GameCraft · Tools

Milestone and Assumptions Planner

15 checks across four sections. Nothing is stored — your answers live in the URL.

Nobody believes your date. They believe that you know what would break it.

A plan with no uncertainty in it reads as a plan written by someone who has not shipped. The credibility is not in the confidence — it is in naming the two or three things that could move the date, what each one costs, and what you would see first if one started to happen.

The cheapest credibility available

If you do one thing from this: write down what your date depends on, put a number in weeks against each item, and have someone who was not in the planning room read the list. It takes a few days, almost nobody does it, and it changes how every subsequent conversation about the date goes.

01 · The milestone

What it is, who owns it, and whether the date came from the work or was handed to it.

The date came from the work, not the other way round
An estimate built up from the remaining work, then compared against any fixed external date.
Usually missing: The direction. Most dates are set by a funding round, a publisher, or a conference, and the work is then declared to fit. That is a legitimate constraint and a dishonest plan — the honest version states the imposed date, states the bottom-up estimate, and names the gap.
Done when
- A bottom-up estimate exists and is written down, even if a fixed date overrides it.
- The gap between the two is stated in weeks.
- The plan says what is cut to close the gap, rather than assuming it closes itself.
The milestone has a definition, not a name
A written statement of what is true when it is reached, checkable by someone who did not build it.
Usually missing: Anything checkable. "Vertical slice" and "alpha" mean different things in every studio and often different things to two people in the same room. The argument about whether it was hit arrives at the worst possible moment, which is the week it was due.
Done when
- The definition names what a person can see or do in the build, not a percentage complete.
- Someone outside the team has read it and agreed it is unambiguous.
- It has not changed since the plan was agreed, or the change is dated and explained.
One person is accountable for it
A named person who reports on it and is expected to raise a problem early.
Usually missing: The name. Milestones owned by a discipline or a team are owned by nobody, and the first person to notice it slipping has no obvious duty to say so. Accountability is what converts a private worry into a reported risk.
Done when
- One person is named per milestone, and the team knows who.
- That person has reported at least one problem without being asked.
- They have the authority to trade scope, or a named person who does.
Milestones are ordered by risk, not by convenience
The riskiest unknown is scheduled early, where being wrong is still cheap.
Usually missing: Risk ordering entirely. Teams schedule what is easy to start or what a discipline is ready for, so the genuinely uncertain thing lands late — at the point where discovering it does not work costs the whole plan rather than a sprint.
Done when
- The two riskiest unknowns are named.
- Both are scheduled in the first half of the plan.
- There is a written reason for anything risky that is deliberately late.

02 · Assumptions

What has to be true. The section that produces the artifact teams keep using afterward.

Each assumption carries a cost in weeks
A number next to each one, for what happens to the date if it breaks.
Usually missing: The magnitude. An assumption without a number cannot be prioritized, so every risk gets equal attention, which means the one that matters gets no more than the one that does not.
Done when
- Every assumption has a week count.
- The counts differ from each other — if they are all the same, nobody estimated them.
- The largest is at least twice the smallest, or there is a reason why not.
The list is revisited at every milestone
A dated review where assumptions are confirmed, retired, or added.
Usually missing: The revisit. The list gets written once for a funding conversation and then goes stale, so by the second milestone it describes risks that resolved months ago and misses the ones that emerged since.
Done when
- The list has been reviewed at least once since it was written.
- At least one assumption has been retired or added.
- The review is on the milestone checklist rather than remembered.
Someone outside the estimating room has read them
A named person who was not in the planning session, and what they challenged.
Usually missing: The outside reader. Assumptions written by the people who set the date reliably omit the uncomfortable one, not through dishonesty but because it is invisible from inside the plan.
Done when
- A named person outside the team has read the list.
- They challenged at least one item.
- The challenge is recorded, whether or not it changed the plan.
The assumptions are written down
A list of things that could turn out false, each one able to move the date.
Usually missing: The list. Nobody believes your date; they believe that you know what would break it. This is the cheapest credibility available in a plan and almost nobody does it, which is exactly why doing it is worth so much.
Done when
- Every listed assumption could plausibly be false.
- Anything that cannot be false has been removed — it is a fact and belongs elsewhere.
- The list is short enough that someone reads all of it.

