What k factor means
K factor is the growth team's word for how many new users each existing user brings in. The formula is k = i × c, where i is the number of invites a user sends and c is the share of those invites that convert.
The idea comes from epidemiology, where the same number is the basic reproduction rate of a disease: how many people each infected person goes on to infect.
The thresholds follow from the formula. At k above 1 a product compounds on its own. At k of 1 it holds steady.
Below 1 the loop decays, though it still adds users on top of whatever else is bringing them in, and Dropbox ran at about 0.7 and WhatsApp at about 0.4 on their way to being worth what they became.
The discipline of building these loops belongs to a particular era. Andrew Chen's account of it, written in November 2025, puts the golden age at 2005 to 2010, when address book imports let a product send thousands of invites per user, and says that mobile ended the classic loop and the knowledge went with it.
Most AI and B2B products launched since then have no invite step at all. Their k factor as a product is close to zero, and their growth has to come from somewhere else.
A launch has a k factor too
A post on X spreads the way a product with a loop does. Someone sees it, and either it stops with them or it goes on to reach people who do not follow the account.
The carriers are shares, quotes, replies that pull the post back into feeds, and the recommendation system deciding to show it to strangers. Each carrier is an i, and the audience it reaches is a c.
The cleanest way to put a number on that is to compare where a launch ended up against where it started. Every account has a following, which is the audience the post would reach with no spread at all.
Views divided by followers is the launch's reproduction number: how many people the post reached for every one who already followed. The launch directory records both for every AI launch on X we track, so the number exists for all of them.
Measured on 16 September 2026, across 619 launches:
| Launches | Views per follower |
|---|---|
| Bottom quarter | 7 |
| Median | 38 |
| Top quarter | 168 |
| Top tenth | 603 |
The median launch reached 38 times the account's own following. 89 percent of launches reached more people than followed the account, and a third reached more than a hundred times as many.
The 71 launches that came in under 1, meaning the post reached fewer people than the follower count, are the ones that never left the timeline of the people who already knew the company.
Those are large numbers for anyone used to product k factors of 0.4. They are large because the loop on X runs in hours instead of weeks, and because the feed carries a post to people who took no action at all. Nobody had to send an invite.
The multiplier falls as the account grows
The number moves most with the size of the account posting. Smaller accounts get far more reach per follower, and the largest accounts get close to none.
| Account followers | Launches | Median views per follower | Median views |
|---|---|---|---|
| Under 1,000 | 119 | 150 | 53,907 |
| 1,000 to 5,000 | 165 | 91 | 232,843 |
| 5,000 to 20,000 | 140 | 37 | 340,722 |
| 20,000 to 100,000 | 99 | 14 | 611,338 |
| Over 100,000 | 96 | 1.4 | 709,893 |
Read the two right-hand columns together. Going from under a thousand followers to over a hundred thousand multiplies the account by more than a hundred and the median launch's views by 13.
The following sets a floor. The spread is what the launch adds, and it is worth the most to the accounts that have the least.
This is the same pattern the algorithm guide describes from the other side: the feed shows a post to people outside the following when the early signals are strong, and a small account's launch is nearly all outside reach. How the X algorithm works covers the mechanism. This page is what it produces.
Shares explain less than you would think
If the loop ran on viewers sharing, the launches with the highest share rate would be the biggest. In the directory the opposite holds. Split the 619 launches into quarters by reposts and quotes per thousand views:
| Shares per 1,000 views | Median views | Median views per follower |
|---|---|---|
| Lowest quarter (under 0.22) | 1,120,288 | 254 |
| Second quarter | 444,322 | 60 |
| Third quarter | 145,468 | 21 |
| Highest quarter (over 0.77) | 35,197 | 13 |
The launches that reached the most people were shared the least per view, by a factor of 30 in views between the top and bottom quarters. The correlation between share rate and views across the directory is negative and strong.
The median launch earns 0.42 shares per thousand views. Shares get counted because they are easy to count, and they are a poor guide to what a launch did.
What this says about the loop matters for anyone planning one. The bulk of a big launch's reach comes from the recommendation system deciding, early, to put the post in front of people who follow nobody involved.
That decision is made on the first hour's signals, and the accounts that carry the post in that hour set it. A launch's k factor is mostly decided before most of its viewers arrive. Preferential attachment is the reason the same names keep winning it.
A worked example
TypeSafe AI came out of stealth on 15 September 2026 from Diogo's account, which had 1,500 followers that morning. The post stands at 40,007,554 views, 16 of the 1978 launches we track. Against the following it launched from, that is a reproduction number above 11,000.
The directory shows a smaller multiplier of 26,672, because it records the account as it stands today, at 154,182 followers. The difference between the two numbers is the part of the loop that stuck: the viewers who followed. The case study has the day by day, including how the first hour was seeded.
How to use the number
Take the views on your launch post and divide by the followers the account had the morning it went up. That is your launch's k factor, and the table at the top of this page places it.
Under 7 is the bottom quarter. Above 38 is better than half the AI launches we have measured. Above 133 is the top quarter, and above 491 is the top tenth.
Judge it against your account size, using the second table. A launch at 20 times followers from an account of 200,000 is a very good result, and the same multiple from an account of 800 is below the median for its bracket.
Then look up your own category in the directory, which lists every launch by vertical with the same figures, so the comparison is against companies like yours.
Questions we get asked
What is a good k factor for a launch on X?
The median AI launch in our directory reached 38 times the account's followers. Above 133 is the top quarter and above 491 the top tenth. Accounts under 1,000 followers should expect more, with a median of 150, and accounts over 100,000 should expect close to 1.
Is k factor the same thing as viral coefficient?
Yes. Both names describe the same number, how many new people each existing one brings in, and the formula is invites per user times the conversion rate of those invites. Growth teams say k factor; investor decks tend to say viral coefficient.
Does a high share rate mean a launch went viral?
In our data it tends to mean the opposite. The quarter of launches with the lowest reposts and quotes per view had a median of 1.12 million views, and the quarter with the highest had 35,197. Big launches are carried by the feed showing the post to non-followers, and that decision is made on early signals from the accounts involved.
Can a product with no referral loop have a k factor?
Its k factor as a product will sit near zero. Its launches can still compound, because the loop runs on the network instead of inside the product. 155 launches in our directory passed a million views, and 59 of them came from accounts with fewer than 10,000 followers.