Writing up research: findings, limitations and recommendations people can act on (Research and Evidence for Better Decisions) | READOS

Research is only useful if it changes what people know or do. A careful survey, a set of well-run interviews or a rigorous evaluation can still fail at the final step, when the results are written up. Reports that bury the answer under pages of method, list every number collected without saying what it means, overstate weak evidence, hide inconvenient results or end with vague recommendations are read briefly, filed and forgotten. Meanwhile, decisions are made on impressions, and the effort that went into the research is wasted.

Writing up research well is a skill distinct from doing it. It requires deciding what the findings actually are, ranking them by importance to the decision, presenting the evidence for each in a form readers can follow, being honest about what the research cannot show and turning findings into recommendations that someone can act on. It also requires knowing the audience: a board, an operations team, a customer and a funding body need different amounts of detail and different emphasis, even when the underlying research is the same.

This article explains how to plan a research report around the decision it informs, structure it so the answer comes first, write findings that are tied to evidence, describe methods and limitations honestly, use tables, charts and quotes well, frame recommendations, review the report before release and present findings in meetings. It closes the series on research and evidence for business decisions, with a worked example.

Start with the decision and the reader

Before writing anything, answer three questions:

What decision or question does this research inform? The research question set at the start, as described in the series article on choosing a research approach, should anchor the report.
Who will read it, and what do they already know? A managing director may need a two-page summary; an engineer implementing a change may need the detailed data.
What should readers do or understand after reading? If the answer is unclear, the report will be too.

Reports written for the writer, recording everything that was done in the order it was done, read like diaries. Reports written for the reader start with what the reader needs.

Put the answer first

Business readers are busy, and many will read only the first page. An answer-first structure puts the main conclusions at the start and supports them in the body. A typical structure is:

Summary: the question, the main findings, the key limitations and the recommendations, on one page.
Background and question: why the research was done and what it set out to answer.
Method: what was done, briefly, with detail in an appendix.
Findings: organised by importance or by question, each with its evidence.
Limitations: what the research cannot show and how that affects confidence.
Recommendations: specific actions, owners and next steps.
Appendices: questionnaires, detailed tables, technical notes and data sources.

The summary should stand alone. A reader who sees nothing else should know what was found, how confident to be and what is recommended.

Write findings, not results

A result is a number or an observation: 62% of respondents rated delivery as good or very good. A finding says what the results mean for the question: delivery is broadly satisfactory, but satisfaction falls sharply for orders delivered to regional sites. Readers need findings, supported by results.

Good findings are:

specific, stating who, what and how much;
tied to evidence, with the supporting data, quotes or observations shown or referenced;
ranked, with the most important first; and
qualified to match the strength of the evidence.

Headings that state findings, such as “Regional customers wait twice as long for deliveries”, tell readers more than headings that name topics, such as “Delivery results”. Readers who skim the headings alone should be able to follow the argument.

Match the language to the evidence

The words used should reflect how strong the evidence is. “Customers who received the new service rated it higher” describes an observation; “the new service improved ratings” claims a cause; “the new service will improve ratings across all regions” makes a prediction. Each requires stronger evidence than the one before.

Useful conventions include:

describing associations as associations unless an experiment or strong comparison supports cause;
giving sample sizes alongside percentages, especially for small groups;
reporting ranges or confidence intervals for key estimates, not just single figures;
saying when a finding comes from a small number of interviews and illustrates, rather than measures, a pattern; and
separating what the data shows from the writer’s interpretation of it.

Hedging every sentence makes a report unreadable, while stating everything with certainty makes it untrustworthy. The aim is calibration: confident where the evidence is strong, cautious where it is weak and clear about which is which.

Describe methods briefly and honestly

Readers need enough method to judge how far to trust the findings: who was asked or observed, how many, how they were selected, when and how data was collected, what was measured and how it was analysed. Most of this fits in a short section, with questionnaires, sampling details and analysis notes in appendices.

Method sections should report what was actually done, not what was planned. If the response rate was lower than expected, a question was changed partway through or one site could not be visited, the report should say so. Readers who discover such gaps later lose trust in the whole report.

