First, find the decision the study was for and the finding that was meant to change minds
Open the brief before the report. The research design guide wants every study to begin with a decision stated as a sentence that has a verb and a date, plus the sponsor's own account of which finding would make them change course. Read the decision field. "Understand the loyalty opportunity" is a topic, and the guide is blunt about what a brief built on a topic produces: a report, not an answer. If the field names a decision, note who owned it and when it was due. If it names a topic, the first cause on this page is already found, and the rescue starts by attaching the findings to a decision that is still open.
Then find the line about what would change the sponsor's mind. The design guide calls it the point most often skipped and most often regretted, because a sponsor who never said up front what would make them drop a plan will explain away whatever finding puts it at risk. If the line exists, find the one finding in the report that bears on it and note the page it sits on. That page number is the test the inverted pyramid applies to any report: could a reader who stopped after the opening act correctly? A finding on a late slide failed it, however strong it was.
Finally, ask the sponsor one question: what did you decide after the readout? The answers sort the rest of the table. Nothing was ever to be decided; something was, but the answer never reached them; something was, but they stopped trusting the study; they agreed, but nobody owned the action; they could not check the agency's figures; or they decided and the decision was lost. Each is a different cause with a different fix, and the what, so what, now what framework gives every rewrite the same order: what was found, why it matters to the decision and how sure the team is, and what should happen, with an owner and a date.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| The sponsor calls the study interesting and the plan continues unchanged; nobody can say what it was meant to decide | The brief's decision field for a sentence with a verb and a date, and the line on what finding would change the sponsor's mind | No decision named before fieldwork, so no finding could change a plan | Attach the findings to a decision that is still open, with its owner and date, and write the line about what would change minds before a second readout |
| A deck that opens with method and sample and ends with themes, which readers stop reading before the finding | Where the answer sits, and whether the first page states the decision, the answer, the confidence and the main limit | A report written in the order the research was done, with the answer buried | Rewrite answer-first, with findings as statements and their strength, and move the method to the appendix |
| Someone outside the team found a gap in the sample after the decision, and the study is now spoken of as "that survey" | The position and specificity of the limits against the survey design guide's rule that they sit plainly near the top | Limits found by a sceptical reader after the decision, so the whole study lost trust | Put each limit on the first page beside the answer it qualifies, with what it means for the recommendation |
| Recommendations that begin "consider" or "explore", with nobody's name on them | Each recommendation for an action, an owner, a date and a measure | Recommendations with no owner, date or measure | Rewrite each as an action with an owner, a check date and the measure's baseline, or an explicit decision to change nothing |
| Agency slides with figures but no bases, wording or method statement, and questions the team cannot answer | The deliverables against the brief's questions and the commissioning guide's acceptance list | An agency deck accepted without bases, wording or a method statement | Ask in writing for the tables, wording, verbatims and method statement, and re-read the findings against them before acting |
| The same question briefed again later, and the earlier study found in a deck by accident | The repository, or its absence, for the study's finding records and the decisions they informed | Findings left in decks and folders, so the question is researched again instead of answered | Write the study's findings as records with evidence, strength and decisions informed, and make a search the first step of every brief |
Fix, verify and prevent each cause
No decision was named before fieldwork
The first question in the research design guide asks what the business would do differently for each possible result; where nothing would change, the study is decoration, and where the answer is "we will know more", it is curiosity that should be named as such. Naming the decision sets the deadline, sets the standard of evidence and forces the sponsor to say before fieldwork what finding would change their mind. A study briefed on a topic has none of those, so its findings arrive with nowhere to land.
Fix: find a decision the findings bear on that is still open, name its owner and its date, write the finding that would change it, and take the study back to that owner as evidence for that decision rather than as a report about a subject. Where no open decision exists, say so and file the findings as records for the next one rather than forcing a readout.
Verify: the decision owner can state the decision, the date and the finding that bears on it in their own words, and the second readout is on their calendar before that date.
Prevent: fill the decision field of the research brief template on every study, including who makes the decision and what finding would change it, and decline a brief that cannot fill it, as the customer research programme playbook declines requests for evidence to support a decision already made.
The report was written in the order the research was done
The guide to reporting research findings names the commonest shape of an unread report: it opens with the method, walks through every question and ends with a list of themes, so it is read by nobody past the first pages and the decision is made on whatever the sponsor already believed. The inverted pyramid puts a summary first, giving the main finding and the decision at stake ahead of any chart, and the reporting guide sets out what that summary holds: the decision, the answer and the team's confidence with its reasons, then the main limit and the recommendation.
Fix: rewrite the first page to that shape, in a sentence a reader who stops there could act on; turn every theme heading into a finding stated as a sentence, with its evidence and words that match its strength, as the guide's table of evidence types sets out; move the method, the question list and the detailed tables to the back. Then say plainly whether the finding the sponsor said would change their mind appeared.
