Guide

Feature requests: why they mislead and how to track them.

A feature request is a customer's guess at a fix. Track the problem behind it instead: their exact words, what they do today, who they are and whether you asked. Then count companies per problem, not votes per request.

Updated

What they asked for

A Slack alert when a report changes

↓ the problem underneath

Caught out by changes before the Monday meeting

Why do feature requests mislead?

Because the customer is doing your job for you, without the information you have. When someone asks for a filter on the reports page, they are telling you about a problem and guessing at the answer in the same sentence. The problem is usually right. The guess often is not.

Requests also arrive unevenly. A customer who talks a lot asks for more. A large customer's request travels faster through the company. And a request you led them to with "would it help if we built X?" sounds the same in a list as one somebody raised on their own, though it means much less.

Kept as a list of requests, all of that turns into a count that looks objective and is not.

What should you ask when you hear one?

One more question is usually enough. Any of these turns a request into a problem you can work with:

  • "What were you trying to do when you noticed it was missing?"
  • "What do you do today instead?"
  • "How often does that happen?"

The second one matters most. A workaround someone built in a spreadsheet is proof the need is real, because they paid for it with their own time.

A feature request template

For anyone who talks to customers: sales, success, support, founders. Fill it in straight after the call, while the words are fresh. It pastes into a form, a doc, a ticket or a spreadsheet row.

Feature request

What they asked for:
[The request, as they put it]

What they said, word for word:
"[Exact sentence]"

What they were trying to do:
[The job or moment where it came up]

What they do today instead:
[The workaround, and roughly how often]

Who:
[Company, person's role, plan]

Raised it themselves, or answered our question?
[Themselves / We asked]

Call and date:
[Type of call, date, link to the recording or notes]

The first field is the only one most request forms ask for. The rest is what makes the request useful later.

What does a good entry look like?

A made-up example:

What they asked forA Slack alert when a report changes
Word for word"I find out a report's been updated when my boss asks why the numbers moved."
Trying to doPrepare the Monday numbers for the leadership meeting
Instead todayOpens every report on Monday morning to check what changed, about an hour a week
WhoNorthvale Systems, head of operations
Raised itThemselves, before we asked about reports
CallRenewal call, 9 October

The request is a Slack alert. The problem is being caught out by changes before an important meeting. That problem might be solved by an alert, a weekly digest, or a changes view on the report itself, and you only see those options because the entry kept the problem.

How do you track them over time?

Keep every entry in one place, and file each one under a problem, not a feature. "A Slack alert", "an email when it changes" and "a changelog on the report" all go under one problem: people get caught out by changes they did not see.

Then, per problem, count how many companies raised it, and how many did so without being asked. That count is what goes into planning, not the number of rows and not the size of the loudest customer.

ProblemRequestsCompaniesWithout being asked
Caught out by changes they did not see1496
Exports have to be cleaned up by hand2141

The made-up numbers above show the usual pattern: the problem with the most requests is often not the one the most customers have.

The guide How to analyse customer feedback from your calls goes through grouping and counting step by step.

Should you let customers vote?

A public voting board is a good way to hear from many users at once, especially when most of them never talk to you. If thousands of people use your product without ever talking to you, it earns its place.

Know what a vote measures, though. It counts who found the board and agreed with a request someone else wrote. It carries no reason, no workaround and no sense of how much it hurts. And the customers you talk to every week rarely visit the board, so what they say on calls is missing from it.

If you use a board, treat votes as one more source of entries, and file them under problems with everything else. Choosing between kinds of tool? See How to choose a customer feedback tool.

How do you close the loop?

When something ships, tell every company filed under that problem, including the ones who asked for something different. "You told us in October that changes caught you out before your Monday meeting. Reports now show what changed since you last looked." That message is the best reason a customer has to tell you more next time. Closing the customer feedback loop has messages to copy for each step.

Can the logging happen on its own?

The template works when people fill it in. After a busy week of calls, most do not, and the requests that get logged are the ones someone remembered.

Ordagrann reads the customer calls you record in Granola, groups every problem and request into themes, counts how many companies raised each one, and turns a theme into a one-page brief your team can build from. It reads every customer call you approve, so nothing depends on someone filling in a form.

Get started

Works with Granola Business. Any other transcript can be pasted in.