Growth

For a long time, growing a product was cut up and handed out by department. Marketing owned getting people to show up. The product team owned shipping features. Sales owned closing contracts. Each group was judged on its own stretch of the road, and the people who vanished between one stretch and the next appeared on nobody's scorecard.

So the meetings kept landing in the same place. A company doubles its ad budget, collects 10,000 signups in a month, and finds 400 of them still around thirty days later. Asked what happened, marketing points out that it delivered the signup number it promised. The product team points out that it shipped every feature that was requested. Both statements are true, which is what makes the situation so hard to argue with. The loss was happening in between — in the minutes after signup, when someone lands on an empty screen, can't tell what they're supposed to do first, and closes the tab. No job description covered those minutes.

Growth exists to stitch those separated stretches back into one line. Treat the path as a loop — someone hears about the product, arrives, tries it, keeps using it, and brings someone else — and it becomes possible to see which section of the loop is leaking. Pick the leaking section, change something small, check whether the number moved, and undo the change if it didn't. The word points at that sequence rather than at a department on the org chart.

The terms under this heading split ten ways: who does the work (roles), what explains growth (models), how the path gets cut into pieces (funnel), how a single attempt is run (execution), how the outcome gets read (measurement), and then five branches that each attach to one piece of the path — bringing people in (acquisition), moving them to the next step (conversion), keeping them from leaving (retention), getting users to bring other users (referral), and handing the selling motion to the product itself (product-led growth).

Roles divide the people doing the work. There are two seats, separated by whether the person decides the direction or builds the thing.

1. A growth lead judges which section is leaking worst and decides where the quarter's attention goes. Something like thirty ideas are always sitting in the queue while only five can actually be run, so most of this job is closing the other twenty-five.
2. A growth engineer builds the chosen attempt into the product and deletes the losing side once results arrive. What separates this from ordinary engineering is the assumption that more than half of what gets built will be thrown away. Given the choice, this seat spends two days putting two versions in front of users rather than two weeks polishing one.

The unit those two sit in is a growth team: product, engineering, design, and analysis mixed into one group that is handed a metric to move instead of a feature list to ship. The mixing matters because a single attempt usually touches all four disciplines. File a request with a neighboring team and wait your turn and one lap takes three weeks, which allows four laps in a quarter and no more. The count matters more than it looks: if three attempts out of four change nothing, a team running four laps a year gets one real improvement out of the whole year.

Models write growth down as arithmetic. A growth model expresses how users and revenue accumulate using addition and multiplication. Active users this month, for instance, are last month's active users multiplied by the share who stay, plus new arrivals. Put real numbers in and the place to push shows up on its own. Start with 10,000 active users, keep 70 percent of them, and add 2,000 new: next month is 9,000, so the product is shrinking. Raising new arrivals to 3,000 only gets back to even, while raising the share who stay to 80 percent reaches 10,000 without a single extra signup. Without the arithmetic, that comparison turns into an argument about opinions.

The funnel cuts the path into countable pieces. AARRR stages divide the route a user travels into acquisition, activation, retention, referral, and revenue, then count how many people clear each one. The cutting is what makes causes findable. Knowing only that 10,000 people signed up and revenue is flat tells nobody what to do. Knowing that 1,200 of those 10,000 created a first document, and 900 of those 1,200 came back a week later, narrows the work to one specific piece.

Execution is about running a single attempt. A growth experiment means writing down in advance what change is expected to move which number by how much, splitting users so that only one group sees the new version, and then checking whether the gap between the two groups is larger than chance would produce. Writing the prediction first matters because once results are visible, a convincing story attaches itself to any number at all.

Measurement produces and interprets those numbers. Growth analytics works out which piece is blocked and whether a change really caused what followed it. It covers everything from recording what users click on a screen to grouping everyone who signed up in the same week and tracking how many are still around in week four. When this branch is weak, the other nine run on instinct.

Acquisition covers the doors people come through. An acquisition channel is a route to the product — search, paid ads, word of mouth, content, partnerships. Each route differs in how many people it delivers, what one of them costs, and what share of them stick around afterward. A thousand people from an ad and a thousand from search are not the same thousand people.

Conversion pushes arrivals to the next piece. Conversion rate optimization (CRO) extracts more results from traffic you already have instead of buying more of it. It's the work of cutting a signup form from five fields to two, which pays off when plenty of people arrive and few of them reach the end.

Retention keeps the people who arrived, and it splits in two. Engagement looks at how deeply someone is involved with the product. A person who opens it once a month to read something and a person who edits documents every day both count as active users, yet their odds of being here next month are nothing alike. Churn prevention catches the signs of a departure early enough to act. A team that logged in daily and has gone quiet for two weeks has effectively left, whether or not anyone has clicked cancel.

Referral turns existing users into a source of new ones. Referral growth designs the routes along which one user produces another: invitations, shared links, rewards for bringing someone in. A product where this branch turns keeps its inflow even as ad spend comes down, though a reward bolted on without a reason to share mostly attracts people who take the reward and leave.

Product-led growth hands the selling motion to the product. PLG tactics replace a salesperson explaining and persuading with a free tier people try on their own, an invite flow that pulls in teammates, and a payment prompt placed exactly where someone hits a limit. Where the other branches each own one piece of the path, this one pulls the whole path inside the product.

The ten branches don't run in parallel. The model picks where to push, the funnel locates that push as a specific piece, acquisition and conversion and retention and referral and PLG hold the actual levers on that piece, execution turns a lever once, and measurement reads what happened. Roles determine how many weeks one full lap takes.

"This quarter we're ignoring acquisition and only looking at activation. Pouring more in the top just leaks out there anyway."

Fixing the second piece of the funnel instead of the first, which also means deferring any work on inflow for a quarter.

"The success metric on this test is signup conversion, but you can raise that and still have the same number of people around a month later."

The number being measured isn't connected to the thing the team is actually trying to grow.

"Four attempts last week, one winner. We rolled the other three back."

Reported on the assumption that failure is the normal outcome rather than an incident.

At Pagedeck, Ravi opens every Monday with last week's numbers spread out in front of him. Signups climbed last month while the share of people still around after two weeks went down. Walking down the pieces, he found that 14 percent of new signups created even one document, and that the ones who did were five times as likely to be there two weeks later as the ones who didn't. So every attempt that quarter aimed at the same place: getting a first document written.

Ravi writes the prediction, Mia checks it against where the product is headed, and Theo puts two versions of the screen up for a week. Three times out of four there's no real difference and the change comes back out; the one winner stays. The biggest win that year was seeding the empty screen with three example documents.

The most common mistake is reading growth as a new label for marketing. The visible wins tend to arrive on the inflow side first, and plenty of people in these seats came from marketing, so the reading is understandable. But most of the hands-on work happens inside the product. The signup flow, the first screen someone sees, the moment a payment prompt appears — changing screens rather than buying ads is what fills the queue. Treat the two as the same thing and the team gets built with an ad budget and no authority to change the product, which means the leaking piece stays exactly as it was while the amount poured in the top goes up until the budget runs out.

The second mistake is believing that running more attempts produces growth. It's true enough that a winner shows up eventually, so the pull toward raising the count is strong. But this is a tool for tuning a structure that already works by a few percent, and in a product nobody has a reason to keep using, a hundred button colors leave the share who stay exactly where it started. Lifting conversion from 2 percent to 2.4 percent is worth something only when the people converted are still there afterward. The order is what gets reversed: teams reach for the tuning tool because it produces visible weekly activity, while the question of why anyone would come back next month has no experiment attached to it. Reverse that order and the team stays busy for a year while the metrics sit where they began.