Verify: a colleague who reads only the first page can state the answer, the confidence and the recommendation, and the finding that bears on the decision is now on that page rather than deep in the deck.
Prevent: draft the summary before the rest of the report on every study, and use the inverted pyramid's test on it before the readout is booked.
A sceptical reader found the limits after the decision
The reporting guide puts it in one line: a sponsor who hears about a sample gap second-hand, once the decision is made, loses trust in the next study as well. The survey design guide sets the rule the report broke: give the base for every figure, show differences between groups only where the sample supports them, say when a subgroup is too small to read, and state the sample's limits plainly near the top, not in a footnote. The professional codes the site's research guides describe from the Market Research Society and ESOMAR expect limits to be reported, not hidden, so a buried limit is a professional fault as well as a tactical one.
Fix: list the specific limits, who the study did not hear from, what the method cannot establish, where the sample sizeThe number of users or sessions an experiment needs to detect a given effect reliably. is too small to separate figures, and what has changed since fieldwork, then put each on the first page beside the answer it qualifies, with a sentence on what it means: the recommendation stands, needs a condition, or needs follow-up research first. Where the limit is a sample that does not represent the people the decision concerns, the diagnosis for survey results that are not representative says what the survey can still claim.
Verify: the sceptical reader, shown the rewritten first page, finds nothing to add, and the recommendation reads with its condition attached rather than without it.
Prevent: write the limits section before the findings section on every study, because a limit found by the team is a caveat and a limit found by a reader is a verdict.
The recommendations had no owner, date or measure
The reporting guide defines a recommendation as an action with an owner and a date, or an explicit decision to change nothing, and lists "consider" items nobody owns among the things a report must not contain. The what, so what, now what framework supplies the discipline: the now-what is what will change, who owns it and when it will be checked, and vague intentions such as "monitor closely" are its named failure. The guide adds that a recommendation which moves a target, a message or a KPIA metric tied directly to an objective. names how its success will be measured and the BaselineAn estimate of what a metric would have done without a given change, campaign or promotion, against which the effect is measured. that measure starts from, so the effect can be judged afterwards.
Fix: rewrite every recommendation through the three questions, name the finding each one rests on, give it an owner who has agreed to own it, a date for the check and the measure with its baseline, and where the honest recommendation is to change nothing, write that as a decision rather than leaving the item blank. Where the team disagrees, say so and give both views for the readout to settle.
Verify: the framework's own test passes: reading only the now-what lines gives a coherent plan with owners and dates, and each owner can be held to their line at the next review.
Prevent: open the next report, and the next readout, with what happened to the last set of actions, which is how the framework closes its loop and how owners learn that their names on a recommendation mean something.
The agency deck was accepted without bases, wording or a method statement
The guide to commissioning a research agency treats a deliverable as accepted once the business can see how each finding was reached, not when the slides land, and it lists what to ask for behind the report: the final questionnaire or discussion guideThe plan for a research interview or group: a warm-up, the episode to reconstruct, open questions about past behaviour and probes, used as a plan, not a script., data tables that show bases and any weighting, anonymised verbatims, the codeframe used for open answers, and a method statement covering who took part, how and when they were recruited, and who was left out. Slides without those cannot be checked or reused, which is why a team that could not answer the first hard question in the readout stopped defending the study.
Fix: ask the supplier in writing for the materials the guide lists, re-read each finding against its base and its exact wording, and check the deliverables against the brief's questions before any further readout; a finding that rests on a small base or a chart without its wording is a reason to ask, not to reject the study.
Verify: every figure the team intends to act on now carries its base and wording, and the method statement names who was missed in the same terms the report uses on its first page.
Prevent: write the deliverables into the brief as findings with their evidence rather than a slide count, judge proposals on what their analysis plan and deliverables will contain, and treat the tables and method statement as the deliverable, with the deck as its summary.
The findings were left in a deck, so the question was researched again
The tutorial on building a research repository lists why research gets run twice: the first study is invisible, its findings sitting in one team's deck; someone finds the deck but cannot separate a finding from a hunch or tell whether it still holds; or the findings are filed against the decision they served, so a team with a related question never sees that they apply. A study that changed nothing the first time is the most likely to be repeated, because nothing in the business carries its memory.
Fix: write the study's findings as records now, one observation per record with its evidence, counter-evidence and strength in words, tagged the same way as the rest of the store, with a field for the decisions each one informed, and add the study record and the brief beside them. Where the study informed no decision, the empty field is itself a finding about the programme.
Verify: a search of the repository on the question the study asked returns its records with their strength and dates, and the next brief on a related question cites them in its "what is already known" field.
