Research Report Compilation

How do I compile a research report from lots of different sources?

Inventory everything first, before reading any of it for content, and give each input a source code you will use in every later reference. Then work objectives, evidence hierarchy, findings, narrative, architecture, evidence map, gaps, visuals, draft, QA, in that order. The sequence matters: deciding how sources rank when they disagree, before you know which findings you like, is what stops the hierarchy becoming a justification for a preferred story.

Should a report follow the questionnaire order?

Usually not. Questionnaire order is the default only because it requires no decision, and it reliably produces chapters that end without conclusions. Report in the order the argument needs. The honest exceptions are a tracking study read side by side against a fixed template, a tender or regulatory format specified by the buyer, a data book or appendix, and the case where the instrument already followed the journey that is the natural narrative. In the first two the narrative moves into the executive summary and chapter openers rather than disappearing.

What is answer-first communication in research reporting?

Leading with the conclusion and then the evidence. "62% of respondents prefer messaging" leaves the reader to do the interpretive work. "Messaging is the dominant communication preference, chosen by nearly two thirds of respondents" does that work for them and is exactly as true. The limit is strict: never let the framing claim more than the evidence carries. "Customers have abandoned phone and email for messaging" is overreach, because preference was measured and behaviour was not.

What makes a good report headline?

A claim, not a topic. "Customer Satisfaction" is a label. "Satisfaction falls sharply after customers experience their first service issue" is a headline: it states the finding, it can be disagreed with, and it is safe if someone quotes it with no chart attached. Test that it is true of the page beneath it and nothing more, and watch for causal words the design does not license. One primary idea per page.

How do I make a report written by several people read as one voice?

Mechanically, not stylistically. Fix one term per concept, one base and rounding convention, one set of confidence words, one headline grammar, and one level of abstraction per section, and write them down in a conventions register before drafting begins. Matching prose style without matching these conventions produces a report that sounds uniform and contradicts itself.

What do I do when the story needs a section the data cannot support?

One of four things, and never a fifth. Fill it by going back to the analysis if the material exists. Reframe the section at the level the evidence does support. Demote the question to a "what we could not establish" section, where it becomes the strongest part of the next brief. Or cut it and adjust the narrative. Writing across the gap with confident connective prose is the commonest way an honest project produces a dishonest report, and it is most tempting when the narrative is otherwise elegant.

Two analysts gave me different numbers for the same thing. Which do I use?

Neither, until you know why they differ. Investigate the discrepancy to its origin, which is usually a different base, a different filter, a different rounding point, an excluded "don't know", or reworded questions between waves. Each of those is itself worth reporting. Never resolve it silently, and never adopt the more convenient figure. If the origin cannot be established, state the range, name both sources, and flag it for a researcher.

How should I handle client-supplied data and previous waves in a report?

Label client-supplied figures as client-supplied at every appearance, not only the first, because they have not been through your analysis and their provenance is not yours to vouch for. Date material from earlier studies and note its method. Before any wave comparison, confirm the question wording, base definition and method match. Where they do not, report the waves separately and say why rather than drawing a trend line across an inconsistency.

How do I keep evidence traceable when a report is compiled from many inputs?

Build the evidence map during assembly rather than afterwards. Each row names the claim, its location in the report, its level (finding, interpretation, insight, implication, recommendation), the source codes it rests on, the base and question reference, and its confidence. Reconstructed traceability is memory dressed as record. Every recommendation gets its row populated backwards to the findings beneath it, and one that cannot be traced backwards is deleted rather than defended later.

When should I not start compiling a report?

When the analysis is not finished. Compilation over incomplete analysis disguises the gap rather than surfacing it: the architecture creates slots, the slots demand content, and content gets written from the strongest available fragment. The tell is the sentence "we can put a placeholder in for now". Also wait if data is still in field, if the dataset's quality is genuinely in doubt, or if the conclusion has already been fixed by a stakeholder and your task is to arrange evidence behind it.

What should the QA pass at the end actually check?

Every major claim: each headline, each executive summary sentence, each recommendation, each number in a chart title or callout, each quote, and each comparative, trend or significance claim. Walk the chain back to source and confirm the number, base, question, direction and wording. Then sweep the whole document for the same number reported identically everywhere, one term per concept, and every objective either answered or listed as unanswered. This is the author's own pass and does not replace independent review.