Original research

Original research is the only capability in the educate category that produces something other people cite. That's the whole argument for it, and it's also why the standard is different: a blog post that's wrong is a bad post, while a benchmark that's wrong is a number that circulates for years with the company's name attached.

The operational shape

Trigger. A question the market argues about with nothing underneath it, or data the company holds that nobody else does. The second is rarer and worth more.

Reads. Whatever can actually be measured. The argument, from the market map.

Cadence. ⚑ Open. Quarterly benchmark against annual flagship is an unresolved question in the ContentOS spec, and the cost-to-impact tradeoff has not been validated. This page describes the capability, not a schedule.

Gate. Approve before publish, and the strictest treatment applies to every number — statistics sit behind a named human without exception.

What good looks like

The method is on the page. Sample size, how respondents were found, what was excluded and why, and what the question actually said. Research without a method is an assertion with a chart, and the readers most likely to cite it are the readers most likely to check.

The sample survives division. A finding that holds at n=400 and evaporates when split by company size was never a finding. State what the data can't support as clearly as what it can.

The raw counts are published, not only the percentages. "68 percent" from 25 responses is 17 people, and a reader deserves to know that without doing the arithmetic.

It says something the company would rather it didn't. Research that only ever confirms the positioning gets read as marketing, usually within a year, and that reputation doesn't recover. The inconvenient finding is what makes the convenient one believable.

The date is in the title. Not the metadata. This is the only content type where the date is part of the claim.

When it goes stale

It's dated by design, not by decay — which makes it the easiest type to handle honestly and the easiest to handle badly.

Handled well, an old benchmark stays useful forever as a point in a series, and the second edition is worth more than the first because it can show movement. Handled badly, an undated three-year-old figure keeps circulating as current, and the company that published it gets the blame for every misuse.

Two triggers worth watching. The next edition ships, at which point the previous one gets a link forward rather than a deletion. And the figure starts being quoted without its date, which is a signal the page isn't making the date prominent enough.

What it feeds

• Citations — the reason this capability exists; it is the surface that gets referenced by press, analysts, and answer engines
• Every other page's evidence — a claim with a number behind it, sourced to work the company did, outranks a claim without one
• Sales conversations — a benchmark is the rare asset that a prospect will forward internally

Where ContentOS fits

As of September 2026 this capability is in build. The open question is cadence, not whether — and the ⚑ on it is deliberate, because a research programme committed to before the first edition ships is a commitment made without data about its own cost.