Case study · 2021-2024

From twenty engineers to one hundred, without losing the plot.

Three years inside a US healthcare marketplace as it became a $2B platform. The work that grew here became the spine of three of our frameworks.

20→100Engineers, three years
$2BValuation reached
~40%Sev-1 incidents, year on year
~3×Deploy frequency, sustained
Context

A marketplace under load.

US healthcare facilities, professionals, regulatory weight on every side. Demand was outrunning the team's ability to ship safely. The architecture had grown by accretion, not by design. The window to fix both without stalling the business was narrow.

The brief was direct: scale engineering to match the business without breaking what made the business work. Three years later the company crossed $2B in valuation, and the engineering organisation that crossed with it was still recognisable to the people who'd been there from day one.

Approach

Three things, in order.

The work split into three movements. None of them original. The value was in choosing them, naming them, and refusing the fourth.

N° 01

Service boundaries that match team boundaries

We rewrote the architecture map and the org chart on the same week. Where they disagreed, one of them moved. By month three they fit.

N° 02

Decisions written down, defaults distributed

An ADR habit, a written escalation path, a decision log every Friday. The CTO stopped being the bottleneck for the small things, and became more useful for the large.

N° 03

Validation built into delivery

Every quarter shipped at least one Validation Sprint as a first-class citizen. The roadmap began to look like a series of falsifiable bets, not a wish list.

"The team that grows during a hard scaling year is the team that survives the next one."

Jędrzej Tabaczyński
Outcome

What stayed.

The numbers above are useful. The structures below are what the team carries now, after we left.

Stayed N° 01

A decision log everyone reads

A single place where every architectural and organisational decision is written, dated and revisited. Replaced four meetings a week.

Stayed N° 02

An on-call that does not burn people

Rotations, runbooks, incident-review-as-learning rather than blame. Sev-1 incidents fell forty percent year on year.

Stayed N° 03

A Validation Sprint cadence

One per quarter, baked into the planning ritual. The roadmap stopped being a list and became a portfolio of bets.

Stayed N° 04

A team shape that absorbs growth

Squads of seven to nine, owning a service end-to-end. Doubled headcount three times without a re-org costing more than two weeks.

Start a conversation

If this is the year ahead of you, let's talk.

Not every engagement looks like ShiftKey. Most are smaller. None are louder.