Estimating in two-week increments instead of twelve-month plans
Long-range plans are precise about the wrong things. Increments make the uncertainty visible where it can still be acted on.
Key takeaways
- Estimate the next increment precisely and the horizon in ranges.
- Every increment ends with something demonstrable, not a status update.
- Track forecast accuracy per increment — it is the only estimation feedback loop that works.
Why estimate in increments?
Because accuracy decays with distance. A two-week increment can be estimated from what the team knows now; a twelve-month plan is a series of guesses presented with false precision. Increments keep the uncertainty visible while there is still time to act on it.
The problem with the twelve-month plan
Long-range plans are not wrong because teams are careless. They are wrong because the information needed to make them accurate does not exist until the work is underway.
The damage is not the inaccuracy itself. It is that the plan becomes the thing being managed, so discovering something new looks like failure instead of progress.
What an increment has to produce
Something demonstrable. Not a status percentage, not a document — a working slice a stakeholder can use or reject.
That constraint does most of the work. It forces vertical slices through the stack, surfaces integration problems early, and makes the question at the boundary a real one: is this worth continuing.
How precision changes with distance
| Horizon | What we commit to | Confidence |
|---|---|---|
| Next increment | Specific outcomes | High |
| Next quarter | Ordered themes | Medium, as a range |
| Beyond a quarter | Direction and constraints | Directional only |
Measure the forecast, not just the delivery
At each boundary we record what we planned against what shipped. After five increments, that history forecasts the next quarter better than any estimation technique we have tried.
It also makes the team measurably better at estimating, because the feedback loop is two weeks long instead of twelve months.
Making the horizon honest without a fake date
Stakeholders still need to plan around something. The answer is not to refuse a date, it is to give a range with the confidence attached and to narrow it as increments complete.
We publish three numbers after every increment: what shipped, what the observed throughput is, and the current forecast range for the remaining scope. After four or five increments that range is usually tight enough to plan a launch around, and it is derived from this team on this codebase rather than from an industry benchmark.
The cultural change is that the forecast moving is normal rather than a failure. A range that narrows from twelve to sixteen weeks down to fourteen to fifteen is a project behaving correctly. A single date that never moves until it moves by three months is the alternative most organisations are used to.
FAQFAQ
Frequently asked questions
About the author
Sofia runs delivery across engineering engagements — discovery, estimation, increments and handover. She writes about migrations, pricing models and the undocumented business logic that turns a clean estimate into an overrun.
- Enterprise migrations
- Project estimation
- Delivery management
- CRM and ERP systems
