Ever participated in a backlog grooming or an estimation meeting that took forever? Ever worked on a project where leadership judges you for delivery only against your original estimates? Read on!
If so, you are not alone. This post has some ideas I have found to work most of the times. Have you applied any or all of these ideas before? What is your go to idea?
References: http://bit.ly/3HVbuE8 and LinkedIn post [By Alistair Cockburn]
Thesis: Estimation is unavoidable, but the win is making it cheap, decision-oriented, and revisited on signals (not on a calendar).
1) Start with an appetite (timebox)
Decide what the work is worth before shaping it.
-
Example: “Two weeks max; scope flexes to fit.”
-
Effect: Shifts the conversation from perfect prediction to value and trade-offs.
2) Shrink variance by slicing
Smaller batches beat clever estimates.
-
Slice work into small increments (one week or less) with a clear and just-enough Done criteria.
-
Limit WIP so cycle times stabilize.
-
Visible outcome each slice → easier to steer.
3) Forecast quickly (no precision theater)
Use reference-class data or a rough throughput range; avoid point-level fantasy.
"We ship X points per two weeks" or “We ship ~3 to 5 slices per week” is enough to shape scope inside the appetite.
-
If discovery risk is high, reserve capacity for spikes up front.
4) Map assumptions
Identify what might fail, and decide what to test first, set expectations with dependency partners:
Map known dependencies and external approvals needed.
Identify failure modes - what if a dependency is delayed, or is poor quality?
Create fail-fast mitigations - incremental, small wins surface progress and issues early.
5) Re-estimate on triggers, not calendar
Define the few moments that must force a rethink. When a trigger fires: pause, resize, or reorder. No blame, just a new decision:
-
A new dependency or external approval appears.
-
Defect rate spikes or quality gates fail twice.
-
Rollovers repeat (same slice slips twice).
-
Major scope change, staff change, or >1 day blocked.
6) Operate in flight
Keep delivery light and boring:
-
Stale-signal review in standup: any trigger fired?
-
Adjust scope before the appetite breaks.
-
Celebrate “stopped early” when value capped out.
Tiny templates
Appetite Card
-
Timebox: ___ (e.g., 2 weeks)
-
Success looks like: ___
-
Must-have by week+1: ___
-
Nice-to-have if time remains: ___
Trigger Card
Re-estimate if any of the following happen:
-
New dependency or approval discovered
-
Defect spike / quality gate fails twice
-
Same slice rolls over twice
1 day blocked, or major scope/staff change
_____
Goto metrics (to learn, not to punish)
WIP and blocked time
-
Cycle time distribution (is it tightening?)
-
Hit rate vs appetite (did slices land on time?)
-
Trigger count per initiative (are risks front-loaded?)
Red/Yellow/Green to next milestone
Risks, Dependencies, Mitigations that need to stay on track
TL;DR
Estimate to decide. Reduce variance with slicing. Revisit on triggers, not the calendar.
That’s how estimation becomes a lightweight steering tool instead of precision theater.
No comments:
Post a Comment