k, and the clock it runs on

“Viral” is one of those words that does damage by sounding like magic. It suggests a thing that either happens to you or doesn’t, a lightning strike you can pray for but not engineer. Underneath the word, though, there’s a small, unglamorous equation, and once you can see it, most of the mystique drains out and something more useful takes its place: a couple of levers you can actually pull, and a clear-eyed read on which loops are real and which are stories people tell after the fact.

The equation everyone half-quotes

Start with the number people mean when they say “viral coefficient” or k-factor. It’s just this: take an average new user, and ask how many successful new users they bring in before they’re done. That’s two things multiplied — how many invitations the average user sends, times the fraction of those invitations that convert into active users.

k = (invites sent per user) × (conversion per invite).

If k is above 1, every cohort of users produces a larger next cohort, which produces a larger one again — the definition of exponential growth with no outside spend. If k is below 1, each cohort produces a smaller echo, and the echoes fade out. That threshold at 1 is why the number gets worshipped.

But two facts about k get quietly dropped, and they’re the whole story.

The first: sustained k above 1 is close to mythical. The reason is baked into the same network you’re spreading through. Your earliest, most enthusiastic users invite the most and convert their friends the best, because the best-fit people go first. As you grow, you’re recruiting less-enthusiastic inviters reaching more-saturated, worse-fit friends. Both terms in the equation sag. So k almost always decays cohort over cohort — the loop that looked explosive in month one is a different, weaker loop by month six. A brief k over 1 during a launch is real and worth catching; a permanent one is the growth equivalent of a perpetual-motion machine.

The second dropped fact is the one that actually matters more day to day: the clock. k tells you the multiplier per generation; it says nothing about how long a generation takes. Cycle time — how long from a user joining to that user bringing in the next one — is the other half of the engine. A loop with a modest k and a fast cycle (invite happens in the first session) will outrun a loop with a bigger k and a slow cycle (invite happens after a month of use), because it gets through far more generations in the same calendar. When people say a product “grew fast,” they’re usually describing cycle time as much as coefficient. You compound on the number of turns, and the clock sets the turns.

The trap, and the more honest goal

Two traps sit on top of this, and they’re where growth stories go wrong.

The first is believing your own virality. Most “we grew virally” narratives are, on inspection, strong word-of-mouth plus good retention plus some paid — not a literal, self-sustaining k above 1. That distinction isn’t pedantic. If you think you have a k-above-1 loop and you actually have great retention driving referrals, you’ll spend your energy optimizing invite mechanics when the thing actually holding the number up is the product being worth coming back to. Measure it honestly and per cohort — blended lifetime averages will flatter you, because they mix your best early inviters in with everyone since. Watch k’s trend, not its snapshot: a k that’s drifting down cohort over cohort is telling you the loop is aging, no matter how good this quarter’s blended figure looks.

The second trap is treating k below 1 as a loss. It isn’t. A sub-1 loop is an amplifier on every other channel you run. The math is a geometric series: if k is 0.6, every 100 users you bring in the door — from paid, from content, from anywhere — turn into roughly 100 ÷ (1 − 0.6) = 250 over the life of the loop. That’s a 2.5× multiplier on all your acquisition, quietly, forever, as long as the loop and the retention behind it hold. A durable k of 0.5 on a fast clock is a far better asset than a fragile k of 1.1 that collapses the moment you stop hand-feeding it with referral bonuses — and referral-bonus growth is exactly the kind that attracts users who came for the bonus, don’t stick, and don’t invite, so next cohort’s k caves in. (Two earlier pieces sit underneath this one: why most referral programs fail, and why the best referral makes the sender look good rather than richer. This piece is the arithmetic those live inside; and it’s distinct from the paid-vs-owned-loops piece, which is about cost compounding — this one is about the shape of the loop itself.)

So the goal isn’t to chase the mythical number above 1. It’s more useful and more attainable than that: build a durable loop with the best k you can honestly sustain, run it on the fastest clock you can design, and protect the retention underneath it — because a user who churns before they complete the cycle contributes exactly zero to either term. You can’t out-loop a leaky bucket. Get those three right and even a “sub-viral” coefficient becomes one of the most valuable things you own: growth you don’t rent.

The research, to look up: the viral coefficient (k-factor) and cycle time as the two drivers of loop growth (widely covered in growth writing — Andrew Chen and Reforge among the clearer treatments); the geometric-series amplification 1 ÷ (1 − k) for k < 1; and the coupling to retention (the leaky-bucket / steady-state argument). These are modeling heuristics, not physical laws — real loops are noisier than the clean equation, and k varies by cohort, segment, and season.