How the 2027 survey works

It is a voluntary survey of people who find their way to AuthorAZ. That is a self-selecting group, and no amount of responses makes it a random sample of independent authors. Nobody knows the size or shape of that population, so nobody can weight a sample to match it.

So the results describe the authors who answered. They will be reported that way — never as “indie authors earn X”, always as “X% of the N respondents who answered this question”. Where a question has too few answers to say anything, that will be stated instead of a percentage.

The value compounds with repetition. One year is a snapshot of a self-selected group; the same questions asked the same way across years show what is changing within that group, which is a genuinely useful thing even without a representative sample.

18 questions, of which 2 are required and the rest optional. 17 are marked longitudinal: their wording is fixed and will not be revised between editions, because rewording a question silently breaks the comparison it exists to support.

New questions may be added in later years. When a question changes in any way that affects comparability, the change will be noted alongside the results rather than quietly absorbed.

The full question set is published before the survey opens, so you can read exactly what is asked before deciding to take part.

The survey does not ask for your name, pen name, book titles, country, age, or website. Revenue and spending are ranges. Location is a broad region with a “prefer not to say” option. There is no account, no login and no cookie identifying you.

If you give an email address to receive the report, it is written to a separate table with no foreign key back to your answers. This is structural rather than a promise: there is no column joining the two, so the answers cannot be attributed to an address even by us, and adding such a column later would be a visible schema change rather than a quiet one.

The consequence, which is worth stating plainly: because responses carry no identifier, they cannot be edited or withdrawn after submission. There is nothing to look up.

Submissions are limited to 3 per client per day. That check uses a salted hash of the network address combined with the current date — never the address itself — so the same visitor produces a different value tomorrow. Those rows are deleted after 24 hours.

That is deliberately weak as identification and sufficient as rate limiting. It stops a script submitting a thousand times; it cannot be used to build a profile, to link two submissions across days, or to recover an address if the table were ever exposed.

Answers are validated on the server against the published question set. Values that were never offered as options are rejected rather than stored, so aggregate counts cannot be skewed by a hand-crafted request.

One question is free text, optional, and capped at 1000 characters.

Free-text answers are never published automatically. Anything quoted in the report is read by an editor first, with identifying detail removed, and quotes are used only where they illustrate something the numbers already show.

Alongside the results, every edition publishes:

• the number of valid responses;
• the dates the survey was open;
• where respondents were recruited from;
• how many responses were excluded, and why;
• how incomplete responses were handled;
• the exact wording of each question;
• the denominator for every percentage;
• the limitations of the sample.

Percentages are reported as whole numbers against a stated denominator. A figure like “41.37%” from a few hundred responses implies a precision that does not exist, so it will not appear.

The 2027 edition runs from 1 December 2026 to 31 January 2027. Money questions are asked in USD, converted approximately by the respondent — one reporting currency, so the answers can be aggregated at all.