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

Marketing Research › Research Methods

Reporting Research Findings So They Get Used

How to report research findings: lead with the answer to the decision, show the evidence and its strength, say what the study cannot, and record the decision.
Guideintermediate10 min readUpdated

A report is the answer to a decision

The research design guide starts every study from a decision written as a sentence with a verb and a date, and from the sponsor's statement of what finding would change their mind. The report closes that loop. Its first job is to answer the decision; its second is to show how far the evidence supports the answer; its third is to say what the study could not settle. Everything else, including the method, the full question list and the detailed tables, serves those three jobs.

Writing the report this way is also an ethical point, not only a stylistic one. The professional codes of the Market Research Society and ESOMAR, as the methods hub describes them, expect findings to be reported with their methods and limits rather than dressed up as more certain than they are. A report that hides a weak sample on its last page, or that presents a few conversations as if they measured a market, fails that test however well it is designed.

Part of the reportWhat it holdsWhat it must not do
The answerThe decision, the answer, the confidence and the recommendationOpen with background, method or a list of themes
FindingsEach finding as a statement, with its evidence and its strengthPresent a topic heading instead of a finding
Quotes and clipsMaterial chosen to show each finding fairlyPick the most vivid voice regardless of whether it is typical
What the study cannot sayThe sample's gaps and the questions left openSit in a footnote or an appendix
RecommendationsActions with owners and dates, and follow-up research where neededList "consider" items nobody owns
Method and appendixHow the study was run, who took part, the instrumentsCrowd out the answer at the front

Lead with the answer

The inverted pyramid orders a message conclusion first, then the facts that support it, then the background, so a reader who stops early still leaves with the point. Its page notes that a report built this way puts a summary at the top that states the main finding and what is being decided, before any chart. For research, that summary has a fixed shape: the decision, the answer, how confident the team is and why, the main limit, and the recommendation.

Write the answer as a sentence that would make sense to someone who reads nothing else. "Customers value flexibility" is a theme. "We recommend offering the battery-only package to existing customers first, because interviews show a clear reason to buy and the survey shows interest concentrated among recent installs" is an answer. If the study did not answer the decision, say that first, and say what would.

Then go back to the line written in the brief about what would change the sponsor's mind, as the customer interviews tutorial asks at its final step, and say plainly whether that finding appeared. That sentence often matters more to the decision than any chart.

Findings, evidence and how strong each is

Each finding is a statement, not a heading. "Price" is a heading; "Price became the deciding factor only for people who were comparing quotes at the same time" is a finding. Under each one, give the evidence and say how strong it is in words that match the evidence. The what, so what, now what framework is a useful discipline here, as its page describes: state what was found, then why it matters for the decision and how confident the team is, then what should happen.

The language of strength differs by method, and mixing them is the commonest reporting error.

EvidenceHow to describe its strengthWhat it can support
A theme across many interviews"Came up across most of the conversations", with the counter-evidence namedA reason or pattern worth acting on or measuring
A theme from one participant"A single voice", reported as suchA question or a possibility, not a pattern
A survey figureThe figure with its base and the exact question wordingAn estimate for the population the sample represents
A difference between survey subgroupsThe figures with both bases; whether the difference is likely to be realA comparison only where the sample supports it
Behavioural or analytics dataThe count beside the rate and the periodWhat people did, not why
A published or desk sourceThe source, its date and its definitionContext, with the weight the source deserves

Two rules from the site's research guides carry most of the weight. First, interview findings are never reported as prevalence. The qualitative versus quantitative comparison lists "counting interviews" as a common mistake: a purposive sample cannot estimate a proportion, so a theme is reported as a reason, not as a share of customers. The AI research synthesis guide makes the same point about counts and adds the evidence table, in which each theme carries its participants, supporting excerpts, contradicting excerpts and implication; build that table before the report, whether or not a model helped with the coding, and keep it as the report's backing. Second, survey figures carry their base and wording, as the survey design guide requires, and small numbers are read as the guide to reading small numbers advises: counts beside rates, ranges for the figures that drive the decision, and words such as "too few to tell" where they apply. Whether a difference is likely to be more than noise is a question of Sample Size and significance, which the sample size and significance guide covers; the report states the conclusion in plain words and points to the working.

Quotes, clips and fairness

Quotes and video clips are the most persuasive material in a research report, which is why they need the most care. A well-chosen quote makes a finding real to a reader; a badly chosen one makes a single participant's view look like the market's.

Choose each quote to illustrate a finding that the evidence table already supports, not to create one. Prefer a quote that is typical of the theme over the most vivid one, and where a theme had a strong dissenting voice, show that too. Check every quote word for word against the transcript or notes, as the AI synthesis guide requires; a tidied paraphrase presented as a quotation is a misquote however close the meaning. Label each quote with a participant code and the relevant context, such as their segment, rather than a name.

The codes the methods hub describes also say that answers are confidential unless people agree otherwise. Quotes in a report shared beyond the research team should not identify the person, directly or through detail such as a job title in a small company. Clips show a face and a voice, so use them only where participants agreed to that use, and check with whoever owns data protection before a clip leaves the team.

What the study cannot say

Put the limits where the reader will see them: in the summary, next to the answer they qualify. The survey design guide asks for the sample's limits to be stated plainly near the top, not in a footnote, and the same applies to any method.

