For founders
Year-end product audit: five things to cut before January
Pendo's data says 80% of features are rarely or never used. December, when usage is quiet and the roadmap is being written, is when to find out which of yours they are. Five cuts to make before you plan a single new feature.
Prashant Kumar Singh · · 4 min read
In short
- 80% of features in the average software product are rarely or never used (Pendo). Most products are carrying maintenance cost with no return.
- Five cuts: features with no usage in 90 days, settings nobody changes, onboarding steps before first value, dashboard tiles no one clicks, and the pricing tier nobody picks.
- Audit before you plan. A roadmap written on top of unused features inherits their cost.
Pendo's feature adoption research found that 80% of features in the average software product are rarely or never used. The number is old enough to be familiar and still true enough to be uncomfortable. Every one of those features has a cost: code that has to keep working, screens that have to keep making sense, a settings page that keeps getting longer.
December is the right time to do something about it. Usage is quiet, so the data is easy to read, and the 2027 roadmap is being written, so there is a reason to clear the ground first. Five cuts, each with how to find the candidates.
Cut 1: features with no usage in 90 days
Open your analytics and sort features by last-90-day usage. The bottom of the list is the audit. For each one, ask whether a single paying customer would notice its removal. If the answer is no, remove it, or hide it behind a flag and see whether anyone asks. Opinions about which features matter are reliably wrong; usage is not.
Cut 2: settings nobody changes
Every setting is a decision the product declined to make. Check how many users have changed each one from its default. A setting changed by fewer than 2% of users is a default with a UI attached. Pick the better value, make it the only value, and delete the control. The settings page gets shorter and the product gets more opinionated, which is what users mean when they say it feels simpler.
Cut 3: onboarding steps before first value
Count the screens between sign-up and the first moment the product does something useful for the user. Each one loses people. The audit question is whether each step is needed to deliver the first result or merely convenient to collect early. Tours, surveys, setup wizards and "tell us about yourself" screens almost always belong after the first result, not before it.
Cut 4: dashboard tiles no one clicks
Dashboards accrete. Every quarter someone adds a tile, and nobody removes one. Click data on each tile tells you which numbers people act on. A tile that is looked at but never clicked may be fine; a tile nobody looks at is a place to put something they would. The dashboard should fit on one screen and answer one question: does anything need my action today?
Cut 5: the pricing tier nobody picks
Look at the distribution of plans. If a tier has a small single-digit share of customers, it is not a product decision, it is a comparison anchor, and it may not be doing that job either. Three tiers is enough for almost every product. Fold the orphan tier into its neighbour, grandfather existing customers, and watch the conversion on the remaining three.
| Cut | Where to look | Decision rule |
|---|---|---|
| Unused features | Feature usage, last 90 days | Would a paying customer notice? |
| Unchanged settings | Share of users changing each default | Under 2% changed → pick one value |
| Onboarding steps | Screens before first result | Needed for the result, or merely early? |
| Dashboard tiles | Click data per tile | Never clicked → replace or remove |
| Pricing tiers | Plan distribution | Single-digit share → fold into neighbour |
What the audit usually finds
In my experience of doing this on products I inherit, the audit removes between a fifth and a third of the surface area and almost nobody notices, except the engineers, who get a faster codebase, and new users, who get a product that is easier to understand. The features that remain are the ones that were always doing the work. Plan on top of those.
Questions people ask
- What percentage of software features are actually used?
- Pendo's feature adoption report found 80% of features in the average software product are rarely or never used.
- How do you audit a product for unused features?
- Sort features by usage over the last 90 days, check how many users change each setting from its default, count onboarding screens before first value, read click data per dashboard tile, and look at the distribution of customers across pricing tiers.
- When should you do a product audit?
- Before writing the next roadmap. December suits most companies: usage is quiet, the data is easy to read, and the 2027 plan has not been written on top of the old surface area yet.
If you're a founder
Send your product and the free teardown will do the first pass of this audit for you: what to cut, what to keep, what to build first on the cleared ground.
If you run an AI lab or data team
Task sets need the same audit. Tasks that every model passes, tasks nobody can solve, and tasks that measure the same thing twice are the unused features of a benchmark.
Sources