Original research: the one thing a small product can publish that earns links for years

A how-to guide competes with forty other how-to guides. A study competes with nothing, because the figure exists in one place. When somebody writes about the subject next year they need a number, and yours is the number.

• It earns links without asking, because citation is how writing works.
• It gets quoted by AI assistants for the same reason: a page with a figure and a method answers a question no other page can.
• It gives your sales pages something true to point at.
• It is the one page a rival cannot copy, only replicate, which costs them the same week it cost you.

Four sources, in order of how easy they are to reach. Most founders have all four and have never looked at any of them this way.

Pick the question before the source. Write the headline you hope to publish, then ask whether your data can honestly produce it. If the answer needs a stretch, change the question, not the analysis.

Publishing customer data is a trust decision before it is a marketing one. The rules are short and absolute.

• Nothing identifies a customer: no names, no domains, no figures so specific that one account is obvious.
• Report shares and distributions, not individual values.
• Suppress any group small enough to be re-identified. A category with two members is not a category.
• Say in the piece that the data comes from your own users, aggregated. Never imply it is a neutral sample of the world.
• Check your own terms and privacy policy allow it. If they do not, change the policy first or use a crawl instead.

This is a reference document, not an article. Someone will land on it, take one figure and leave, and that is a success.

1. 1

   Title: the finding and the population

   "Half of the status pages we crawled share infrastructure with the service they report on" beats "The state of status pages 2026".

2. 2

   Key findings, four or five lines

   Each a complete sentence with its figure, readable without the rest of the page. This block is what gets copied.

3. 3

   Method, before the charts

   What you collected, when, how many, what you excluded and why. Putting it high signals that you are not hiding anything.

4. 4

   One chart per finding

   Labelled axes, stated units, and an alt text that says the finding in words so it can be read without the image.

5. 5

   The caveats, in your own voice

   Where the sample is biased, what you could not measure, what would change the answer. Writers trust a page that argues with itself.

6. 6

   A licence line and the raw table

   Permission to reproduce with a link, and the underlying table available. Both raise the chance of being cited correctly.

Put it at a stable path, with the year in the page rather than the URL: yourproduct.com/research/status-pages, updated each year, keeps every link it ever earned.

Being citable is a writing problem. Every finding should survive being lifted out of your page and dropped into someone else's paragraph.

• Write each finding as a full sentence including the population: "of the 1,000 status pages we crawled", not "of those we looked at".
• Round consistently and say how you rounded.
• Give the figure once, in one form. Two versions of the same number in one page is how misquotes start.
• Name the study in a way people can say out loud, and use that name everywhere you mention it.
• Keep the page fast and readable without scripts, because the crawlers that feed AI answers do not run JavaScript.

One product mention is enough, at the end, saying who made the study and what the product does. Anything more and the page reads as an advertisement, which is exactly the read you are trying to avoid.

The second edition is worth more than the first, because it turns a figure into a trend, and a trend is a story a writer can cover twice.

1. 1

   Repeat the method exactly

   Same population, same definitions, same exclusions. If you must change one, publish both versions for the overlap year.

2. 2

   Keep the URL

   Update the page in place. Archive the previous edition at a dated path and link it from the current one, so old citations still resolve.

3. 3

   Lead with the change

   "Up from a third last year" is the new headline. The level was interesting once; the movement is interesting every time.

4. 4

   Tell everyone who cited you

   The people who quoted last year's figure have a reason to update their page, and that is the easiest link email you will ever send.

Two editions and you own the subject. Three and people stop looking for an alternative source, which is the whole point of doing the work.