K-factor

The classic viral coefficient is one edge of a loop:

K = i × c

i is invites (or shares) sent per user in a cycle. c is the conversion rate of those invites into new users. If K > 1 and the loop is stable, each user creates more than one new user and growth can explode — until saturation, quality drop or platform limits. If K < 1, virality still helps: it reduces CAC, but you still need other loops or paid fuel.

Walking through the generations

Approximate new users after n generations from a seed of S (ignoring churn and overlap):

generation 0: S
generation 1: S × K
generation 2: S × K²
total ≈ S × (1 − Kⁿ) / (1 − K)   for K ≠ 1

Seed = 1,000 activated users this week. Each sends 4 invites (i = 4). Invite landing converts at 15% (c = 0.15). K = 4 × 0.15 = 0.60. Cycle time ≈ 7 days.

Week 0: 1,000
Week 1: 600 new from invites
Week 2: 360
Week 3: 216
Sum of invite-driven users ≈ 1,000 × 0.60 / 0.40 = 1,500
in addition to the seed (geometric series after seed),
before retention and invite fatigue.

Raise conversion to 25% → K = 1.0 → each generation replaces itself. Raising c (better invite UX, clearer value on the invite page) often beats begging users to send more invites.

Spamming the address book inflates i and destroys c plus brand. Design for high-intent invites ("share this doc with your editor") over blast mechanics.

Two products with K = 0.6: cycle time 2 days means many generations per month; cycle time 30 days means you wait. If cycle time stretches, compounding stalls even when K looks fine on paper.

Worked collaboration tool: 10,000 weekly activated users, average i = 1.2 across all activated, c = 0.25. K = 0.30. Not explosive, but if paid CAC is $40 and a referred signup costs $0 marginal CAC, referred users still improve blended CAC. Raising c to 0.40 via a doc preview on the invite page lifts K to 0.48 without nagging people to spam.

Common mistakes

• Computing K on all signups. Activation leaks hide inside "virality."
• Chasing i with fake gates. "Invite 5 friends to unlock" crushes c and teaches users you are extractive.
• Ignoring cycle time. A pretty K that turns monthly is a slow engine.

Check yourself

i = 4, c = 0.15. What is K, and is that explosive?

K = 0.60. Helpful, not explosive. It still cuts CAC.

You want K = 1.0 with i = 4. What must c be?

c = 0.25. Often easier to raise the invite page than to beg for more invites.

Key takeaways

• K = invites per user × conversion. K < 1 still reduces CAC.
• Cycle time sets speed. Report K on activated users.
• Raise c before you inflate i. High-intent invites beat address-book blasts.

Further learning

Next, which loop fits the product: types of growth loops.