A research, survey or data release must state the sample, the method, the geographic scope, the dates the data covers, the source dataset, the key findings with their numbers, the limitations, and where the full material can be read. When the data is genuinely original, it is one of the strongest release formats for search and AI visibility, because it gives journalists and AI assistants something they cannot find anywhere else. Without the method and limitations, the same numbers are just claims.
Why is research the strongest format?
Most releases announce something that happened. A research release publishes something that is known only because your company measured it. That gives other people a reason to reference you: a journalist writing about water damage in Dubai homes needs a figure, and if yours is the only published one with a clear method, it is the one they use.
AI assistants work in a similar way. They have little reason to cite another generic summary of information available from many stronger sources. An original figure with a stated basis is different. It can be retrieved, attributed and checked. None of this guarantees a citation or a ranking; it makes your information worth citing if a system is looking for it. The post on how AI assistants find and cite companies covers what is and is not in your control.
What must a research release include?
- Sample. How many records, respondents or observations, and what they were.
- Methodology. How the data was collected and classified, and how figures were calculated.
- Geographic scope. Where the data comes from, as precisely as it supports.
- Field dates. The period the data covers, or when a survey was run.
- Source dataset. Whose data it is: your service records, a commissioned survey, a public dataset you analysed.
- Key findings. The main results with their numbers and their base.
- Limitations. What the data does not show and who it does not represent.
- Supporting material. A web page or download with the full findings, charts and method.
For a survey, add how respondents were recruited, whether the sample was weighted, and the wording of the key questions. For internal data, say what was excluded and why.
What does a worked example look like?
This example is taken from our press release guide. Acme Plumbing and every figure below are fictional, used to show structure, not real findings.
Headline
Acme Plumbing Analysis of 4,812 Dubai Callouts Finds Flexible-Hose Failures in 31% of Recorded Water Emergencies
Sixteen words. The number in the headline is the sample, and the percentage is the finding. The headline post explains when numbers belong there.
Summary
Dubai-based Acme Plumbing analysed 4,812 residential emergency callouts completed in 2025. Its internal service data found flexible-hose failures were recorded in 31% of the water-damage emergencies included in the analysis.
Opening paragraphs
DUBAI, UAE, 25 September 2026 — Acme Plumbing LLC has published an analysis of 4,812 residential emergency plumbing visits completed across Dubai between January and December 2025. The company's service records found flexible-hose failures were recorded in 31% of the water-damage emergencies included in the analysis.
The analysis included completed residential emergency jobs categorised by the attending technician. Follow-up visits associated with the same incident were excluded. The findings represent Acme's own service activity and should not be interpreted as a census of all plumbing incidents in Dubai.
The second paragraph is what makes this a research release rather than a marketing claim. It says how jobs were classified, what was excluded and what the data cannot tell you. Stating a limitation plainly makes the finding more credible, not less.
Should the findings go in a table?
Yes, for anything with more than two figures. A short key-facts table is easy for readers to scan and for retrieval systems to extract.
| Metric | Result |
|---|---|
| Emergency visits analysed | 4,812 |
| Flexible-hose failures | 31% |
| Water-heater failures | 22% |
| Leaking valves | 18% |
| Period covered | January to December 2025 |
Every percentage should have a clear base. "31% of the water-damage emergencies included in the analysis" is precise; "31% of leaks" is not.
What are the most common mistakes?
- No method. Numbers with no stated source or sample are indistinguishable from guesses.
- Overstating scope. Your customers are not the whole market. "Acme's service records show" is honest; "Dubai homes suffer" is not supported by one company's data.
- Percentages without a base. Say what the percentage is a percentage of, and how many that is.
- Cherry-picking. Publishing the one striking figure and hiding the rest invites the question of what the rest showed.
- Loaded survey questions. If the question steered the answer, the finding is not worth publishing.
- No source page. The release is a summary. The full data, charts and method need a permanent home on your site.
- Stale data presented as new. Give the field dates, and do not publish data so old that it no longer reflects the market.
- A quote that repeats the numbers. The quote should interpret: what the finding means for customers, or what surprised the people who collected it.
What should a research release link to?
Link the company name to your homepage on first mention, and link the report title to the source page that holds the full findings and method. That page matters more than the release itself: it is what journalists cite, what search engines index as the original, and what AI systems can retrieve and attribute. The first-party source page post covers what to put on it and how to structure its URL. Keep to three links in total.
If you have the data but would rather an editor turned it into a release, you can send the findings to us and the release is reviewed before it is filed.
Frequently asked questions
Does a survey need a large sample to be worth publishing?
It needs a sample large enough to support the claim you make, and an honest statement of its size. A survey of 200 customers can be useful if you say it is 200 customers and do not present it as the whole market.
Can we publish research based only on our own customers?
Yes. Internal service or sales data is often the most original material a company has. Say plainly that it reflects your own customers and activity, as the Acme example does.
How often should we publish data releases?
When you have new data. A recurring report with the same method, published quarterly or annually, is especially useful because readers can compare results over time.
The next post in the series moves from release types to search: why entities matter more than keywords.
media:on Editorial. Written and reviewed by the media:on editorial team, who edit every release before it files.