Original research is data you collect yourself and publish for others to use: a survey of your customers or industry, an analysis of your own anonymised numbers, or a test you ran and documented. Because you are the only source, anyone who quotes the number has a reason to link to you.
What counts as original research?
You do not need a research budget. Three sources work for a small team:
| Source | What you do | Startup example |
|---|---|---|
| Survey | Ask a defined group 5 to 10 clear questions | 150 online shop owners on how they handle returns |
| Your own data | Aggregate and anonymise what your product already records | Which days of the week invoices get paid fastest in your app |
| Test or experiment | Run a repeatable test and publish the method | Load speed of 20 popular site builders, measured the same way |
A real finding changes what someone believes or does. “Our users like us” is not one.
Why does it matter for SEO and AI search?
Google’s own questions for judging helpful content include whether a page provides original information, reporting, research or analysis, and whether it goes beyond the obvious. A page with your own numbers meets both. Pages that rewrite other sources do not.
It also compounds. Journalists, bloggers and AI assistants quote a statistic along with its source, so one study gets cited in many places (see statistics and quotes). Each citation can bring a link and a mention, and the study feeds your digital PR for months.
How do you run and publish original research, step by step?
- Start with the question a reporter would ask. Write down three questions your buyers and the press care about, and pick the one nobody has answered with data.
- Choose the audience and the size. Define who you ask (say, owners of online shops with fewer than 10 staff) and how you reach them: your customer list, a LinkedIn post, a community, or a paid panel. Aim for a sample you can defend. State plainly how many answered, and don’t say “the industry” if you reached 40 people.
- Write neutral questions. Keep them short, avoid leading wording, offer an “other” option and test the form on five people first. Use Tally or Google Forms.
- Collect, then clean. Remove duplicates and obvious junk in Google Sheets. Keep the raw file.
- Find one headline number. Look for the surprising split: a percentage, a gap between two groups or a change from last year. Check each number twice.
- Publish a data page. Put the headline finding at the top, then the charts, the full method (who, when, how many, how asked) and a download of the data. Give the page a stable URL. Add a short “how to cite this” line and consider Dataset markup so Google Dataset Search can list it (Google requires a
nameand adescriptionfor that). - Make charts people can reuse. Clean charts with the source printed on them get copied into articles. The free pie chart generator from Fe/male Switch helps for share-of-total results.
- Pitch it. Send the finding to ten to thirty relevant journalists and bloggers with the chart attached, as laid out in digital PR. A short press release on your own site and a newswire posting give it a second path to discovery.
- Repeat yearly. An annual edition gives writers a reason to update their links and gives you trend data nobody else has.
How do you measure the result?
- Referring domains to the data page in Search Console’s Links report or your backlink tool. See backlink monitoring.
- Citations by name. Set a Google Alert on the study title and search for the headline number.
- Rankings and traffic for the topic the study covers.
- AI citations. Ask assistants the question your study answers and see whether they cite you.
- Sales conversations that mention the study, which you can log in your CRM.
What mistakes should you avoid?
- A tiny or biased sample presented as fact. If you surveyed your newsletter subscribers, say so. Honest limits earn more trust than a big claim.
- Hiding the method. No method, no citations. Journalists check it.
- Leading questions. They produce numbers that reviewers pick apart.
- Burying the finding. Put the number in the headline and the first paragraph.
- Paying for links to the study. Paid links need
rel="sponsored"orrel="nofollow"under Google’s link spam policy. - Publishing and walking away. Without a pitch, the page gets no readers. Budget as much time for outreach as for the research.
- Handling personal data carelessly. Anonymise results and ask for consent where your law requires it, such as under GDPR in the EU.
Where can you learn more?
The free digital marketing module of the Fe/male Switch startup school covers how a startup finds the questions customers ask. Next steps on SEO100: give the result a home as a linkable asset, and turn the numbers into a visual with infographics.