Publish the Methodology, Not Just the Number

A number with no visible method cannot be responsibly cited, and careful writers won’t cite it. They may not even be able to say why they skipped you — the page just didn’t give them enough to stand on.

The methodology section is not academic decoration. It is the part that converts your finding into something another person can put their name behind. Everything else on the page is the headline; this is the load-bearing wall.

What a citable methodology actually contains

You are writing for someone who has to defend their choice to quote you. Give them the answers to the questions they’ll be asked.

The question you asked. Stated as a question, in one sentence, before any result. If a reader has to reverse-engineer what you were trying to find out, they can’t tell whether your answer applies to their argument.
The population, and where it came from. Who or what did you measure? Where did those records come from — your own systems, a public dataset, a survey panel, a list you assembled? Provenance is the single most-checked item.
The time period. Both the window measured and the fact that it’s a window. “Last year” ages badly; a stated date range does not.
Sample size and how the sample was selected. Not just how many but how chosen. Everyone who did a thing, a random draw, whoever answered — these produce different findings, and the difference belongs on the page.
What you excluded, and why. Test accounts, duplicates, records below a threshold, incomplete responses. Exclusions are normal and disclosing them reads as competence, not weakness. Hiding them means someone else’s recount won’t match yours.
A definition for every term you counted. If you report “active accounts”, say what makes one active. Counting requires a boundary, and the boundary is a decision you made.
The known limitations. In their own short section, in plain language.
The publication date, visible on the page. Not buried in a byline.

None of this needs to be long. A tight methodology runs a few hundred words and answers each item once.

Limitations sections increase citations

This surprises people, because it feels like handing over ammunition. The mechanism is simple: a writer who cites you has to describe your finding, in one sentence, accurately. Your limitations section is where you tell them how to phrase it.

A page that says “this covers only businesses with fewer than fifty employees, so it says nothing about enterprise behaviour” has pre-written the caveat. The writer copies the framing, keeps their own accuracy, and moves on. A page with no stated limits leaves them guessing at the scope — and a guess they can’t verify is a risk they’ll avoid by finding a different source.

The same logic applies to results you didn’t expect and can’t explain. Say that you can’t explain them. “We don’t know why” is a sentence that makes the rest of the page more believable, not less.

The self-serving-data problem

If your numbers come from your own customers, users, or logs, say so in the first line of the methodology and again anywhere the finding is summarised.

Your customers are not the market. They selected into your product, which means they differ from everyone else in ways you can’t fully list. That doesn’t make the data worthless — internal data is often the only data that exists on a question — but it changes what the finding is about. A number describing your users is a finding about your users.

Writers notice this, and they notice attempts to blur it. Language like “we analysed the industry” over a dataset of your own accounts is the fastest way to lose a citation you nearly had: it tells the reader the framing was chosen for flattery rather than accuracy. Saying “this is our own book of business, and here’s why that matters” costs nothing and buys you the benefit of the doubt everywhere else on the page.

Reproducibility is a spectrum

You do not have to publish raw data to be credible, and in most cases you can’t — the records are confidential, or personal, or commercially sensitive. Reproducibility is not all-or-nothing. Work up the spectrum as far as you’re able:

The aggregate table, so a reader can see the shape of the result rather than only your summary of it. One row per period, one column per category is usually enough; the point is that the numbers you’re describing are visible somewhere.
The definitions and thresholds, so a person with similar data could count the same way.
The exact questionnaire wording if you surveyed anyone. Question wording changes answers more than most people expect, and a survey without its wording is uncheckable by design.
The processing steps in prose: what got joined to what, what got rounded, what got weighted.

Undisclosed weighting deserves its own warning. If you adjusted your sample to look more like some reference population, that adjustment is part of the finding, and a reader who discovers it later will assume you hid it on purpose.

Version the method so editions are comparable

If you intend to publish this annually, the method becomes a fixed asset in its own right. Give it a version, keep it stable, and when you do change a definition, say so plainly on both the new edition and the old one: what changed, when, and whether the earlier figures were restated.

This is what makes a repeat study more valuable than a one-off. A comparable series lets writers describe a trend, and trends get cited far more often than snapshots. A series where the definitions quietly drift lets them describe nothing, and the second edition undermines the first.

What disqualifies a method page

Three patterns end the conversation for a careful reader:

Circular definitions. Defining a term using the thing you’re trying to measure, so any result is guaranteed.
Undisclosed weighting or adjustment, discovered by inference.
“Based on our analysis of industry data.” This is the absence of a method wearing a method’s clothes. It names no source, no period, no size, and no definition.

What to do next

Method only matters if the numbers are yours to begin with. Start with the original data you already have — most sites are sitting on a countable question and haven’t noticed — then document how you counted it before you publish a single figure.

It helps to know what you’re competing with: a well-documented finding travels because the alternative is usually a statistics roundup full of numbers nobody can trace. Then spend time on how the finding is phrased so people can quote it — a defensible number in an unquotable sentence still doesn’t get used.