Hiring Slower, Shipping Faster: Engineering Capacity Myths

It is planning season, which means engineering leaders are being asked how many people they need next year. The question assumes a relationship between headcount and output that frequently does not hold.

Teams that double in size rarely double in throughput. Many deliver less for a period, and some never fully recover the difference.

Coordination cost grows faster than capacity

Each new team adds communication paths, and the number of paths grows faster than the number of teams. That overhead is invisible in a headcount plan and entirely visible in delivery.

This is why organisations sometimes ship more after a reorganisation that reduced team count, which makes no sense in a capacity model and perfect sense in a dependency model.

Measure what teams wait for

The most useful diagnostic we know is to trace a representative sample of completed work end to end and record every handover and every wait state.

Most organisations discover that active work occupies a small fraction of elapsed time. The rest is queuing for another team’s review, approval or capacity. Adding engineers to a team that spends most of its time waiting adds waiting.

Platform teams become involuntary bottlenecks

A common pattern: a shared platform team is created to reduce duplication, and within a year every other team’s work passes through it.

The fix is not more platform engineers. It is converting the platform from a service that performs work on request into infrastructure other teams use without asking. That is a different design goal and it needs to be stated explicitly.

Onboarding is a capacity decision

A new engineer consumes senior engineering time for months before contributing net capacity. Hire ten at once into a team of fifteen and you have removed capacity for a quarter.

Organisations that stagger hiring and invest in a structured first fortnight recover the investment much faster. The measure worth tracking is time to first meaningful contribution, and most organisations have never calculated it.

Before you approve the headcount

Ask one question of each team requesting people: can you deliver something a customer would notice without another team’s involvement?

Where the answer is no, headcount will not fix it and may make it worse. Where the answer is yes, additional capacity converts to output reasonably efficiently. That single question will usually reallocate a hiring plan more effectively than the plan itself.