In The B2B Numbers Census, a website-content audit of 614 B2B companies observed in September 2026 and published in October 2026 by ClaimTheStory Research, 7 in 10 companies used someone else’s numbers, and 1 in 3 never published a number from their own data.
When you run a survey of your own, publishing the numbers is only half the work. If readers, analysts and journalists are going to trust what you found, your report has to show how the survey was run, what each figure counts, what the responses support and where the limits are. This guide walks through the six parts of a survey report, in the order of the sections below: summary, methodology, findings, limits, conclusions and publication.
What a survey report is, and what it must let others check
A survey report states the question a survey asked, how it was run, what the responses show and what they don’t support. Each of its six parts lets an outside reader check something different:
- Summary: the main finding, how many responses it counts, and who was surveyed.
- Methodology: who was invited, when fieldwork ran, how many responded, and what evidence justified removing any responses.
- Findings: each group’s share against its base, the exact question wording, and how small groups were handled.
- Limits: who is missing, how non-response could affect the numbers, and what the question didn’t measure.
- Conclusions: that every takeaway points to a stated finding and stays within its limits.
- Publication: a clean public table, the release rules and a complete methods note, for journalists and analysts to inspect.
Internal and published reports share those parts. A published report also has to give readers outside your company enough detail to verify the method and the figures themselves. The last section of this guide, Publishing the report: what others need to check and reuse it, covers that. Designing the questionnaire, running fieldwork and cleaning raw responses are covered in survey research for content; this guide starts once the data is clean.
Open with the main finding and its base
Lead your report with the main finding, not the methodology. But state the base (how many responses the figure counts) and who the respondents are in that same opening summary.
Avoid
“59% of pest control companies raised prices.” No base, no period, and the whole industry instead of one vendor’s customers.
Use
59.0% of 312 customer companies (one response each, fielded 2–16 September 2026) said they raised residential prices between 1 January and 30 June 2026. Customers of one software company, not the whole industry.
The strong opening has four parts:
- Who acted: the respondents’ companies.
- On whom: their residential customers.
- What is counted: completed responses after cleaning, one per company (base 312).
- The period: 1 January to 30 June 2026.
Repeat the exact wording of the measure wherever the figure appears again in the report, and never put a number in the summary that isn’t in your findings section.
Write the methodology section
Write one methodology paragraph that covers nine items, in this order:
- Who the survey was meant to describe
- Where invitees came from
- How they were invited (and how duplicate responses were prevented)
- Exact fieldwork dates (kept separate from the period asked about)
- Total invitations sent
- Completed responses received
- Responses removed during cleaning, with the evidence for each removal
- Final retained base
- Completion rate among those invited, with its formula
- Invited 2,400 customer accounts, one unique link each.
- Completed 340, a completion rate among those invited of 14.2%.
- Removed 17: the respondent was not an owner or manager.
- Removed 11: failed an instructed-response item and finished in under a third of the median time.
- Final base 312 companies.
Use the exact name completion rate among those invited whenever you cite that percentage. Never call a sample “representative” without saying what it represents and how you checked.
Report a margin of sampling error only when your responses come from a probability sample, drawn at random (probability-based panels count). Even then, say that it leaves out non-response and other errors. In the pest control example, every customer account was invited and 14.2% completed the survey. The respondents chose themselves, so the report gives no margin of error and deals with non-response in the limits section instead. Whenever a finding depends on a specific question, quote that question word for word.
Report it
The responses come from a probability sample drawn at random, including probability-based panels. Say that it leaves out non-response and other errors.
Don’t report it
Respondents chose themselves, as in this survey (14.2% of those invited completed it). Deal with non-response in the limits instead.
Present each finding with its base, definition and group sizes
Every share in your findings needs four things beside it: what is counted, the base, the exact question, and how the other answers (here No and Not sure) are treated in the base.
Set one minimum base before you look at the results, and apply it the same way in every table and sentence. The pest control example uses an illustrative minimum of 30 completed responses per group, a choice for this dataset, not a standard. One size group falls below 30, so it’s merged with its smaller neighbor into 6 or more (base 90, 70.0% said yes). Merging keeps every respondent in the total.
In the published report, say only that groups below 30 are merged with the neighboring smaller size group. Never publish the merged groups’ own bases, counts or shares: a reader could take either one away from 6 or more and recover the other.
Name what your columns can’t answer instead of guessing. The pest control survey counts how many companies said they raised prices, not how much prices rose or why. Answering those would need size-of-increase and reason fields, and this questionnaire didn’t collect them. A share of companies that raised prices isn’t a measure of price growth.
Pair the table with one chart: a horizontal bar for each published size group (1–2, 3–5, 6 or more), with the group’s base in its label (base 118, base 104, base 90). Turning data into content covers choosing a chart and explaining it in words.
State the limits honestly
Put the limits in the main text, not in an appendix, and tie each one to the finding it affects. Cover six points, in this order:
- Who was invited, and who wasn’t: the list the invitations came from, and who that leaves out.
- Who responded: the completion rate, and the chance that those who answered differ from those who didn’t.
- Question wording: what the question asked, and what it didn’t.
- Self-report: whether answers were checked against records.
- Period: the period the question covers, and when fieldwork ran.
- What the measure leaves out: what the data can’t show.
A limits paragraph at the bottom doesn’t fix a headline that still describes the whole industry.
Draw conclusions only the data supports
Keep three things apart at the end of your report: a finding (what the responses show), an interpretation (a possible reason) and a recommendation (an action tied to a stated finding).
Supported
More of the larger respondents reported raising prices (50.0% at 1–2 technicians, 70.0% at 6 or more).
Qualified
Larger operators may face higher costs; the survey didn’t measure costs.
Dropped
“Small operators are losing pricing power.” The data covers six months of price rises, not pricing power.
Publishing the report: what others need to check and reuse it
Before publishing, separate your internal working table from your public table, so the suppressed groups never slip into the published report.
Set a clear publication policy with the data owner before releasing anything: publish group-level shares and bases only, and keep individual responses and company names private. Assess disclosure risk with the data owner too. A minimum base of 30 responses never makes data anonymous on its own.
Publish
Group-level shares and bases, the exact question, and the methods note.
Not published
- Individual responses and company names.
- The merged groups’ own bases, counts or shares.
Publish a complete methods note beside the report, so journalists, analysts and peers can check every detail. It covers 12 items:
- Who ran the survey and for whom
- Recruitment method (and how one response per company was enforced)
- Fieldwork dates and the calendar period asked about
- Exact question wording and answer options
- Counts of invited, completed, removed (with the evidence for each removal) and retained responses
- Completion rate among those invited, with its definition
- Denominators and how Not sure or missing answers are treated
- How small groups are merged, without giving the merged groups’ own figures
- Weighting (or confirmation that results are unweighted)
- Limits of the sample, self-report and measure
- Suggested citation line
- How readers can request data access
A transparent methods note makes your survey easy to inspect, but it doesn’t guarantee media coverage, backlinks or citations. If the survey is for press outreach, PR surveys covers commissioning and judging a survey, and data-led PR covers press angles and the press-ready package. If you’re reporting recurring operational records, see how to create a benchmark report. Our hub on data-driven content marketing shows how all the guides connect.
If you already publish research, we find your next research question, one only your company could answer, and check what has already been published on it before you commit. Book a demo