Limitations: an honest account of what the research cannot show

Every study has limitations, and stating them is a sign of competence, not weakness. Common limitations in business research include:

sample limitations: small samples, low response rates or groups that are under-represented;
self-selection: respondents who chose to take part may differ from those who did not;
measurement limitations: self-reported data, questions that may have been misunderstood or records of uneven quality;
design limitations: no comparison group, a short period of observation or other changes happening at the same time;
generalisability: findings from one site, segment or period may not apply elsewhere; and
researcher influence: interviewers known to participants, or analysts with a stake in the result.

A good limitations section does more than list problems. For each significant limitation, it explains how it might affect the findings, in which direction if known, and what was done to reduce it. It also distinguishes limitations that change the conclusions from those that merely qualify them. A list of every conceivable weakness, without any assessment of which ones matter, is not useful to readers.

Tables and charts that inform

Tables and charts should each answer a question. Practical guidelines:

Title each chart with its message, such as “Complaints fell after the new checklist”, rather than a description such as “Complaints by month”.
Choose the chart for the comparison: bar charts for comparing categories, line charts for change over time, scatter plots for relationships between two measures and histograms or box plots for distributions.
Show counts as well as percentages, and state the number of responses or cases.
Start bar chart axes at zero, so bar lengths are not misleading.
Sort categories meaningfully, usually by size rather than alphabetically.
Avoid decoration, such as three-dimensional effects, heavy gridlines and unnecessary colours.
Round sensibly: whole percentages are usually enough for survey results.

Tables suit readers who need exact values; charts suit readers who need to see patterns. Detailed tables belong in appendices, with summary tables or charts in the body.

Quotes and examples from qualitative research

Quotes from interviews and open-ended responses bring findings to life and show readers what lies behind the numbers. They should be chosen to illustrate a pattern found across several participants, not because they are the most dramatic. Each quote should be short, attributed only in general terms, such as “workshop supervisor” or “regional customer”, and checked to ensure it does not identify the speaker, as the series article on research ethics explains. Saying how widespread a view was, for example “raised by seven of the twelve customers interviewed”, helps readers judge its weight.

Reporting unexpected and negative results

Research often produces findings nobody expected, including findings that the sponsor would prefer not to hear. An initiative may not have worked, a favoured product idea may have little customer interest or a long-held belief about customers may be wrong. These findings are among the most valuable research produces, because they prevent costly mistakes. They should be reported with the same care and evidence as welcome findings, framed constructively and discussed with the sponsor before wide release, so that the report informs rather than ambushes.

Recommendations people can act on

Recommendations turn findings into action. Weak recommendations, such as “improve communication” or “consider reviewing delivery processes”, give readers nothing to do. Strong recommendations are:

linked to findings, so readers can see the evidence behind each one;
specific about what should be done, by whom and by when;
prioritised, with the most important or urgent first;
realistic about cost, effort and the organisation’s capacity; and
honest about uncertainty, recommending a trial or further research where the evidence is not yet strong enough to justify a full commitment.

Where recommendations involve significant spending, a summary of expected costs and benefits helps decision-makers, using methods such as those in the cost-benefit analysis for business decisions article. Recommendations should also say how success will be measured, so that the next round of evidence is built into the change.

Writing clearly

Plain, direct writing serves research reports well:

use short sentences and familiar words;
define technical terms the first time they appear;
use the active voice where it makes clear who did what;
put one main idea in each paragraph, starting with its point; and
cut words that add nothing, such as “it is important to note that”.

Reading the report aloud, or asking someone outside the project to read the summary and say what they understood, quickly reveals unclear passages.

Reviewing before release

Before a report is released, several checks reduce errors and misunderstandings:

Check the numbers: totals add up, percentages match counts and figures in the text match tables and charts.
Trace each finding to its evidence and each recommendation to its findings.
Ask a colleague to challenge the conclusions, looking for alternative explanations and overstatement.
Check confidentiality: no participant can be identified from quotes, small groups or details.
Confirm the summary matches the body of the report.

For important reports, a short review by someone with research experience is worth the time it takes.

Presenting findings in meetings

Many findings are communicated in meetings rather than read. Presentations should follow the same answer-first logic: the question, the main findings and the recommendations, followed by supporting evidence. Each slide should make one point, stated in its title. Leaving time for discussion matters, because questions often reveal how findings will be interpreted and what further evidence decision-makers need. The reports that change decisions article explains how information is filtered and reshaped as it moves up an organisation, and how to keep the important messages intact.