The useful limits are specific. Who the study did not hear from, such as people who never became customers or a segment that could not be recruited. What the method cannot establish, such as how common a reason is when the evidence is interviews, or why people behave as they do when the evidence is a survey. Where the figures are too small to separate. What has changed since fieldwork. Each limit should say what it means for the answer: whether the recommendation still stands, needs a condition, or needs follow-up research first.

A limit stated early protects the report. A sponsor who learns about a sample gap from someone else, after the decision, stops trusting the next study too.

From findings to recommendations and owners

A recommendation is an action, an owner and a date, or a decision to change nothing. Where a finding needs measuring before the business commits, the recommendation is the follow-up study, with its question written down. Keep the link visible: each recommendation names the findings it rests on, so a reader can see how far the evidence carries it.

Separate what the data shows from what the team concludes, as the survey design guide asks. The finding is the evidence; the recommendation is a judgement, and the report should be clear which is which. Where the team disagrees about the recommendation, say so and give both views; the readout is the place to settle it.

Every recommendation that changes a target, a message or a KPI should state the measure that will show whether it worked and its Baseline, the level the measure starts from, so the effect can be read later. Where the recommendation is a test, the experiment results report is the template for reporting it when it ends.

The readout meeting and the record

Send the report before the readout, so the meeting is spent on the decision rather than on reading. Open with the answer and the limits, take questions on the evidence, and leave most of the time for the recommendation. The meeting should end with one of two outcomes: a decision with its owners and dates, or a clear statement of what is still needed to decide and who will get it. When a finished study has already been received politely and changed nothing, the diagnosis for research findings that are not acted on works back from the decision it was for through the brief, the order of the report, the limits, the owners, the agency deliverables and the repository.

Then write the record. It holds the decision and who took it, the recommendations accepted and declined, the owners and dates, the measure and its baseline, and whether the finding that would have changed minds appeared. Store it with the report, the evidence table and the original brief written from the research brief template, where future teams can find them, so that the next study starts from what this one settled. The tutorial on building a research repository sets up that store so findings can be found and trusted later. When the date for checking arrives, the result goes back into the same record. Teams that report on a regular rhythm rather than study by study can gather these records into the quarterly readout that the customer research programme playbook runs, with a decision log showing which findings changed which decisions. To review whether reports, readouts and their records hold up across every study rather than one, the marketing research audit scorecard scores them beside briefs, samples and consent and ranks the weaknesses into a fix list.

A solar installer reports a mixed-methods study (illustrative)

Illustrative example — not a real result

Trivoxa Solar rewrites a findings deck to lead with the answer

Trivoxa Solar, an invented installer of rooftop solar panels, ran a study to decide whether to offer battery-only installs to households that already have panels. It held 14 interviews with existing customers and people who had asked for a quote, then surveyed customers on its own list, with 620 responses.

The first draft of the deck ran to 38 slides and opened with 6 slides on method and sample. The finding that mattered sat on slide 24. It also said "9 of 14 interviewees want a battery", presented as a pie chart. The rewrite opens with the decision and the answer: offer battery-only installs, starting with customers whose panels were fitted in the last 3 years, because interviews show a clear reason (using daytime generation in the evening) and the survey shows interest concentrated in that group, 41% against 17% among older installs, with both bases shown.

The interview count becomes a reason, labelled as such, with 3 quotes checked word for word and a dissenting voice about cost. The sample's main gap moves to the first page: the survey reached only existing customers, so it says nothing about households with panels from other installers. The readout ends with the head of sales owning a pilot offer, a launch date 6 weeks out, the battery attachment rate as the measure, with a baseline of zero, and a follow-up question on the households the survey could not reach.

Common mistakes

  • Opening with the method. The answer buried behind slides about how the study was run.
  • Themes instead of findings. Headings such as "price" and "trust" that tell the reader nothing until they read every line.
  • Counting interviews as if they were a survey. Percentages and pie charts from a handful of conversations.
  • Figures without bases or wording. Survey results the reader cannot judge.
  • The most vivid quote, not the typical one. One memorable participant setting the story.
  • Limits at the back. The sample gap found by a sceptical reader after the decision.
  • Recommendations nobody owns. "Consider" and "explore" with no owner, date or measure.
  • No record. The decision taken in the meeting and lost, so the next study repeats the work.

Frequently asked questions

Should a research report start with the method?

No. Start with the answer to the decision the study was commissioned for, then the findings that support it, then what the study cannot say, and leave the method detail for later in the report. The limits that change how the answer should be read belong near the top, not with the method at the back. A reader who stops after the opening should know the answer, how confident to be in it and what is recommended.

How do I report how many interviewees said something?

Carefully, and never as a percentage. Counts in qualitative work describe breadth within the people you spoke to, not how common a view is among customers. Say that a theme came up across most of the interviews, or from a single participant, and present it as a reason or a pattern worth acting on or measuring. If the decision depends on how common the view is, that is a question for a survey with a suitable sample.

What should happen at the end of a research readout?

A decision, or a clear statement of what is still needed to make one. The meeting should end with the decision taken, who owns each action, when the effect will be checked and the measure that will show whether it worked. Record that alongside the report, together with whether the finding the sponsor said would change their mind appeared, so the next study starts from what this one settled.

Terms used on this page

Frameworks

Where this page is used

Sources

  1. Market Research Society — MRS
  2. ESOMAR — ESOMAR

Related