Compounding Growth

Growth targets used to be written as addition. Three thousand new users this month, three thousand again next month. It's easy to set and easy to score, which is why it lasted. The trouble is that a target like that puts everything already built to no use. Whether the product has 10,000 users or 100,000, next month still means finding another 3,000.

So the same number described completely different situations without any way to tell them apart. Two companies each add 300 users this month; one started from 1,000 and the other from 10,000. The first grew 30 percent, the second grew 3 percent. On a report written in addition, both lines read 300.

Pagedeck noticed this late. The monthly signup count Ravi reported kept going up. In the sixth month Grace asked what the number would be two years out at this pace. Working it through showed a figure that only held as long as the ad spend held — stop spending and it stops there.

Compounding growth is the term for that conversation. It's a way of writing into the plan how much the accumulated base contributes to next month's growth, and the name comes straight from finance, where interest joins the principal and enlarges the next month's interest.

Multiply what you have now by a steady rate to get the size of the next period, and let that result become the base for the period after. Three steps.

1. Set a base and a rate. Say 10,000 active users today and a monthly growth rate of 5 percent.
2. The first month adds 500 and lands at 10,500. A target of "add 500 a month" gets the same result, so at this point nothing separates them.
3. The second month splits. Five percent now applies to 10,500 rather than 10,000, so 525 join, and the third month adds 551.

The gap only becomes visible with time. Hold 5 percent a month for twelve months and 10,000 becomes 18,000; hold it for twenty-four and it becomes 32,000. Adding 500 a month gives 16,000 and 22,000 over the same stretches. The first year differs by 2,000 users, and by the end of the second the distance is 10,000.

Because the curve has that shape, the first six months look flat whichever path a team is on. A chart that moved from 10,000 to 10,500 looks like a chart where nothing happened.

Holding the rate takes two things at once: a steady flow of new arrivals, and people who don't leave after arriving. The second is retention, how reliably existing users keep coming back. If 5 percent arrive each month while 4 percent leave, the real growth rate is 1 percent, and twenty-four months out the number is 12,000 rather than 32,000. Water pouring in and draining out at the same speed, with the level never rising, is what the leaky bucket describes.

Compounding growth comes up when deciding what unit a target gets written in, and when choosing what to work on this month.

1. The unit of the target. Instead of 3,000 new users, write 7 percent monthly growth. A rate scales the absolute number up along with the product, so the fact that a tactic which worked six months ago is no longer enough shows up in the target itself.
2. Sorting the work. Separate work that pays off only this month from work that attaches as a rate every month. A big event or an ad that happens to land belongs to the first group; building a growth loop, where output returns as input, belongs to the second. Loops come in a few kinds: a viral loop, where existing users bring in new ones; a content loop, where what users write draws search traffic; a paid loop, where revenue goes back into advertising.
3. Explaining the business to investors. The number asked for in a funding conversation is usually the monthly growth rate, not last month's signups. With evidence that the rate holds, a small product can still be discussed in terms of where it lands in two years.
4. Subscription revenue. In a business where existing customers stay while new ones stack on top, revenue accumulates in the same shape, and here too the share that stays sets the curve.

"This quarter I want the target as 7 percent monthly growth, not a headcount."

Write the goal as a rate so the bar rises along with the size of the product.

"The curve looks flat right now, but at this rate we double in eight months."

The early stretch is supposed to look like that, so don't change direction over it.

"That one only lifts this month. Nothing compounds off it."

The effect ends in one go and adds nothing to next month's base.

At the quarterly planning meeting, Ravi proposed raising the ad budget to double this month's signups. Grace asked what happens the month after, and the answer was that pulling the budget back pulls the signups back with it.

Noa put a different option on the table in the same meeting: the invite screen takes three clicks to bring in a colleague, and it could take one. The effect this month would be far smaller than the ad spend. Mia worked out both. The ads bring in 2,000 extra users this month and stop there, while the invite change moves the monthly growth rate from 4 percent to 5 percent. Starting from 10,000, the two are nearly identical after twelve months, and after twenty-four they stand at 25,600 and 32,200. The team decided to do both but to hold the extra budget to a single month.

Compounding growth often gets read as a synonym for fast growth, because the steep part of the curve is the part that sticks in memory. In practice this is the slowest approach at the start. Five hundred users in the first month looks poor next to a month that brought in 3,000 by addition. Teams that can't sit with that difference change direction at the six-month mark, which means stopping right before the curves separate — every time.

The other common belief is that once it starts, it runs itself. That comes from the finance picture, where interest accrues without anyone doing anything. Holding a rate gets harder as the product grows. Five percent of 10,000 is 500, but 5 percent of 100,000 is 5,000, so the same rate means bringing in ten times as many people each month. When invites and search don't scale with the size of the base, the rate drifts down on its own. A falling growth rate is closer to the default than to a sign that something broke.