Skip to content
Roy A. Bradley

What we thought readers wanted

A subscriber survey that asked the wrong question, and what it took to notice.

Client
The Margin Review
Role
Research and writing
Year
2025

Before we watched anyone read, we asked them. Six hundred subscribers answered a survey about what they wanted from the journal, and we spent most of a year building toward the answers. The answers were not wrong, exactly. They were about a different magazine than the one people were reading.

What we asked

Eleven questions, all of them the kind that sound reasonable in a planning meeting: which topics would you like to see more of, how often would you like to hear from us, would you read a longer piece if it were serialised.1

Every question asked readers to predict their own behaviour. None of them asked what readers had actually done.

What we heard

More interviews. More criticism of new fiction. A monthly digest. A serialised long read, which forty-one percent said they would follow to the end.

We shipped the digest and commissioned the serial. The digest’s open rate settled below the weekly letter it was meant to complement, and the serial lost two-thirds of its readers between the first and third instalment.2

What we should have asked

Nothing, probably. The survey cost four months and told us what our readers admired, which is not the same as what they read. A reader who says they want more criticism is telling you something true about their self-image and almost nothing about their Tuesday evening.

The following year we stopped asking and started watching. That study is the second part of this series, and it contradicted the survey on nearly every point that mattered.

What we kept

One question earned its place: when did you last abandon a piece partway through, and do you remember why? It was the only question about something that had already happened, and it was the only one whose answers survived contact with the study.

Footnotes

  1. The full instrument is in the appendix of the 2025 annual report. It is worth reading as a catalogue of leading questions. ↩

  2. Instalment three was the longest of the five, at forty-one hundred words. We assumed at the time that length was the cause. The study the following year suggests it was the absence of a way back in. ↩