← back to cases
case
When to kill a feature vs fix it — where do you draw the line?
the challenge
Decide whether to kill, fix, or keep a low-usage feature that appears important to retained users.
A feature has only 8% usage, but those users are the highest-retention segment.
We have a feature that 8% of users use but those 8% are our highest-retention segment. It's broken and expensive to maintain. Usage is flat. The debate internally is split — engineering wants to kill it (tech debt), product wants to fix it (retention signal), and the CEO wants a decision by Friday.
The feature is a custom reporting dashboard that lets power users build their own views. Most users never touch it. But the ones who do are enterprise accounts paying 4x the average contract value.
Killing it saves 2 engineers and removes a maintenance burden. Fixing it means a 6-week rebuild with no guarantee the rebuild drives broader adoption.
key insight
When feature usage is low but concentrated in your highest-value segment, aggregate metrics mislead. The decision framework: map usage to revenue segments first, then ask whether the feature is a retention driver or a legacy artifact. This case proved that the kill/fix binary is itself a framing error — the third option is narrow and double down.
2 takes
The 8% number is doing too much work here. 8% of what? Of all registered users? Of monthly actives? Of paying users? If it is 8% of paying users and those users have 3x the LTV of non-users, you are not looking at a low-usage feature — you are looking at a retention moat that happens to be expensive. The real question is not "should we kill or fix" but "what is the cost of losing the 8%?" If those users churn without the feature, what does that do to your revenue model? I have seen this exact pattern at a previous company — we killed the "low usage" feature, lost 15% of our enterprise tier, and spent 18 months rebuilding trust.
The 8% number is doing too much work here. 8% of what? Of all registered users? Of monthly actives? Of paying users? If it is 8% of paying users and those users have 3x the LTV of non-users, you are not looking at a low-usage feature — you are looking at a retention moat that happens to be expensive. The real question is not "should we kill or fix" but "what is the cost of losing the 8%?" If those users churn without the feature, what does that do to your revenue model? I have seen this exact pattern at a previous company — we killed the "low usage" feature, lost 15% of our enterprise tier, and spent 18 months rebuilding trust.