How much does feature flag technical debt cost?

It comes down to four numbers only your team has: how many stale flags you carry, how long one takes to remove, what an hour of engineering costs, and how much a stale flag slows people down each month. Most figures that circulate online have no published method behind them. Put your own numbers in and the calculator shows every step, so the total is one you can defend when someone in a planning meeting asks where it came from.

Why does waiting cost only the carrying cost?

Because the number of removals stays put. Clear 35 flags now and remove new ones as they go stale, or wait a year and clear the bigger pile. Every flag still gets removed once, at the same minutes each. Waiting adds the months those flags spend in the code, where someone has to read around them or test both paths, and that carrying cost is what the calculator charges for delay. If your older flags really are harder to remove, raise the minutes per removal.

What counts as a stale flag?

A temporary flag that now serves the same value everywhere. Think of a release flag rolled out to 100% six weeks ago, or one past its expiry date. A flag no service has evaluated in a month counts too. Kill switches, ops toggles and entitlement flags can serve one value for a year and still be doing their job, so leave them out of both counts.

Does a free removal tool always save time?

It saves time from the point where the minutes it saves on each removal add up past the setup. With a one-hour setup and 45 minutes saved per removal, that point is the second removal. A free tool still costs someone's time, though: wiring up the workflow, then a few minutes reviewing each pull request it opens. With a handful of flags, a slow trickle of new ones and a short manual removal, that first hour can take more than a year to win back, and the calculator shows that case.

How do I find my own numbers?

Count active and stale flags in your flag dashboard or with a search for your SDK calls. Take the flags created and removed per month from the last quarter of flag history or merged pull requests. Time the next manual removal from the first search to the merge. For the hourly cost, ask finance for a loaded rate, or use salary plus benefits divided by working hours.

Feature flag cleanup cost calculator

Stale flags cost something. How much depends on your team, so this page asks for your counts and your rates and shows the arithmetic as you type. Clearing the backlog has one price. Then there's the slow monthly bill for leaving those flags where they are. If you're weighing a removal tool, it also tells you when the setup pays for itself.

The starting values are placeholders. Replace every one with your own numbers.

Your flags

Flags in the codebaseOf those, flags safe to removeFor example, rolled out everywhere or past their expiry date.Temporary flags created per monthLeave out kill switches and other flags meant to stay.Flags removed per month today

Your time

Minutes for one manual removalFind the references, delete the dead branch, fix the tests, review, merge.Loaded cost per engineering hour (USD)Minutes a stale flag costs per monthTime spent reading around it or testing both paths. Set 0 if you think it costs nothing.

If a tool opened the removal PRs

Minutes to review one automated removalHours to set the tool up

Clearing today's backlog

$3,500

35 flags × 60 min × $100/h = 35 hours. That is 29% of your flags.

Carrying cost, next 12 months

$5,700

Each month's backlog × 5 min × $100/h, summed. Right now that runs $292 a month.

Backlog in 12 months

83 flags

6 new − 2 removed = 4 more stale flags every month.

Keeping up, per month

$600

6 new flags × 60 min × $100/h = 6 hours a month, removing each one as it goes stale.

Putting cleanup off for a year costs about $5,700 in carrying cost. The removal work itself costs the same whenever you do it, so that figure is the whole price of waiting.

Does a removal tool pay for itself?

Each automated removal saves 45 minutes. The $100 setup is paid back after 2 removals, which your current backlog covers on its own. Over a year (107 removals), that is 79.3 hours back after setup.

Every figure here comes from your inputs. The calculator assumes a removal costs the same whenever it happens. If your older flags take longer to remove, raise the minutes per removal.

How the numbers are worked out

There are five formulas, and the calculator shows the working under each result as you type. All five come down to flags times minutes, priced at your hourly rate.

Clearing the backlog is stale flags × minutes per removal ÷ 60 × hourly cost.
The backlog's direction is flags created per month minus flags removed per month. A positive result is how many stale flags the backlog gains each month, and a negative one counts down to an empty backlog.
Carrying cost is each month's backlog × minutes a stale flag costs per month ÷ 60 × hourly cost, added up across the next 12 months.
Keeping up is flags created per month × minutes per removal ÷ 60 × hourly cost. It is what cleanup costs once the backlog is gone.
A removal tool's break-even is setup minutes ÷ (manual minutes − review minutes), rounded up to whole removals. The hourly rate cancels out of that one, which is why it reads as a count of removals.

The model assumes a removal takes the same time whenever you do it. Delay then changes how long the flags sit in your code, and the removal hours come out the same. If your old flags have tangled themselves into newer code, raise the minutes per removal until it matches what you see.

Why there are no averages here

Search for the cost of flag debt and you'll find confident figures, like a share of flags that never get removed or a fixed number of hours per flag. We went looking for the method behind them. None of the ones we checked publish one, and a figure with no method tends to die the first time a manager asks for the source.

Pete Hodgson described toggles as "inventory which comes with a carrying cost", and this page is built on that idea. All it asks is that you put the carrying cost in minutes, for your own codebase.

Once you have a number

To work through the backlog without breaking production, start with the flag cleanup playbook and its two-PR removal pattern. Keeping the pile from growing back takes a monthly audit. The lifecycle guide lists the filters. Owners and expiry dates set when a flag is created keep the "safe to remove" count honest from day one. The flag expiration guide covers how Featureflip handles both.

Say the break-even comes out in a tool's favor. The Featureflip flag cleanup Action opens one pull request per removal candidate and archives each flag once its removal ships. It comes with every plan, Free included, so its only cost is the setup and review time the calculator already counts. Other tools work differently, and the cleanup tools comparison puts them side by side.

Featureflip finds the stale ones for you

Featureflip marks a flag Stale or Dead from its real evaluation traffic, with the reasons attached, on every plan. Paid plans add cleanup emails and unlimited flags at a flat price. See the pricing page, or browse the other cost calculators.

Start free on Featureflip

Frequently asked questions

How much does feature flag technical debt cost?: It comes down to four numbers only your team has: how many stale flags you carry, how long one takes to remove, what an hour of engineering costs, and how much a stale flag slows people down each month. Most figures that circulate online have no published method behind them. Put your own numbers in and the calculator shows every step, so the total is one you can defend when someone in a planning meeting asks where it came from.
Why does waiting cost only the carrying cost?: Because the number of removals stays put. Clear 35 flags now and remove new ones as they go stale, or wait a year and clear the bigger pile. Every flag still gets removed once, at the same minutes each. Waiting adds the months those flags spend in the code, where someone has to read around them or test both paths, and that carrying cost is what the calculator charges for delay. If your older flags really are harder to remove, raise the minutes per removal.
What counts as a stale flag?: A temporary flag that now serves the same value everywhere. Think of a release flag rolled out to 100% six weeks ago, or one past its expiry date. A flag no service has evaluated in a month counts too. Kill switches, ops toggles and entitlement flags can serve one value for a year and still be doing their job, so leave them out of both counts.
Does a free removal tool always save time?: It saves time from the point where the minutes it saves on each removal add up past the setup. With a one-hour setup and 45 minutes saved per removal, that point is the second removal. A free tool still costs someone's time, though: wiring up the workflow, then a few minutes reviewing each pull request it opens. With a handful of flags, a slow trickle of new ones and a short manual removal, that first hour can take more than a year to win back, and the calculator shows that case.
How do I find my own numbers?: Count active and stale flags in your flag dashboard or with a search for your SDK calls. Take the flags created and removed per month from the last quarter of flag history or merged pull requests. Time the next manual removal from the first search to the merge. For the hourly cost, ask finance for a loaded rate, or use salary plus benefits divided by working hours.