How do you improve engineering processes?

Question

Accepted Answer

Improve process by **finding the real bottleneck with data, changing one thing, and measuring the effect** — not by importing a methodology wholesale. Process exists to serve the team; when it stops serving, change it.

## A loop for improvement

```text
1. MEASURE       → where does work actually slow down? (cycle time, review wait,
                    deploy frequency, incident rate)
2. FIND THE CONSTRAINT → optimizing anything but the bottleneck is wasted effort
3. CHANGE ONE THING → small, reversible experiments
4. MEASURE AGAIN → did it help? keep, adjust, or revert
5. INVOLVE THE TEAM → they know the pain; they'll own the fix
```

## Example

If features take three weeks but only two days are coding, the constraint is *wait time* — code review queues, environment availability, approvals — not developer speed. Adding pressure on coding would do nothing.

## Pitfalls

- **Cargo-culting** another company's process ("let's do it like Spotify").
- **Process for its own sake** — ceremonies that outlived their purpose.
- **Changing five things at once**, so you can't tell what worked.
- **Imposing process top-down** without the team's buy-in.

## Why it matters

Small process improvements compound.

Cutting review wait time or flaky-test pain frees hours every week across the whole team.

Equally important, involving the team turns process from something done *to* them into something they own — which is the difference between a change that sticks and one that quietly gets abandoned.