Capacity planning without the spreadsheet gymnastics

The capacity spreadsheet is a familiar ritual: rows for people, columns for days, a heroic sum at the bottom that says the team can absorb exactly this much work. It's wrong by Wednesday. Someone takes a sick day, a production issue eats an afternoon, a 'small' task turns out to be three, and the beautiful arithmetic quietly stops describing reality. The problem isn't the math — it's that a static snapshot is modeling a moving system.

Why static capacity math fails

A spreadsheet freezes assumptions the instant you save it. It assumes a full roster running at full efficiency, no unplanned interruptions, and scope that holds steady. Real teams violate all three every week. It also can't see the work already in flight — the half-finished items carrying over — so it plans as if everyone starts the cycle empty-handed. The result looks precise and is quietly fictional, which is worse than an honest guess because people trust it.

Plan from live signals

The alternative is to compute capacity from signals that update themselves. Three do most of the work: real availability (who's actually here, accounting for time off and split assignments), current work-in-progress (what's already carrying over), and historical throughput (how much this team actually completes per cycle). These are live facts, not projections, so the capacity number moves as reality moves instead of lying still while the world changes.

Throughput over effort estimates

The single most reliable capacity signal is what a team has actually finished in recent cycles. Throughput — items completed per period — quietly captures everything an estimate tries and fails to model: interruptions, meeting load, the real difficulty of the work, the team's actual size after the org chart met reality. It needs no estimation ceremony, because it's measured from work that already happened. This is the same insight behind forecasting delivery without estimating: count what completes, don't debate what might.

WIP is the real ceiling

There's a hard limit above the arithmetic one, and it's set by flow. Little's Law ties work-in-progress, throughput, and cycle time together: push more work into the system than it can flow and cycle time stretches, so everything takes longer and less actually finishes. That means the honest capacity is often lower than the sum of everyone's available hours — because a team running at nominal 100% is a team whose queue is growing. Protecting flow is exactly why you keep work in flow with real WIP limits rather than treating capacity as a bucket to fill to the brim.

Make it computed, not a meeting

When availability, WIP, and throughput are live in the same system as the work, capacity stops being a quarterly spreadsheet exercise and becomes a signal you can read at a glance. Over-commitment shows up before the cycle starts — the draft is visibly heavier than recent throughput supports — so you cut scope while it's cheap instead of discovering the miss in the retro. Pair this with AI-assisted sprint planning and the agent's draft respects a capacity number that's actually true. Start from the planning pillar to wire the signals together.