How to collect customer feedback, six ways compared

Collecting customer feedback is not one job, and the six ways people search for are not six versions of the same thing.

Each one reaches a different group of people, and the group it reaches is almost always more important than the question you asked them.

That is the whole argument of this page. A method is not good or bad. It is a filter, and knowing which filter you just applied is the difference between data and decoration.

What is the fastest way to collect customer feedback online?

Put the place to say something on the screen where the thing goes wrong, not on a page people have to find.

That sounds obvious and almost nobody does it, because the collecting surface usually lives on a help site and the problem lives in the product. The distance between those two is where most of your feedback is lost, and it is lost silently, because nobody files a complaint about a form they never found.

Three surfaces work online, in descending order of how much they collect.

A control beside the thing itself. A single button next to the missing feature, on a screen somebody is already using. It collects the most because it costs the least.

A public board. A page anybody can read, where the requests already filed are visible. It collects less than the first but it does something the first cannot, which is let two people discover they want the same thing.

A form on a contact page. It collects from people who were already motivated enough to go looking. That is a small and unrepresentative group, and it is the surface most companies rely on.

How do you collect anonymous feedback, and should you?

Let people say something without an account, and accept that you will not know who said it.

The trade is genuine in both directions and neither side of it is obviously right.

Anonymous collection gets you volume and honesty. People say the blunt thing when their name is not on it, and the ones who would never have created an account get counted. On a board where a signed out visitor gets an identity minted in their own browser, voting and commenting work from the first page load, and that is the whole reason public boards collect anything at all.

What you lose is everything downstream. You cannot tell whether a request came from your largest customer or from somebody who will never pay you. You cannot tell the person it shipped. You cannot check whether one enthusiastic person voted eleven times from three browsers.

The practical answer for most products is both doors. Anonymous for anybody who arrives, and an identified path for customers already signed in to your own product, so their requests carry a name. What an anonymous vote is worth in detail is a page of its own.

How do you collect customer feedback by email?

Reply to something, rather than sending something.

A feedback email that arrives unprompted is an interruption, and the response rate shows it. A feedback question attached to a message the person was already going to read is a different object entirely: the end of a support thread, a receipt, a release note about something they asked for.

Two rules make email work and both are about restraint.

One question, in the body, answerable by replying. A link to a form is an extra step and it halves the answers. If the question genuinely needs a form, say so in one sentence and make it the only link.

Never to everybody. A list send produces answers from the people with the most time, which is not the same group as the people with the most at stake.

Email is also the only channel here that does the closing half well, which is the half most teams skip. On our own board that is wired to the vote rather than to a mailing list: the changelog emails the people whose own vote is in the release, and nobody else is told.

What about by phone, or in person?

Both produce the best feedback you will get and the least of it, and both are worth doing for a reason that is not the obvious one.

A conversation is the only method where you can ask the second question. Everything written gives you the answer to what you asked. A call gives you the answer, then why, then the thing behind it that the person did not know was the point until they said it out loud.

The catch is volume and bias in equal measure. You will speak to a handful of people, and they will be the ones who agreed to speak to you, which is its own filter.

So the honest use of a call is not measurement. It is discovery: finding the request you did not know existed, so that it can then be put somewhere countable. Ten conversations early in a product's life will teach you more than a hundred votes. A hundred votes later will tell you which of those ten things to build first.

In person collection at a desk or an event has the same shape with one extra problem. People are more polite to your face, and politeness is indistinguishable from approval in a notebook.

How do you turn any of this into data you can act on?

Put everything into one list where the same request from two people becomes one row with a two on it.

That is the step that converts feedback into a decision, and it is the step that gets skipped, because merging is manual and nobody owns it. A support inbox, a survey export and a board full of votes are three piles that each look like evidence and cannot be added together.

Three things have to be true of the list.

One row per request, not per message. Eleven people asking for the same thing is a card with eleven on it, not eleven cards.

The channel recorded on each. A request typed into a cancellation form and the same request typed into a survey are not the same evidence, and a list holding only the words has thrown away the half that told you which.

Something to sort by other than recency. A count is the minimum. Demand, effort and priority is the version with the formula printed, because a ranking you cannot recompute by hand is a ranking you cannot argue with.

What decides whether you get a second message?

Whether anybody heard back about the first one.

Every method on this page is about collection, and collection is the easy half. The half that decides your next quarter's volume is whether the person who wrote to you in March ever found out what happened.

Three moments are worth reporting and only three: when you decide not to, when the work starts, and when it ships. The first one is the one everybody skips, and silence and refusal look identical from the outside.

Ten worked examples of real customer feedback, sorted by what each one can actually decide, is the page to read next. If you want the form itself rather than the method, twelve forms with every field written out is that page.