Module 4 — Estimation and Time · Lesson 4.4
Capacity Planning at Workspace Scale
Sprint, quarter, year — and the difference between commitment and intent
~12 min
What you'll learn
- Distinguish the three planning horizons and what each is for
- Run a sprint plan that uses calibrated estimates without inflating commitments
- Commit sprints at p90 rather than the median, and know where those numbers come from
- Use the year horizon to set capacity-shaped strategic bets rather than ambition lists
Planning is the act of comparing what the team wants to do against what the team has time to do, at a chosen horizon. Done well, it produces a credible commitment about the near horizon and a credible direction about the far one. Done poorly, it produces a list of intentions that nobody believes and nobody is held to. This lesson covers the three horizons — sprint, quarter, year — and how Kavanah supports each.
Sprint: the firm horizon
A sprint is the horizon over which the team commits. Whatever you commit at sprint start should ship by sprint end with the same reliability you would expect from any other team commitment — high. Sprint commitments that miss should be the exception, not the rule.
Kavanah's sprint planning surface is the Planning view (/planning), on the Sprint planner tab. Set the sprint length and the realistic per-person working hours (capacity here is the honest number, not the theoretical maximum), pick the confidence level in the 'Commit at' control, and draft. The planner packs the backlog against capacity and shows the forecast at both p50 and p90, so you can see what the choice costs before you make it.
The critical word there is p90. Committing at the median means missing half the sprints. Committing at p90 means hitting almost all of them, with the trade-off that you fit slightly less in each sprint than the median rollup would suggest. The right trade-off is to hit consistently — missed sprints are expensive in trust, and the cost of carrying less work per sprint is far less than the cost of intermittent late delivery.
Those two numbers are not a rule of thumb. Kavanah derives them from your own history: every completed task carrying both an estimate and logged hours contributes a ratio of actual to estimated, and the p50 and p90 of that distribution become the multipliers applied to each person's estimates. A teammate whose work reliably takes 1.5x their estimate gets planned at 1.5x, not at face value. Someone with too little history of their own inherits the team's distribution; until the workspace has any, the planner says so and takes estimates at face value rather than inventing a confidence level it cannot support.
The AI agent has the same tool. Ask it to plan the next sprint and it will show you both numbers — what fits at median and what fits at p90 — the points you give up by choosing safety, and which teammate's uplift came from their own record versus the team's. It recommends; it does not commit. Choose p90 unless you have a specific reason not to.
Quarter: the planning horizon
The quarter is too long to commit at sprint reliability but short enough that capacity is a real constraint. Quarter plans should be commitments about direction and rough capacity allocation, not about specific tasks.
A good quarter plan answers two questions per project: how much capacity will this project get, and what are the milestones that would tell us we are on track. The capacity allocation is expressed as a range — '20–30% of team capacity for Q3, biased toward the back half' — because reality will move within the quarter.
Kavanah's Portfolio view (/portfolio) is where the quarter plan lives. Set each member's weekly capacity and their allocations to projects on the Resources tab, and the quarter panel rolls those up: every project's share of the team's total capacity for the quarter, summed. Where the total exceeds 100%, it says so — and the visible rollup makes that cut a conscious decision rather than a quiet overload.
Two details worth knowing, because they change what the number means. Allocations rarely line up with quarter boundaries, so hours are counted pro rata: an allocation running August to November contributes only its August–September portion to Q3, not all of it. And the panel shows plan-to-actual drift per project once actuals are logged — at quarter end, the gap between what a project was promised and what actually flowed there. Drift consistently over 20% means the quarter plan is decorative.
The per-member view still matters alongside the rollup: a quarter that sums to 90% can still have one person at 130%. The rollup catches over-commitment; the overallocation flags catch the person it lands on.
Year: the direction horizon
The year is too long to plan in terms of capacity at task or project granularity. Year plans should be about bets, not about commitments. Three to five strategic bets, each tied to the workspace Vision, each with a rough share of the team's annual capacity, each with an explicit Negation of what it is not.
The bets live on the workspace KVN charter (/workspace-kvn), under the three axes. Each carries a label, a share of the year's capacity, and the line explaining how it ladders to the Vision. The shares are summed: over 100% is flagged, because you cannot spend 130% of a year. The count is flagged too — under three may be under-ambitious, over five will not fit — but neither blocks a save, since a charter passes through two bets on its way to three.
The annual ritual is to revisit the workspace KVN, confirm or change the bets, and reset the project portfolio accordingly. Projects that no longer ladder up to a current bet get closed out, not allowed to drift.
The trap at the year horizon is ambition without negation. Listing eight strategic priorities is the same as listing zero, because the team's actual capacity can serve at most three or four. The discipline is to choose, in advance, the things the team will not pursue — and to write those into the workspace Negation so the rest of the year's conversations have a sharp instrument for filtering opportunities.
One asymmetry to understand: the agent reads your bets and will tell you when work ladders to none of them, but it will not refuse that work. Only the Negation is enforced. Bets shape the conversation; the Negation is the hard no.
