Ten projects in flight is a queue pretending to be progress

A team of six is running ten initiatives. Ask how it is going and every one of the ten is "in progress", which is technically true and materially false: the team is doing about two things, slowly, while eight wait in slices too thin for anyone to notice the waiting.

Why the count drifts up

Nobody decides to run ten things at once. The count drifts, one rational yes at a time, because starting is the cheapest concession there is: "we'll get going on it" ends an uncomfortable conversation without displacing anything visible. Starting feels like responsiveness. Every stakeholder can honestly be told their thing is underway — and the cost lands somewhere no status report looks: the finish dates of everything already running, each pushed a little, none pushed enough to trigger an alarm.

What the in-flight count actually costs

Attention, not headcount, is the scarce unit. Every parallel initiative taxes it three ways: the switching itself, the re-loading of context that decayed while the work waited, and the coordination surface — ten things touching each other in more places than four things ever could. The arithmetic is unkind: work that would take a focused month takes a fragmented quarter, so the ten-in-flight team delivers later and less than the four-in-flight team with the same people. Parallel work feels like motion and delivers like a queue — because it is one, just round-robin instead of first-in-first-out, with the switching tax added.

And the queue is invisible. A sequenced backlog shows its waiting honestly: item seven has not started, everyone can see it. Ten "in progress" bars show no waiting at all — the same delay, relabelled as progress.

Why WIP-limit slogans don't hold

The standard remedy — "limit work in progress" — is correct and does not survive contact with the first escalation, because a limit with no mechanism is a preference. The vice-president's request arrives, the limit is "temporarily" lifted, and the count is eleven. Limits hold when exceeding them costs something visible at the moment of decision, which is a property slogans do not have and constraints do.

The mechanism: pay for parallelism at commitment time

Make the trade visible where it is decided. Each team's honest capacity — FTE-months, standing costs subtracted — is a line, and demands walk down it in priority order until it is spent. Under that arithmetic, "we'll run it in parallel" stops being free: the eleventh initiative does not fit, so it must either name which commitment moves, take a recorded not-now, or cross the line as a recorded exception with a name on the risk. The count stays low not because anyone polices it, but because raising it now has a price tag someone must sign.

Sequencing follows naturally: when starting is no longer free, the conversation becomes "which of these finishes first?" — and the mid-quarter arrival becomes an edit to a named commitment instead of an eleventh plate to spin.

The uncomfortable close

Finishing is a throughput strategy. Starting is just inventory — and inventory of the worst kind, because it looks like progress in every report and decays like milk. The teams that ship the most are rarely the ones doing the most at once; they are the ones whose capacity line makes "at once" a decision somebody has to own, out loud, with the displacement attached. DeliverySheet holds that decision's record — per-team capacity figures on every commitment, summed in the committed-capacity report, with a named exception when a promise runs ahead of its evidence — at commitment time, which is the only place a WIP limit has ever actually lived.