K-Factor
The K-factor measures how many additional users each user brings in, borrowing the term from epidemiology to describe how a product spreads. It is effectively the viral coefficient applied to a defined time window.
Key takeaways
- K-factor is a viral coefficient stated per cycle; the window is part of the number.
- Below one, total users from one entrant approach one divided by one minus k.
- Lengthening the measurement window raises reported k without improving anything at all.
- Saturation makes the geometric projection overshoot, so forecasts from one cycle overstate growth.
- K counts arrivals, not fit; pair it with a qualification measure.
In depth
The K-factor is a viral coefficient with a clock attached. You state it per cycle: the number of new users each existing user produces within one defined window, such as a week. That bound is what makes the number compoundable. Over successive cycles, an entering cohort produces k times itself, then k squared, and so on, so total users generated from one entrant approaches one divided by one minus k whenever k stays below one. That geometric sum is the whole reason the metric is worth stating precisely.
Two adjustments move the reported figure without changing the product at all. Shortening the measurement window lowers k, because fewer hops finish inside it, and lengthening the window raises k while making the number slower to react. Real improvements come from the same places as any spread: how many people each user reaches, and how many of those act. The epidemiological origin also imports a warning. Above one, a contagion model assumes an unbounded population, and no product has one.
Teams state k as a pair with the cycle length, write both on the dashboard, and never compare a weekly k against a monthly one. A quiz funnel makes the window easy to choose, because completion is fast and the natural cycle is the gap between one person finishing an assessment and a colleague they sent it to finishing theirs. Measuring that gap first, then computing k over it, produces a figure two teams can actually reconcile.
The geometric sum assumes k holds steady, which it does not. Each cycle draws from a smaller pool of unreached contacts, so the real series terminates well before the formula's limit. Projecting a quarter of growth from one cycle's k therefore overshoots, sometimes badly. The number is also silent about who arrives: k counts heads, not fit, so a funnel can post a healthy k while the invited people never qualify. Pair it with a quality measure or it flatters.
Example in practice
How to measure it
Fix the cycle length first, using the median time from one user's share to the invited person's own share. Then, for each cycle, divide new users attributable to invitations by the number of users present at the start of that cycle. Report the series, not a single value, because the shape across cycles tells you whether the engine is holding or fading.
Compare the k series against paid acquisition in the same cycles. The useful derived figure is the amplification factor: total users a cohort eventually produces divided by the users who entered it. If that factor sits near 1.4, every hundred paid signups behave like a hundred and forty, which converts k directly into an effective cost per acquisition your finance team can use.
Common mistakes
Teams quote k without the window and then compare figures that measure different things. A weekly 0.3 and a quarterly 0.3 describe products an order of magnitude apart, yet both appear as 0.3 in a deck. Always write the cycle beside the number, and when you inherit a figure from someone else, ask what period it covers before you plan against it.
The second error is forecasting with a constant k. Plugging one cycle's value into a compounding model produces a curve that never arrives, because each generation faces a more saturated audience than the last. Forecast with a k that decays across cycles, fit the decay from your own generation data, and treat any projection beyond three or four cycles as a scenario rather than a plan.
Frequently asked questions
Is the K-factor the same as the viral coefficient?
They are nearly identical and often used interchangeably. The key distinction is that the K-factor is usually measured against a specific viral cycle time, which makes growth modeling more accurate.
Why does viral cycle time matter for the K-factor?
A K-factor only tells you how many users each user brings; cycle time tells you how fast. The same K-factor with a shorter cycle compounds into far more users over a given period.
How do I track K-factor in a quiz funnel?
Tag every share link from your result pages with unique parameters and measure how many invited recipients complete the quiz. That acceptance rate, multiplied by shares per finisher, gives you a measurable K-factor.
What K-factor do I need for self-sustaining growth?
Above one within a cycle short enough to matter, which very few products reach and none sustain indefinitely. Below one the product still grows, just not on its own: each entrant brings a fraction of another, and acquisition remains necessary. Treat one as a theoretical boundary rather than a target, because approaching it usually means the invitation mechanic has taken over the product.
How do I choose the cycle length for K-factor?
Use the median time between a share and the recipient's own share, measured from your own data rather than picked for convenience. If that median is five days, a weekly cycle captures most hops without stretching. Picking a month because reporting is monthly inflates k and slows your feedback loop, which is the opposite of what the metric is for.
Why is the K-factor borrowed from epidemiology?
Because the arithmetic of spread is the same: one carrier, some number of contacts, some probability each contact is affected. Growth teams adopted the vocabulary along with the model. The borrowing is imperfect, though, since disease models assume a large susceptible population and products face a bounded, quickly saturating one, which is why product k values fall over time in a way epidemic curves do not.
How do I turn K-factor into a CAC number?
Compute the amplification factor, one divided by one minus k, then divide your paid cost per acquisition by it. A k of 0.25 gives roughly 1.33, so a paid signup that cost a hundred effectively costs about seventy-five once the invitations it produces are counted. Use a decayed k rather than the launch value or the discount will be overstated.
Can K-factor be negative or greater than several?
It cannot be negative, since it counts users produced and the minimum is zero. Values well above one do appear briefly, usually during a launch when a large pre-built audience is being converted at once. Those readings describe a backlog draining, not a repeatable engine, and the next cycle almost always returns to a fraction of the first.
Should sales-led B2B companies track K-factor?
It is worth tracking where the product or content spreads inside accounts, and not worth much where every deal starts with an outbound conversation. In the first case the cycle is short and the network is a company, so k is measurable and actionable. In the second, invitations are rare enough that the number is mostly noise and account-level referral counts serve better.