Where the three horizons meet
Each horizon is the input to the next. The year sets the bets. The quarter allocates capacity to each bet. The sprint commits specific tasks within the quarter's allocation. The bottom-up signal — actuals from each sprint — feeds back up the chain, recalibrating the quarter's expected output and, over multiple quarters, the year's bets.
That feedback is not a metaphor at the sprint horizon: logged hours against estimates are what move the p50 and p90 multipliers, so a quarter of consistently optimistic estimating shows up as a planner that packs less into the next sprint. The loop closes on its own, but only if actuals get logged. A team that estimates and never records what the work took has disabled its own calibration.
The healthy operating cadence is: weekly sprint plan + retro, monthly quarter check-in, annual year reset. Each takes a different amount of time and produces a different artifact. Conflating them — running a sprint plan as if it were a quarter conversation, or vice versa — produces meetings that drag and decisions that nobody owns.
One caution about reading the metrics. /reports' Progress tab reads weekly, monthly, quarterly, or yearly, and /portfolio's quarter panel is quarter-scoped — but these are trailing windows, not calendar periods: 'quarterly' there means the last three months, not this quarter to date. There is deliberately no 'this sprint' view on /reports, because sprints belong to projects: a workspace running five projects has five concurrent sprints with different dates, and a single workspace-wide sprint window would be a number nobody could reason about. Read sprint-scoped questions on /planning, where the sprint is unambiguous. Deciding which horizon you are in remains your job, not the tool's.
Run each horizon deliberately
- 1
On the Sprint planner tab, set Commit at to p90 and draft against honest per-person hours. Reject the temptation to add 'just one more.'
- 2
On the Resources tab, confirm each project has an allocation and a milestone. Read the quarter rollup; if it is over 100%, cut until it isn't.
- 3
Three to five bets, each with a capacity share, each laddering to the workspace V. Write the explicit N — what you are choosing not to do.
- 4
Build the operating cadence into the calendar
Weekly sprint slot, monthly quarter check-in, annual year reset. Each is a recurring calendar event with a fixed agenda.
Planning health
- Sprint commitment hit rate
- Fraction of the tasks a sprint committed to at its start that shipped by its end date. Shown on the Sprint planner tab (/planning) for your recent completed sprints. Measured against the commitment as it was recorded on day one, so tasks added mid-sprint neither help nor hurt it.
- Healthy signal: Above 85%. Below 70% means committing at median, not p90.
- Estimate calibration ratio
- Each person's actual hours divided by their estimate, over the last 90 days. Shown on the Estimation calibration tab (/planning); its p50 and p90 are what the sprint planner uses to uplift estimates.
- Healthy signal: Near 1.0 and steady. Persistently above 1.0 is fine and self-correcting — the planner adjusts for it. A drifting trend is the thing to look at.
- Quarter allocation total
- Sum of every project's allocation as a percentage of total team capacity for the quarter, on /portfolio's Resources tab. Allocations straddling the quarter edge count pro rata.
- Healthy signal: 100% or slightly under. Above 100% is implicit overload, and is flagged.
- Plan-to-actual drift
- Per project, the gap between planned allocation and the actual hours that flowed there — shown beside each project's share on the Resources tab, once actuals exist.
- Healthy signal: Under 20%. Larger drift means quarter plans are not load-bearing.
- Year-bet count
- Number of strategic bets on the workspace KVN charter (/workspace-kvn), with their capacity shares summed.
- Healthy signal: 3–5, summing to 100% or less. Both are flagged; neither blocks a save.
Key takeaways
- ·Three horizons: sprint (commit), quarter (allocate), year (bet).
- ·Commit sprints at p90, not median — and log actuals, because that is what makes p90 mean anything.
- ·Quarter allocations sum to 100% or less. Year bets are 3–5, with explicit negations — and only the Negation is enforced; bets inform.
- ·Each horizon feeds the next; conflating them produces decisions nobody owns.
With assignment and estimation in place, the next module turns to the operating cadence — the rituals that hold KVN steady against the noise of the week.
Sources
- 1.From Nobel Prize to Project Management: Getting Risks Right
Bent Flyvbjerg · Project Management Journal 37(3), 5–15 · 2006
The percentile-commitment logic: the median carries a 50% overrun risk; committing a sprint at a high percentile is the same discipline as uplifting a forecast to a confidence level you can live with.
- 2.Kingman's formula
J. F. C. Kingman · Mathematical Proceedings of the Cambridge Philosophical Society 57(4) · 1961
Mean wait scales with ρ/(1−ρ) — planning to full utilization (ρ→1) means unbounded delay; realistic-hours capacity leaves the slack that keeps cycle time finite.
- 3.Little's Law as Viewed on Its 50th Anniversary
John D. C. Little · Operations Research 59(3), 536–549 · 2011
L = λW: committing more work than throughput supports lengthens cycle time proportionally — why allocations over 100% of capacity are implicit overload, not a stretch goal.
- 4.What You Should Know About Megaprojects and Why: An Overview
Bent Flyvbjerg · Project Management Journal 45(2) · 2014
The 'iron law of megaprojects' — nine in ten overrun. Over-committing capacity is the default failure mode, so the discipline is to choose in advance what not to do.