03 · Slip

What would move the date, by how much, and what you would notice first.

Each risk has an early warning sign
For every named risk, the thing you would observe first, before the slip is certain.
Usually missing: The observable. A risk you cannot detect early is not managed, it is merely anticipated — and the difference is whether you find out with three months to react or three weeks.
Done when
- Every named risk has a sign someone could actually notice.
- The signs are things that happen before the slip, not the slip itself.
- Someone is watching each one.
What actually happened last time
The last two milestones, planned versus actual, with the reason recorded.
Usually missing: Any record. Studios slip in the same way repeatedly — the same discipline, the same phase, the same cause — and nobody notices because nobody writes it down. Two data points are enough to see the pattern.
Done when
- The last two milestones have planned and actual dates recorded.
- A reason is written for each variance.
- Someone has looked for a pattern across them.
Named risks with magnitudes, not adjectives
Two or three risks, each with weeks attached and a rough likelihood.
Usually missing: The numbers. A risk register with no magnitudes is a list of worries and gets read as one. "Console certification, four to eight weeks, landing on the launch window" is a risk someone can plan around.
Done when
- Each risk has a range in weeks.
- The magnitudes are consistent with the assumption costs in section 02.
- Two or three risks are named, not fifteen.
A decided response, before it is needed
For the top risk, what gets cut or moved, decided in advance and by whom.
Usually missing: The decision. Everyone knows what they would probably do; nobody has agreed it, so when the risk lands the first two weeks go on deciding rather than acting — which is the two weeks that would have absorbed the slip.
Done when
- The top risk has a written response.
- A named person can execute it without convening anyone.
- The response has been checked as still viable at the last milestone.

04 · Communication

Whether the status is checkable, and how long a slip takes to reach the people who can act on it.

One plan, and everyone is looking at it
A single artifact the team, leadership, and any investor all read.
Usually missing: The singularity. There is a team plan, a board version, and a publisher version, and they diverge quietly. The first time anyone notices is when two of them are in the same room, which is a bad way to find out.
Done when
- One artifact exists, and everyone can say where it lives.
- Any external version was exported from it rather than rewritten.
- The dates in the deck match the dates in the plan today.
A slip reaches leadership within a week
The gap between the team knowing and leadership knowing, measured in days.
Usually missing: The measurement. Every week a slip stays invisible is a week of dependent work built on a false date, and that rework typically costs two to three times the original delay.
Done when
- The last slip has a known date-known and date-reported, and the gap is under a week.
- The report named when it was first known, not just that it happened.
- Nobody was punished for it.
Reported status is checkable against the build
Every completion claim names where in the build it can be seen.
Usually missing: The tie to the build. "Combat complete" is a status; "combat is what you play in the first ten minutes, minus the third enemy type" is a claim someone can check. The difference is what separates a plan from a wish.
Done when
- Completion claims name a place in the build.
- Someone outside the team has checked two of them at random.
- Anything complete in the plan but not in the build is marked as such.

Three rules about the plan itself

An assumption that cannot be false is a fact. Facts belong elsewhere. A list padded with them hides the two or three things that could actually move the date.

Every risk needs a magnitude. A register with no numbers is a list of worries and gets treated as one. Weeks, as a range, not an adjective.

Review it at every milestone, not when someone gets nervous. Attach it to the milestone review so it happens mechanically. Anything that depends on remembering does not happen.

How to score it

Count the boxes. There are 15.

Score	Read	What to do
0–5	A date, not a plan.	Start with section 02. Writing down what the date depends on is a few days of work and it is the difference between a claim and a position.
6–10	A plan with the uncertainty still implicit.	The structure exists. Put numbers on the assumptions and get someone outside the estimating room to challenge them.
11–14	Better than most studios manage.	Focus on section 04. A good plan that leadership hears about late is worth much less than it should be.
15	Unusually well run.	Check that the assumption list has actually changed since it was written. A list that never changes is being reread rather than reviewed.

01 · The milestone

02 · Assumptions

03 · Slip

04 · Communication

How to score it

Your plan, checked