How to track feature requests from support chats

Short answer. Capture requests where customers say them, in support chats and right after they end. Group different wordings of the same need into one request, keep each customer's own words and who they are, and rank by how many accounts asked rather than by who asked loudest. Keep questions about features you already have on a separate list, and when something ships, tell everyone who asked.

The best feature requests you will ever get are not on a voting board. They are in support chats, said by a customer in the middle of a real problem: "can I export this to CSV, our accountant needs it every month". Most of them are answered, closed, and forgotten. A few weeks later nobody remembers that eleven other accounts asked for the same thing.

Here is how to keep them, with a spreadsheet if you are just starting, and what a support tool can do for you once there are too many to keep by hand.

What to keep for each request

FieldWhy
The need, in one line"Export invoices to CSV", not "CSV". One entry per need.
Every account that askedAccounts, not messages. Five messages from one customer are one vote.
Their own words"Our accountant wants a spreadsheet of what we've paid" tells you the real problem and the scope.
When they askedA request that grew by five this month matters more than one that stopped.
StatusOpen, planned, building, shipped, declined, so you know who to tell what.

Doing it by hand

A spreadsheet with those columns works for the first months. After each chat that contains a wish, add the account and the quote to an existing row or start a new one. Once a week, sort by the number of accounts and read the top rows.

It breaks in three places. The same need arrives in different words ("CSV export", "one file with every invoice", "a spreadsheet for our accountant") and ends up as three rows. Quotes get shortened to a word, so nobody remembers what the problem was. And when the feature ships, nobody looks up who asked.

Ranking: how many accounts, and growing

Rank by the number of accounts that asked, then look at the trend. A request at 14 accounts that grew by 5 this month is a stronger signal than one at 20 that has not moved since spring. Read the quotes before you decide: they show whether people want the same thing, and how big a version would solve it.

Keep a separate list for questions about things that already exist. "How do I export my invoices" when there is an export button is not a feature request. It means the docs or the product don't make it findable, and it is fixed with an answer, not with code.

Ask at the end of the chat too

Customers say what they wish for in the middle of a chat, but more of them will tell you if you ask. A short question when a chat ends, "is there anything you wish the product did?", gets answers in the customer's own words, and those go on the same list.

Closing the loop

When a request ships, email everyone who asked for it. A short note saying what was built and where to find it turns a support chat from weeks ago into a reason to stay. It only works if who asked was kept with the request, which is the step most teams skip.

How Afterchat does it

  • Every chat is read when it closes, and what the customer asked for is added to the requests list, grouped with the same need in other words.
  • Requests are ranked by how many accounts asked, with the trend this month and every customer's own words.
  • When a chat ends, customers can say what they wish your product did, and it lands on the same list.
  • Questions about things that already exist go on their own list, as gaps for the docs.
  • Your coding agent can load a request's brief, with the problem, the quotes and the scope, ask the people who requested it a follow-up question, and, once you approve, email everyone who asked when it ships. See customer support from Claude Code over MCP.

On FanWake, a chat tool for creators, the list has 8 open requests after two weeks. None has been built and shipped through the loop yet, so this is the part we have the least to show for so far.

Questions

What is the best way to track feature requests?

Capture them where customers already talk to you, which for most small SaaS teams is support chat, and keep one entry per need with every account that asked and their own words. A spreadsheet works at first; it breaks when the same need arrives in ten different wordings.

How do you prioritize feature requests?

Start with how many accounts asked, then look at whether that number is growing and who the accounts are. The customers' own words tell you the size of the problem, which a vote count alone does not.

Should we use a public feature request board?

Boards collect votes from the customers who find them, which is rarely most of them. Requests made in support chats come from customers in the middle of a real problem, so they are worth capturing even if you also run a board.

How do you tell customers a feature they asked for shipped?

Keep who asked with each request, then email them when it ships, in plain words and with a link to how it works. It is the cheapest way to show customers you listen.

Read next

Buyer's guide AI customer support for SaaS: a checklist for small teams What to look for in an AI support chat when your team is small and answers its own customers: sourced answers, honest handovers, learning from every chat, knowing what to build, and working with your coding agent. Coding agents How to use Claude Code for customer support How I run customer support from Claude Code over MCP: the setup, what the agent can do, what waits for approval, and 825 real tool calls on FanWake. Knowledge base An AI support chatbot trained on your docs and your code How to give an AI support chatbot your own knowledge: why crawling your docs is only the start, how short checked entries built from your docs, code and chats answer better, and how to keep them true.