Viral Coefficient

While word of mouth was described only as happening or not happening, the description never turned into a decision. A report that word of mouth is picking up draws the question of whether advertising can be cut, and the conversation stops there. Between the feeling that something is spreading and the judgment that users will grow without ad spend, one number was missing.

Adding an invite feature didn't change much. A month later signups are up by 1,200, and there's no way to separate how much came from ads and how much from invites. Advertising kept running while the invite feature went on top, so the two are mixed, and counting arrivals through invite links gives a number with nothing to judge it against.

The viral coefficient turns that into one threshold. Convert everything into a single value — new users per existing user — and the line at 1 separates a product that grows by itself from one that doesn't.

Multiply the number of invitations one user sends by the share of those invitations that get accepted. A hundred users each sending four invitations, with 10 percent of them signing up, gives a viral coefficient of 0.4.

Two multiplications are the whole formula, so the difficulty sits in four definitions.

1. The group in the denominator — dividing by all users mixes in accounts that signed up years ago and sent their invitations long since, which drags the value down. Bundle people who joined in the same period and measure inside that bundle. If 500 September signups sent 1,800 invitations that month and 190 converted, the value is 0.38.
2. What counts as an invitation — only links sent by pressing an invite button, or also a shared document that put your product's name on someone else's screen. Including viral surfaces, the points where the product's own output becomes visible to others, can triple the number for the same product. Either choice is defensible; changing it later is not.
3. The window — how many days after signup an invitation still counts. Thirty days produces a stable value, seven produces a low one. Products where people get pulled in as work begins see most invitations in the first week, so seven days works there.
4. What counts as acceptance — clicking the link, creating an account, or reaching activation, the point where someone first feels the product's worth. The further along you set it, the smaller and more trustworthy the number.

Once the value is set, the generations compound. At 0.4, the first 100 users bring 40, those 40 bring 16, those 16 bring 6. Altogether 100 users become 167. Above 1, each generation is larger than the last and the growth doesn't stop.

One number says nothing about speed, so the cycle time comes with it. The same 0.4 produces very different user counts three months out depending on whether invitations happen within three days or a month later.

One is the only absolute line. Above 1 users grow without marketing; below it, spread only enlarges acquisition somewhat. Products that actually clear 1 are rare, so moving 0.4 to 0.6 counts as a real result.

Below 1 the number is still useful, read as a multiplier on whatever acquisition you paid for. At 0.4 it's 1.67x; at 0.6, 2.5x. Whether 100 people bought with ad spend end up as 167 or 250 is decided here, and customer acquisition cost (CAC), the cost of acquiring one customer, falls by the same ratio.

Comparing the absolute value against other products has little use. Collaboration tools that require several people to work are structurally high, and a personal budgeting app is low no matter how well it's made. The comparison that works is last quarter's value and the before-and-after of a change.

Raising it splits into two paths that cost differently. Increasing invitations means moving the invite button somewhere more visible or adding surfaces; raising acceptance means fixing the first screen an invited person sees. The second is usually cheaper — going from 4 invitations to 6 touches more screens than going from 10 percent acceptance to 15, and acceptance rests almost entirely on one screen that usually sits untouched.

"Viral coefficient is 0.4. Cut the ads and growth stops."

A verdict that the product is not in a self-sustaining state.

"Invitations went from 4 to 6, but acceptance dropped from 10 percent to 6."

Only one of the two factors moved, so the value is unchanged.

"Are we counting document shares as invitations? Then we can't compare to last quarter."

Changing the counting basis breaks continuity with past values.

At Pagedeck, Ravi pulls this per signup cohort each quarter. Invitations count only links sent deliberately, and acceptance is set at account creation. Last quarter came in at 0.38; after the first screen an invited person sees was changed to show the inviter's name alongside the title of the document they'd be editing, acceptance went from 9 percent to 13. Mia moved further work on that screen into the next quarter.

The most common mistake is reading a higher coefficient as growth. The number connects directly to growth rate, so the arithmetic feels obvious. But this value only comes out of users who are already satisfied. Attach rewards and push invitations while the product still isn't worth much, and invitation counts climb while the people who arrive leave in their first week. The coefficient sits at 0.5, the number of users still around a month later hasn't moved, and the money paid out in rewards doesn't come back. The order runs the other way: produce users who stay, then help them invite.

Treating a measured value as permanent is the other one. A coefficient looks like a property of the product. But early users have a lot of people nearby to call in, and once that layer is used up the same product with the same invite screen returns a lower number. A coefficient of 0.6 becoming 0.2 six months later is ordinary, and fixing the screen at that point tends not to help. So it gets remeasured every quarter, and when it falls the first thing to examine is which new layer of users you could reach rather than the screen.