Keeping research findable

Research has value beyond the decision that prompted it. Storing reports, questionnaires, data and analysis notes in a known place, with a short summary of the question, method, findings and date, lets future projects build on earlier work rather than repeating it. A simple research register, listing past studies and their main findings, helps new staff and future projects find what the organisation already knows.

A worked example

This is an illustrative example. The business and figures are hypothetical.

A building products supplier commissions research into why repeat orders from trade customers have fallen. The research includes a survey of 180 trade customers, with 64 responses, interviews with 12 customers and an analysis of two years of order and delivery records.

The first draft of the report runs to 40 pages and follows the order in which the work was done: survey method, every survey question with a chart, interview summaries and finally the order analysis. The main conclusions appear on page 33.

The revised report starts with a one-page summary. It states three findings, in order of importance. First, delivery reliability for regional customers has fallen: order records show on-time delivery to regional sites dropped from about 90% to about 75% over the period, while metropolitan delivery stayed steady, and eight of the twelve interviewed customers raised late regional deliveries unprompted. Second, customers value the supplier’s technical advice, which most rated highly and which several interviewees described as a reason for staying. Third, survey responses suggest price is a lesser concern than expected, although the summary notes that customers who have already left were under-represented among respondents, so price may matter more to them.

The limitations section explains the 36% response rate, the likely under-representation of lost customers and the absence of data on competitors’ delivery performance. The recommendations are specific: investigate the regional freight contract change identified in the records, with the logistics manager reporting within six weeks; set a regional on-time target with monthly reporting; and interview a sample of lapsed customers to test the price finding before any pricing change. The managing director reads the summary, asks two questions about the freight data and approves all three actions at the next management meeting.

How this applies to a small Australian business

Start with the decision the research informs and the reader who will make it.
Put the answer first, with a one-page summary that stands alone.
Write findings as statements, ranked by importance and tied to evidence.
Match the language to the strength of the evidence.
State limitations honestly and explain how they affect confidence.
Title charts with their message and show counts as well as percentages.
Choose quotes that represent patterns, with care for confidentiality.
Make recommendations specific, with owners, dates and measures of success.
Check numbers and confidentiality before release.
Store reports and data where future projects can find them.

Common mistakes

Burying the answer after pages of method.
Listing results without saying what they mean.
Claiming causes from associations.
Omitting limitations, or listing them without assessing them.
Choosing dramatic quotes that do not represent the pattern.
Hiding unwelcome findings.
Vague recommendations with no owner or timeframe.
Percentages without sample sizes.

Frequently asked questions

How long should a research report be? As short as the decision allows. A one-page summary, a body of a few pages and appendices for detail suit most business research. Length should come from the importance of the decision, not the amount of data collected.

Should limitations go in the summary? The key limitations should, briefly, because they affect how confident readers should be in the findings. A summary that states findings without any qualification invites overconfidence.

What if the research did not answer the question? Say so clearly, explain why, report what was learned and recommend what would be needed to answer it. Inconclusive research still narrows uncertainty and prevents decisions based on false confidence.

Should recommendations be separate from findings? Yes. Keeping them separate lets readers judge the evidence on its own terms and see the reasoning that links evidence to action. Some readers will accept the findings but weigh the recommendations differently, and that is a legitimate part of decision-making.

Who should see the report before release? The sponsor, to check that it answers the question and to prepare for difficult findings, and a reviewer who can challenge the analysis. Participants who were promised a summary of results should receive one.

Questions to ask

What decision does this report inform, and who will make it?
Could a reader who sees only the summary act on it?
Is every finding supported by evidence shown in the report?
Do our words claim more than the evidence supports?
Have we explained which limitations matter and how?
Does each recommendation have an owner, a date and a measure of success?

Bringing it together

Writing up research is the final, decisive step in turning evidence into better decisions. Reports work when they start with the decision and the reader, put the answer first, present findings tied to evidence, match their language to the strength of that evidence, state limitations honestly, use charts, tables and quotes that inform, and end with specific, prioritised recommendations. Checked carefully, presented clearly and stored where they can be found, research reports become part of how an organisation learns, not just a record of what it once asked.

Source: KEVOS editorial notes. Examples and figures in this article are illustrations.