Growth Team

Back when growth work was spread across teams organized by discipline, one small idea had to wait in four separate queues. Someone suggests reworking the signup screen: the product team writes it up, the engineering backlog takes the build, the design queue takes the screens, and the analytics intake form takes the measurement. A full lap ran three weeks, and at that pace a team gets four laps a quarter.

That is exactly what happened at Pagedeck when Ravi proposed cutting the signup form from five fields to two. The request landed twelfth in the engineering backlog and shipped three weeks later, and the screen had no event logging on it, so nobody could tell where people were dropping off. Adding the logging cost another two weeks.

The bigger problem was ownership. The share of new signups who created a first document slid for three straight months and nobody noticed. The product team was judged on features shipped and marketing on people brought in, so the step between the two showed up on nobody's scorecard.

A growth team removes both problems at once. Four disciplines sit together, and the team is handed a single number rather than a feature list. There is nowhere to hand work off to, so there is nothing to wait for, and when the number drops it is that team's problem.

The day opens by checking the numbers on whatever experiments are running. A growth experiment means writing down in advance which change should move which number and by how much, splitting users into two branches, showing the changed screen to only one of them, and then checking whether the gap that opens up is real or noise.

A lap runs in four beats.

1. Read. The Pagedeck growth team opens last week's three result boards on Tuesday morning. The branch with the shortened signup form moved conversion from 18 percent to 24 percent. The branch that only changed the invite button color came in at 11.2 percent against 11.4 percent, which is hard to call a difference at all.
2. Delete. The winner gets turned on for everyone and the loser is pulled out of the code. The button color test was gone by that afternoon. The team works on the assumption that more than half of what it builds gets deleted, so it builds a two-day test version instead of a two-week finished one.
3. Pick. Growth analytics shows which step loses the most people, and the biggest leak gets picked first. If signups are healthy but few people reach a first document, that step is the target rather than more advertising.
4. Build. Whatever was picked gets written and put live as two branches. That part usually takes two days.

Those four beats run three or four times a week. The missing roadmap is the sharpest difference from other teams. What gets fixed at the start of a quarter is one goal — "take first-document rate from 18 percent to 26 percent" — and what to build to get there is left blank. That's also why the work clusters around conversion optimization, which means pulling more results out of the same incoming traffic.

Four seats sit together inside the team, and outside it the team meets the product organization and marketing every week. The Pagedeck growth team is led by Ravi, with Theo and Noa each putting in half their time.

Inside, the seats break down like this.

1. Growth lead — decides which step gets attention this quarter. About thirty ideas pile up and five can actually run, so killing the other twenty-five is the bulk of the job.
2. Growth engineer — builds whatever was picked and deletes the losing branch. Unlike regular product engineering, the code is written expecting to be thrown away.
3. Designer — makes the two branches. Noa keeps experiment screens inside the product's rules while holding them separately so they're easy to discard.
4. Analyst — checks that events are logging correctly and reads the results. Broken logging voids an experiment outright, so when this seat is empty the other three end up guessing.

Outside, the team collides with the product team most often. When both teams change the same screen, nobody can say which change moved the number. Pagedeck drew a line: the signup screen and the first-document screen belong to growth, and everything inside the editor belongs to product.

With marketing the shared subject is acquisition channels. A thousand people from ads and a thousand from search behave differently once inside the product, so both sides look together at which route actually reaches a first document.

"That one should sit with the growth team. It's a signup funnel number."

An argument to split ownership by metric rather than by feature area.

"We can't staff a whole growth team, so let's start with one person on it."

Handing over the role without creating the org unit.

"If growth touches the editor, it'll collide with the product team's experiments."

A request to redraw which screens belong to which team.

Pagedeck built its growth team in a quarter where signups grew every month while paid conversion sat flat. Grace wanted to spend more on advertising and Ravi pushed back. Only 18 percent of new accounts created a first document, so doubling the ad budget would raise paid conversion by the same flat ratio and nothing more.

Of the eleven experiments in the first month, eight produced no difference at all. The remaining three moved 18 percent to 24 percent, which raised the paid customers coming out of the same ad budget by a third.

The most common reading treats it as another name for the marketing team. Both work on growing user counts and the budget comes from a similar place, so they're easy to merge. What they touch is different. Marketing brings people in from outside the product; a growth team changes how people who are already inside move through it. Treat them as one thing and every bad number produces the same answer, which is a bigger ad budget. With the inside of the product untouched, six months of that leaves you spending twice as much for 1.2 times the paid customers.

Reading the team as the thing that starts growth is nearly as common, since drawing one more box on the org chart looks easy. But this team plugs leaks in a stream that is already flowing, and if no water arrives there is nothing to plug. A company that hasn't found product-market fit (PMF) — the state where what you built lines up with real demand in a specific market — will run thirty experiments over a quarter and finish with conversion up one percentage point, because the thing that needed fixing was the product.