Audience survey for a magazine or news site: what to ask and how to run one

Write down this year's decisions, then pick from the groups below. A question about frequency is only worth asking if you could change the frequency.

Two of these do most of the work. "Which one section would you keep?" forces a choice, so it separates the sections readers value from the ones they skim. "What would you miss most?" tells you, in their words, what readers think you are for.

Put questions about the reader last, and only the ones you will use to split the results.

Pew Research Center warns that "even small wording differences can substantially affect the answers people provide". Its guide and the AAPOR best practices give five rules:

The Reuters Institute's Digital News Report 2026 found that 17% of people across the 20 countries it tracks pay for online news, and 10% in the UK. Of those who pay, 46% gave social or values-based reasons, such as supporting journalism, as part of why.

"Do you pay for the magazine now?" has a true answer. "Would you pay £4 a month?" is a guess, and readers who like you guess high. Offer realistic choices, as the subscription question in the bank does, and ask payers why they pay: their reasons are your pitch to those who don't.

LION Publishers, an association of independent news publishers, warns in its 2025 guide to first reader surveys that if a survey takes "more than a few minutes, it will be difficult to get a statistically significant response".

Ten to twelve mostly one-tap questions is a sensible ceiling. Put the time in the invitation and check it with the survey length calculator. AAPOR recommends pretesting with people like those who will answer, so time five readers and ask what each question meant to them.

Each channel brings in different readers:

LION advises starting with loyal supporters, such as donors, members and subscribers, because they are the most likely to complete a survey. In one of its case studies, a local news site, donors were almost five times more likely to complete it than non-donors. Fine, as long as you read the results as their views.

Ask how each person reads you, and compare answers by channel. If newsletter and website readers want different things, a single total hides it.

AAPOR lists incentives among the ways to raise response, and LION suggests offering something back, such as a sticker, a postcard or a T-shirt. Each kind changes your results:

For a first survey, skip the prize. Say what you will do with the answers, then publish the results and what you are changing.

For a large readership, about 385 answers give a margin of error of plus or minus 5 points at 95% confidence, and about 100 give plus or minus 10. The sample size calculator works it out for your list size.

The groups are the catch. If 500 people answer but only 60 read you on the website, any figure about website readers carries about plus or minus 13 points. Decide which groups you need before you send, and keep the survey open until each has enough.

When you share results, say how many people answered, through which channels and over what dates, as AAPOR asks.

A reader survey rarely needs to know who answered. UK GDPR requires personal data to be "adequate, relevant and limited to what is necessary", and the ICO says you "must not collect personal data on the off-chance that it might be useful in the future".

To link answers to subscriber records, for example to compare payers with non-payers, talk to your DPO or lawyer first.

Keep questions about single stories out of the yearly survey: "Did you enjoy our March cover story?" gets answers from memory, months later. Ask one short question at the end of every story instead, such as "Was this story worth your time?", and compare sections through the year.

In the example above, that question would show whether the quick weekday recipes you publish next land with everyone who finishes them, not only with subscribers.

Asklet handles the per-story half. One snippet in your story template adds a short survey to every story, and path patterns such as /recipes/* give each section its own questions and results. Readers answer anonymously: the widget sets no cookies, stores the answer with the page path and time, and removes emails and phone numbers typed into free text.

It doesn't run the yearly survey. It shows on a web page, not inside emails, and it can't contact readers or split answers by subscriber. Use a form tool for that. The Free plan covers one website and up to 100 responses a month, with no card needed.