional groups and social threads hold unprompted questions from people who have not yet learned the vocabulary, and they show who the audience already trusts. Reading a community is slow and the sample is loud, so use it for language and for the earliest-stage questions rather than for priorities. A useful habit is a standing list of the places the audience gathers, checked briefly each month, with new questions added to the bank with their source and date. Where the brand already runs social listening, its queries and hand-tagged samples can feed the same question bank.
From findings to a content map
Raw findings are a pile of quotes, tickets and queries. The content map is what the calendar is built from, and making it is a tagging exercise. Each question in the bank gets a stage (using the See-Think-Do-Care stages or the team's own), a theme, the audience segment it came from, the source and date, and a rough measure of demand — interview frequency, ticket count, search volume band. Group by theme and the pillars start to show themselves; group by stage and the gaps appear, usually at the earliest stage, where the brand has been writing about its product while the audience was still describing symptoms. Tagging each question with the job the person is trying to get done, as well as the segment, keeps both lenses in the map, and the comparison of buyer personas and jobs-to-be-done explains which planning questions each answers.
The map then records what exists. For each question, is there a page or asset that answers it well, badly, or not at all? Answered well means keep and distribute; answered badly means refresh or rewrite; not at all means brief. This step turns research into a prioritised list rather than a list of ideas, and it is the direct input to the content pillars and calendar guide, which turns the themes into pillars and the prioritised questions into a schedule the team can keep. The content strategy hub shows where the map sits among the other strategy decisions.
A research sprint for a home-insulation installer
An installer with one marketer runs a two-week sprint before planning the year. She reads 180 enquiry emails and 60 recorded sales calls, tags 410 distinct questions, and interviews 9 customers and 4 people who chose a competitor. Search data for the 35 strongest phrasings shows demand concentrated on cost, grants and disruption during installation, with a cluster of early-stage queries about damp and condensation that the site never mentions. The map has 4 themes and shows that 70% of existing pages sit at the decision stage while 60% of the questions sit earlier. The first quarter's calendar is set accordingly: 2 pillars at the awareness and consideration stages, 1 at decision, and a refresh of the 6 grant pages that sales calls showed were out of date.
Keeping research alive
Audience questions change faster than segments do: a new regulation, a competitor's launch, a change in the product or the economy all shift what people ask within weeks. Research done once and filed becomes a description of a past audience. Three routines keep it current without turning it into a project.
The first is a standing intake. Sales and support send recurring questions to one place — a shared document, a form, a tag in the ticketing system — and the content team reviews it fortnightly, adding to the bank with source and date. The second is a quarterly re-read: before each calendar is planned, re-run the search data for the pillars, read the last quarter's tickets and calls, and interview a few recent buyers, then retire questions that have stopped appearing and promote the new ones. The third is a feedback loop from published content: comments, replies, on-page questions and the search terms that bring people to each page show whether the answer matched the question, and often supply the next one.
Dating every entry matters more than it seems. A question bank where each item carries a source and a date can be weighted towards the recent, and old assumptions can be seen for what they are rather than carried forward as fact. The same dating, applied to findings from every study the business runs, is the core of the tutorial on building a research repository.
Common mistakes
- Researching the product instead of the decision. Interviews about features produce feature content; interviews about the moment the search began produce content people look for.
- Search data only. It shows people who already know the words and misses the earliest, most valuable stage.
- Interviewing only happy customers. The people who did not buy hold the questions the content fails to answer.
- Rewriting the audience's words. Their phrasing is the title, the heading and the query; the brand's vocabulary is not.
- Findings without a map. A pile of quotes does not become a calendar; tag by stage, theme and demand.
- One-off research. Questions move within a quarter; keep an intake and a dated bank.
- Skipping what exists. Calls, tickets and search logs are already recorded and cost nothing to read first.
