Viral Loop

While acquiring users was entirely the company's job, every month started from the top of the funnel. Spend on ads and people arrive; stop and it stops. The growth plan and the budget plan were the same document.

A question Grace asked while reading the monthly numbers exposed another path. Of 3,000 new users last month, 1,100 came from ads, and nobody could say where the other 1,900 came from. Tracing the paths, most had clicked a document link someone else sent them. Not a cent of ad spend sat behind that traffic, and the path was not designed — it was a side effect of people using the product.

The viral loop exists to pull that side effect into the realm of design. Once something accidental gets written down as a repeating structure, you can work out which segment to touch to turn those 1,900 into 2,400.

One turn moves through four positions.

1. A new user arrives, whether from an ad or from the previous turn's invitation.
2. Using the product creates a reason to bring somebody else in — a document that needs showing to someone.
3. An invitation goes out, as a link or an email.
4. The recipient signs up and starts using it, which returns to position one.

Whether it turns is read from how many new users one user brings on average. That figure is the viral coefficient. At 0.4, a hundred users bring forty, those forty bring sixteen, and the thing fades out. At 1.2, the same hundred bring 120, those 120 bring 144, and it grows with nobody touching it. Above one or below one is the fork.

Products that clear one are rare, which does not make 0.6 useless. Bring a hundred people in with ads and those hundred bring sixty, the sixty bring thirty-six, and the total reaches about 250. That is the same as the cost per user falling from $15 to $6.

As important as the coefficient is how long one turn takes. Between a product at 0.6 with a three-day cycle and a product at 0.9 with a thirty-day cycle, the first one is larger six months later, because it completes ten times as many turns. Shortening the cycle means shortening the distance between receiving an invitation and actually using the product, which is usually a question of what the recipient sees first rather than of how the invitation was worded.

The coefficient factors into a multiplication: the share of users who sent any invitation, how many each sender sent, the share accepted, and the share of accepters who go on to invite someone themselves. The first factor is usually the lowest, and where team invite sits moves the coefficient more than anything else.

Turns come in two shapes. One invites people directly; the other relies on a viral surface, where the product's output is visible to outsiders. The first carries intent and the second does not — a product name in the corner of a document opened from a shared link belongs to the second.

In products that only pay off with several people in them, this becomes the main channel of growth: document tools, video calls, scheduling, anything that needs a counterpart to be complete. At Pagedeck, fixing one invite screen carries the same weight as doubling the ad budget.

In the growth model document, where the growth structure is written as a formula, this loop appears as one term. Estimating next quarter's new users means multiplying paid arrivals by the coefficient and adding. A coefficient sliding from 0.6 to 0.5 shows up immediately as the same budget producing 200 new users instead of 250.

It also comes up when deciding where budget goes. There is a point where adding $7,500 of ad spend and raising invitation accept rate by ten percentage points are the same size, and the second one keeps working next quarter. That asymmetry is why loop work tends to outlast campaign work even when both show the same number this month.

The limit surfaces in the same place. When there is nobody left to invite, the coefficient falls. A team that has already brought in everyone at their company has nobody else to add, and getting to the next company needs a different path. A coefficient that has declined for six straight months is usually describing this state rather than a broken invite flow.

"Coefficient is 0.62. A hundred users from ads actually turns into about 260."

Reading the loop as a multiplier on paid acquisition.

"Coefficient held, but one turn takes three weeks. Let's start with the accept screen."

Cycle length sets the growth rate.

"That's not a loop, that's one spike. Next month won't look like this."

Separating a repeating structure from a one-off event.

Ravi brings the coefficient to every quarterly review split into four factors. Last quarter, 4 percent of users sent any invitation, senders averaged 2.6 invitations, and 62 percent were accepted, putting the coefficient at 0.064. The first factor was badly low.

After Noa and Ravi moved the invite placement out of settings and into the document share button, the first factor rose to 13 percent. The accept rate fell from 62 percent to 34, though, so the coefficient stalled at 0.115. What Mia pointed out was that the two factors are coupled: more invitations sent means links reaching people with no account, which made the recipient's first screen the next piece of work.

The most common conflation is treating a viral loop as the same thing as going viral, since the word also describes a widely shared ad. Something that spread once is not a loop. Twenty thousand arrivals in a day from one video is a large amount of traffic, but if those 20,000 bring nobody, next month is zero. A loop is a structure where this turn's output is next turn's input, so what separates it from an event is whether it repeats.

The other one is that the coefficient tells the whole story, because a single figure looks like a verdict. If the people it brings in don't stay, the wheel stops. In a product with a 0.9 coefficient where invited users retain at 12 percent after two months, most of them leave before reaching the second position, so the wheel that actually turns is smaller than the coefficient claims. Teams that put the figure on the goal sheet tend to push invitation volume, while the larger gain usually sits in making arrivals stay.