The average number of new users an existing user brings in through sharing — a coefficient above 1 means growth that feeds itself.
In practice
Standard formula, often written K: the average number of invitations sent per user, multiplied by their conversion rate. K > 1 means each user brings in more than one other on average — growth that feeds itself without spending more on acquisition. In practice, a sustained K > 1 is rare and valuable; most products instead aim for a K that meaningfully lowers effective CAC without claiming pure virality. This tool's own growth dashboard computes its K-factor continuously, using exactly this formula, on real completed analyses.
The formula
K = invitations sent per user × conversion rate of those invitations
- Invitations per user
Every exposure a user creates: an explicit invite, a shared link, a public artefact someone else sees. Count them per existing user over a fixed period — and count the users who send zero, which is most of them.
- Conversion rate
The share of those exposures that become a new user — not a click, a user. A shared result that gets 100 views and 4 new sign-ups converts at 4%.
- K
New users each existing user brings, on average. K = 0.3 means every 100 users bring 30 more, who bring 9, who bring about 3 — a finite boost of about 43%. K > 1 means the chain never ends on its own.
The second number that matters is cycle time: how long from a user joining to their invitees joining. K = 0.5 with a one-day cycle outgrows K = 0.8 with a two-month cycle for a long while. Most "viral" products won on cycle time, not on K.
Worked example
What a K below 1 still buys you
- 1,000 users. 20% of them share once; each share is seen by 15 people on average: 1,000 × 0.2 × 15 = 3,000 exposures, i.e. 3 per user.
- 5% of exposures become a user: K = 3 × 0.05 = 0.15.
- Total users the loop eventually yields from those 1,000: 1,000 ÷ (1 − 0.15) ≈ 1,176. Referral added 17.6% on top of whatever acquisition paid for.
- Double the share of users who share (40%) and K becomes 0.3: 1,000 ÷ 0.7 ≈ 1,429, +43%. At a €500 CAC, that loop is worth about €215 of acquisition per paid user.
You don't need K > 1 for referral to matter. You need to know your K, because it is the multiplier on every euro of acquisition — and a multiplier nobody measures gets optimised by nobody.
Orders of magnitude
- A sustained K above 1 is rare and almost always temporary — the handful of famous cases (early social networks, some messaging and collaboration tools) had a product that was useless alone. For most products, 0.15-0.5 is the realistic range, and a reliable 0.3 is a serious competitive advantage.
- K is not stable over time: it falls as you saturate the network your early users belong to, and it differs by channel of arrival — users who came through a share tend to share more themselves.
- Watch the denominator: a K computed only on users who shared at least once is not a K, it's a share conversion rate. The formula divides by every user, silent majority included.
How to improve it
- Put the share where the value is, not in a menu: the moment a user gets a result worth showing is the moment to offer the share. A referral button on the settings page measures nothing but its own irrelevance.
- Make what gets shared worth receiving. The conversion half of K depends on what the invitee sees: a rich preview, a personal result, a concrete reason to try it themselves — not a generic "X invited you".
- Shorten the cycle: the faster an invitee becomes a user who shares, the more each K is worth. Remove every step between the shared link and the invitee's own first value.
- Instrument attribution before optimising anything: a reference parameter on every shared link, first touch, deduplicated for self-referrals. Without it you will be tuning a number you can't see.
In the Tour
Do you measure a viral coefficient or equivalent?
- Yes, tracked20 pts
- Vaguely aware of it7 pts
- Never measured0 pts
"Do you measure a viral coefficient or equivalent?" is the third Referral question — 20 for tracked, 7 for vaguely aware of it, 0 for never measured — after whether a sharing mechanism exists in the product and whether customers actually use it. Referral is the stage many teams score lowest on, and the one where a 0 is most often honest: the mechanism was never built, so there was nothing to measure.
Questions people ask
What is a good viral coefficient?
Above 1, growth would be self-sustaining — and almost nobody sustains that. In practice a good K is one you measure, that is above zero, and that you can move: 0.2 to 0.4 is a strong result for most products, and turns every paid customer into 1.25 to 1.7 customers. Judge it by what it does to your effective CAC, not by whether it crosses 1.
Viral coefficient or referral rate — what's the difference?
The referral rate is the share of users who refer at least one person; it is one input. K multiplies how many exposures users create by how many of those convert, over all users — it is the output. A 40% referral rate with invitations that never convert gives a K near zero.
How does Tour de Growth measure its own K?
Directly rather than through the formula: every shared result link carries a reference to the result it came from, and a Tour started from such a link is counted as referred — first touch, ignoring people re-taking their own Tour. K is then referred Tours divided by all Tours. It undercounts (a link copied by hand loses the reference), which is the honest direction to be wrong in.