Prevent: make a search of the repository the first step of every brief, so desk researchResearch from existing sources, the business's own and published ones, done first so that interviews and surveys cover only what nobody has answered. starts inside the business, and, as the repository tutorial does, name what each study leaves behind in the brief's deliverables field before fieldwork, so that no study closes as a deck alone.
Prevention: a brief with a decision, an answer-first report and a readout that records what was decided
The habit that prevents most of this page is written before fieldwork. The research brief carries a decision with a verb and a date, its owner, and the finding that would change it; the timing field names the decision meeting the findings feed; the deliverables field says what is delivered and to whom and, as the repository tutorial asks, what the study will leave behind in the repository; and the "what is already known" field starts with a search of that repository. The methods section collects the design, commissioning, reporting and repository guidance this page draws on, and a brief built from it cannot name a topic in place of a decision without someone noticing.
The report is written in the reader's order, not the researcher's. The first page states the decision, the answer, the confidence and why, the main limit and the recommendation, and says whether the finding that would have changed minds appeared. Findings are statements with their evidence and words that match their strength; survey figures carry their base and wording, interview themes are reported as reasons rather than shares, and the limits sit beside the answers they qualify. Each recommendation names the finding it rests on, an owner, a check date and the measure with its baseline, or records a decision to change nothing. Where the study came from an agency, the tables, wording and method statement arrive before the readout and are read against the brief's questions. The craft of putting that first page in front of a room, with a single message, a single decision and evidence that survives questions, is the subject of the guide to presenting your work to senior leaders, which opens with the answer and the ask for the same reason.
The readout ends in a record. The reporting guide sends the report ahead so the meeting is spent on the decision, opens with the answer and the limits, and closes either on a decision with owners and dates or on a plain account of what is still missing and who will supply it. The record says what was decided and by whom, which recommendations were taken up, who owns each and by when, the measure and where it starts, and whether the mind-changing finding turned up; the marketing decision log template gives that record its columns, including the check date and the actual effect when it arrives. Teams that run research continuously gather those records into the quarterly readout the customer research programme playbook describes, with a decision log that shows which findings changed which decisions, which is the record the playbook judges the whole programme by. When study after study lands without a decision, the marketing research audit scorecard reviews every stage across the studies of a set period, from the brief to the record of what was decided, and ranks which weakness to fix first.
Two neighbouring symptoms have their own pages. When the limit the reader found is a sample that does not represent the customers the decision concerns, the diagnosis for survey results that are not representative says what the survey can still claim. When the study's interviews produced transcripts but nothing a decision could use, the diagnosis for customer interviews that produce no insight works back through the brief, the recruitment and the guide. Both, and the rest of the marketing research troubleshooting section, end where this page does: with a finding connected to a decision, recorded where the next study will find it.
A garden centre group rescues a study nobody read (illustrative)
Fendrasko Garden Centres
Fendrasko Garden Centres, an invented group of 7 garden centres, commissions 16 interviews and a survey of 1,240 loyalty-card members about a loyalty app. The agency's deck runs to 52 slides, opens with 6 on method and ends with 5 themes. The readout is well received. 5 months later nothing has changed, and the marketing director is asked why the group paid for it.
The brief's decision field reads "understand the opportunity for a loyalty app", with no owner, date or line about what would change anyone's mind. The one finding that bears on a decision, that members who shop at more than one centre would use an app for a single points balance, sits on slide 31. The finance director had noticed in the readout that the survey reached only current members and no lapsed ones, which nobody had said, and had treated the whole study as unsafe since. The deck's 5 recommendations all begin "consider" and name nobody, and the data tables were never requested.
The group attaches the study to a decision that is open: whether to pilot an app at 2 centres before the spring season. The agency supplies the tables, wording and method statement within a week. The report is rewritten with the answer, its confidence and the lapsed-member gap on the first page, the slide 31 finding second, and 3 recommendations, each with an owner, a date and a measure; the head of loyalty owns the pilot, with the share of multi-centre members using the app as the measure and a baseline of zero. The second readout lasts 40 minutes and ends with a recorded decision to pilot at 2 centres from 1 March and a follow-up study of 12 lapsed members. 14 finding records go into a new repository, and the next brief, on the plant guarantee, finds 3 of them and narrows its questions accordingly.
Common mistakes
- Blaming the fieldwork. The interviews and the survey were usually fine; the break is between the brief and the readout.
- A topic instead of a decision. "Understand" produces a report; a verb and a date produce an answer.
- Opening with the method. Readers stop before the finding, and the decision is made on prior belief.
- Limits at the back. A caveat the team buries becomes a verdict a reader delivers.
- Recommendations that "consider". No owner, no date, no measure, no change.
- Accepting slides without tables. A finding nobody can check is a finding nobody will defend.
- Closing the study as a deck. The question will be briefed again, and the repository's empty field for decisions informed is the record of why.
