Virality Coefficient Calculator – K, and the Series It Drives
The viral coefficient, usually written K, is one multiplication: K = i × c, where i is the invitations an existing user sends in a cycle and c is the share of those invitations that convert. Four invitations at a 12% conversion rate gives K = 4 × 0.12 = 0.48. The number is not interesting on its own. What makes it worth computing is the series it sets in motion, and the ceiling that series runs into.
The series, not the number
Cycle 0 is the starting cohort. Each cycle multiplies the previous cycle’s arrivals by K, so new users at cycle n are initial × Kⁿ and the running total is the geometric series initial × (1 + K + K² + … + Kⁿ). A thousand users at K = 0.48 add 480 in the first cycle, 230 in the second, 111 in the third — each round smaller than the last, because every arrival is itself multiplied by a number below one.
Why K below 1 converges
A geometric series with a ratio below one has a finite sum. Taken to infinity it collapses to initial × 1/(1 − K), the saturation ceiling this tool draws as a dashed line across the chart. A thousand users at K = 0.48 converge on 1,000 ÷ 0.52 ≈ 1,923. The cohort is not dying — it grows by 92% in total — it simply stops somewhere, and the multiple 1/(1 − K) says where. That multiple, about 1.92× here, is the portable part: it holds at that K whatever size the cohort started at.
Why K = 1 is a different formula
The closed form for the running total is initial × (1 − K^(n+1)) / (1 − K), and at K = 1 that denominator is zero. This is not a rounding nuisance to smooth over with a small number added underneath; it is a separate case with its own formula, total = initial × (n + 1). Every cycle adds the starting cohort again, so growth is linear, not exponential: it never stops and it never accelerates. Five invitations at a 20% conversion land on it exactly, which is common enough that the tool branches on it explicitly and says so on screen.
Why K above 1 is not a forecast
Above one, each cycle adds more than the last and the totals compound — in the arithmetic, forever, because the arithmetic contains no audience. There is no term for the number of people who could ever be reached, none for novelty wearing off, and none for the easiest contacts being used up first, all of which push i and c down over time. The later figures in a K > 1 projection are a property of the equation, not of a product.
K for the same reason: published figures disagree by an order of magnitude and go stale within a year. Compare a cohort against your own earlier cohorts instead.Cycles, fractions and elapsed time
A cycle is one round of existing users inviting and invitees converting — for most products days to weeks, not minutes. Both closed forms accept any non-negative real exponent, so 4.5 cycles is as valid as five. That matters, because the question is usually asked in days: with a four-day cycle, eighteen days elapsed is 18 ÷ 4 = 4.5 cycles. Drive the projection from elapsed time and the fractional count is used as-is rather than rounded, with the part-cycle shown as its own final row. Elapsed time is reported only as a duration; this calculator never prints a calendar date, because a cycle estimate cannot support one.
Where the arithmetic stops
- K = 0 — no invitations, or none converting. The total stays at the starting cohort forever, and that figure is also the ceiling.
- Large K over many cycles —
K^(n+1)leaves the range a double can represent quickly, and long before the overflow a double stops representing consecutive whole numbers. Anything past that reads as out of range rather thanInfinity, and the rest of the table still renders. - A conversion rate above 100% — refused, because a share of something cannot exceed the whole of it.
What the model leaves out
Three assumptions hold the projection up: a constant K in every cycle, a constant cycle time, and an audience that never runs out. None survives contact with a real product for long. Treat the output as a sensitivity model — what these two numbers imply if they hold — which is a different question from what will happen.