How do you estimate work and plan realistically?

Question

Accepted Answer

Estimation is hard because software is full of unknowns. The goal isn't false precision, it's **a realistic, honestly-communicated range** that helps the business plan. Good TLs estimate ranges, surface risk, and re-forecast as they learn.

## How to estimate better

```text
✓ Break work down — small tasks estimate far better than big ones
✓ Estimate as a RANGE or confidence ("2-4 weeks, 70% confident")
✓ Include the invisible work — testing, review, deployment, unknowns
✓ Use history — what did similar work actually take?
✓ Involve the people doing the work — not just you
✓ Separate ESTIMATE from COMMITMENT (deadlines are negotiated, not guessed)
```

## A concrete example

Asked "how long for the new dashboard?", don't say "two weeks." Say: "The known work is about two weeks. There's an unknown around the reporting API, if it's clean, two weeks; if not, closer to four. I'll know more after a one-day spike."

## Manage the padding honestly

Don't secretly inflate estimates, and don't let optimism crush them. Make uncertainty *explicit* so stakeholders can make informed trade-offs instead of being surprised.

## Why it matters

Bad estimates erode trust, both the "always late" team and the "sandbagging" team lose credibility.

Honest ranges plus visible risk let the business make real trade-off decisions, and protect your team from impossible commitments made on their behalf.