The cheapest feature is the one you never build. It costs nothing to test, host, secure or explain to a new hire.
Pendo's analysis of products using its software found that 80% of features are rarely or never used (Pendo, 2019). That's one vendor's customers, not every product, but it matches what I've seen: most roadmaps are mostly guesses.
Here is the five-step order I follow before anything enters the engineering queue.
1. Find the business problem
Before any technology is named. Which step is manual, expensive or error-prone, what does it cost today, and what is "fixed" worth? A request with no answer here is an idea, not a requirement.
2. Decide what not to build
Weigh every request on business value, cost of delay, effort, risk and strategic fit. The SAFe framework turns part of this into a number, Weighted Shortest Job First: cost of delay divided by job duration (SAFe, 2026). Rank by that, and cut the bottom half without guilt.
| Request | Cost of delay | Duration | Score | Decision |
|---|---|---|---|---|
| Automate invoice matching | 8 | 2 | 4.0 | Build now |
| New reports dashboard | 5 | 5 | 1.0 | Later |
| Custom theme per client | 2 | 6 | 0.3 | Cut |
3. Give the team an operating model
Clear roles, a delivery process and a few engineering metrics. Throughput should come from structure, not from pressure or extra headcount.
4. Automate the step that deserves it
AI, automation or a plain script: whichever is proportionate. Keep people on judgement calls and exceptions; remove the repetitive, rule-based work.
5. Measure what changed
Cost, cycle time, error rate, headcount pressure. If the number didn't move, the initiative was wrong, and someone should say so out loud.
Why a lean stack follows
Every feature you skip removes a library, a table, a background job and a page of documentation. In my experience, a small stack is the result of saying no early, not of picking clever tools later.
Test your roadmap
Score each 0 (no), 1 (partly) or 2 (yes):
- Problem: every item names the cost it removes.
- Ranking: items are scored, not argued.
- Cuts: something was removed this quarter.
- Proportion: the simplest fix was considered first.
- Measurement: each shipped item has a before and after number.
| Score | Next move |
|---|---|
| 8–10 | Keep going; review the bottom of the list monthly. |
| 5–7 | Start scoring requests before planning meetings. |
| 0–4 | Freeze new work for a week and rank the backlog. |
Your next step
Take your ten largest open requests. Write the business problem for each in one sentence. Anything you can't write that sentence for comes off the list today.
Sources
- Pendo (2019). The 2019 Feature Adoption Report. Based on Pendo customers' products.
- SAFe (2026). Weighted Shortest Job First. Scaled Agile Framework.