Google's own guidance on helpful content asks whether a page provides "original information, reporting, research, or analysis". AI assistants face the same problem from the other side: when ten pages say the same thing, they cite the one with the specific number. A small business does not need a research department to publish something original. It needs one good question, an honest method and a clear page that presents the answer.
1. Pick a question people already ask and nobody has answered well
The best research topics sit where three things overlap: your customers care about the answer, journalists and bloggers in your field write about it, and you have unique access to the data. Look for questions where the current answers are vague, outdated or based on another country. "How much do Halifax homeowners spend on renovations?" is stronger than "Renovation trends", because a writer can quote the answer in one sentence.
Search your topic plus "statistics" or "survey". If the top results cite data that is five years old or American, you have found a gap.
2. Start with data you already own
The cheapest research is often sitting in your systems. Anonymized and aggregated, it can answer questions no competitor can:
- Booking or job data: average project size, busiest months, lead time by service.
- Quote and pricing records: real price ranges by job type or neighbourhood.
- Support tickets and call notes: the most common problems and questions.
- Website and campaign data: benchmarks such as response times or conversion rates across your clients.
- Before and after measurements from your own work.
Remove anything that could identify a customer, follow your privacy policy and Canadian privacy law, and report only aggregates.
3. Run a small survey properly
If you need new data, a short survey works well. Keep it to eight to twelve questions, most of them multiple choice, with one open question for quotes. Send it to customers, your email list, a professional association or a paid panel. Aim for a sample large enough that one unusual answer cannot swing a result. A few hundred responses is a solid target for a general audience, while a niche professional audience can be meaningful with far fewer, provided you state the number.
Write neutral questions. "How satisfied are you with your current agency?" is fair. "How frustrated are you with agencies that miss deadlines?" produces a headline you cannot defend.
4. Find the three findings worth a headline
Most datasets produce a lot of numbers and a few findings. Look for results that are surprising, specific or useful for a decision. "62% of respondents" means little without context. "Six in ten homeowners got only one quote before hiring" tells a reader something they can repeat. Pick three to five findings and write each one as a plain sentence before designing anything.
5. Publish one findings page that is easy to quote
Put the research on a permanent URL on your own site, not only in a PDF or a social post. Structure it so a journalist, a blogger or an AI system can lift a fact in seconds:
- A title that names the topic, place and year.
- A summary of key findings near the top, each as one sentence with its number.
- One section per finding with a simple chart and a short explanation.
- A methodology section: who was surveyed, how many, when and how.
- A date, an author and a way to contact you for questions.
Charts should be images with descriptive alt text and a text version of the key figure nearby, as covered in our image SEO guide. Add Article or Dataset markup if appropriate, using our schema markup guide.
6. Be transparent about the method and the limits
Methodology is not a footnote. It is what makes the research citable. State the sample size, dates, how respondents were found, and any weighting. Say what the data cannot tell readers. Experienced journalists and editors check this before linking, and a clear limits section strengthens the trust signals described in our E-E-A-T guide.
7. Pitch the findings to people who need them
Publishing is half the job. Make a list of writers who have covered the topic in the last year, local news outlets, trade publications and newsletters. Send each a short, personal note with one finding relevant to their audience and a link to the page. Offer a quote or a short interview. Share each finding separately on social channels over several weeks instead of all at once. Our guide to getting backlinks covers outreach etiquette.
8. Update it and build a series
Research ages. Repeating the same survey every year turns one page into a series with year over year comparisons, which journalists love and which keeps the URL relevant. Keep the same address and update the year in the title, or publish each year on its own page and link them together. Each new edition gives you a fresh reason to contact everyone who cited the last one. Our guide to updating old posts explains how to refresh pages without losing their rankings.
Mistakes that make research impossible to cite
- Hiding the findings behind a form. Journalists and AI crawlers cannot cite what they cannot read. Gate an extended report if you like, but keep the key findings public.
- Rounding away the specifics. "Most people" is less quotable than "71% of respondents". Publish the actual figures.
- Leading questions and convenient samples. Surveying only your happiest customers and presenting it as a market view damages trust when someone notices, and someone usually does.
- No date on the page. Writers need to know whether a statistic is current. Put the fieldwork dates near the top.
- Burying the numbers in images. A chart without the figure in text cannot be searched, read aloud or quoted accurately.
A useful test before publishing: could a reporter write one accurate sentence citing your page in under a minute? If not, simplify the summary until they could.
Sources and further reading
See Google's guidance on creating helpful, reliable, people-first content, its Dataset structured data documentation, and the Office of the Privacy Commissioner of Canada's overview of PIPEDA.
Frequently asked questions
Can a small business really publish original research?
Yes. Many useful studies come from data a business already has, such as aggregated booking, pricing or support records, or from a short customer survey. What matters is a focused question, an honest method and clear reporting, not a large budget.
How many survey responses do I need?
Enough that one unusual answer cannot swing the result. A few hundred responses is a solid target for a general audience, while a small professional audience can be meaningful with fewer. Always publish the sample size so readers can judge it.
Why does original research help with AI search?
AI assistants prefer specific, attributable facts. When many pages repeat the same general advice, a page with a unique, clearly sourced number is more useful to cite, especially if the finding appears in one plain sentence.
Should I publish research as a PDF?
Publish it as a normal web page first. PDFs are harder to read on phones and harder to link to specific findings. You can offer a downloadable PDF as an extra, but the findings page should be the main URL.
How often should I update original research?
Once a year works well for most topics. Repeating the same questions builds year over year comparisons and gives you a new reason to contact everyone who cited the previous edition.