Where the loops went
K factor was built for products that spread through their users. The formula, invites per user times the conversion of those invites, describes an address book import, a referral credit, a document shared with a colleague who then needs an account.
Andrew Chen's history of the discipline puts its peak between 2005 and 2010, and ends with mobile closing off the invite volume that made it work.
The products being launched now mostly have no loop to design. A coding agent, a world model, a security product, an underwriting tool: the buyer uses it at work, and nothing about that use puts it in front of a second buyer.
Product-led growth teams spend years adding a share step that the product did not need. Measured honestly, the k factor of most B2B and AI products is close to zero, and their growth is bought, one impression at a time.
Paid reach has a k of zero
An ad reaches the person it is shown to and stops. There is no second host. Spend more and it reaches more people, at the same price per person, and when the spend stops so does the reach.
That is the definition of a loop with k of zero, and it is fine as far as it goes. It is also where most of a B2B marketing budget goes, and it misses the technical buyer, who scrolls past the promoted label.
A launch is the other kind of reach. The post is shown to the following, and then to the people the feed decides should see it, and then to the followers of everyone who quotes it.
Each of those is a host that costs nothing per person reached. Whether it compounds is a question of the first hour, and that can be engineered.
The launch is the loop
In the launch directory, measured on 16 September 2026, 155 of 619 launches passed a million views. 59 of those came from accounts with fewer than 10,000 followers, and 103 from accounts under 50,000.
The median launch reached 38 times the account's following, and the median launch from an account under a thousand followers reached 150 times it.
Two from the directory, both from companies we have no connection to. Superblocks announced from an account of 6,735 followers and the post did 19,478,893 views, 2,892 for every follower.
Wispr Flow launched from 35,168 followers to 19,382,073. Neither product has an invite step. Both had a day where the k factor of the company was in the thousands.
Nothing else a company without a loop does compounds like this. A launch is the one moment when the network does the distribution, and the reason companies that have run one keep running them.
What stays when the spike is over
Chen's warning about viral loops is that a high k with weak retention is a chain letter: the spike passes and nothing is left.
The same is true of a launch, and on X the unit of retention is the follow. A viewer who follows can be launched to again, at no cost, and the next launch starts from a larger floor.
TypeSafe AI's founder went from 1,500 followers on the morning of the launch to 154,182, about 0.4 percent of the post's 40M views. The company account went from 200 to 61,044.
That is the retained part of the loop, and it is why companies that launch repeatedly did a median of 568,056 views per launch in the directory against 231,359 for companies that launched once. Odyssey is four launches into the year on this pattern, and the Odyssey-3 case study shows what each one started from.
What this changes in the plan
Treat the launch as the growth loop and resource it that way. The founder's account is the host, because smaller accounts reach more people per follower and readers reply to a person.
The first hour is where the multiplier is set, so the accounts that will carry the post are chosen and briefed before it goes up.
The follow is the conversion that matters on the day, and the signup is measured separately, with attribution in place before the post. Then the next launch is planned from the new floor, on the cadence the repeat launchers in the directory run, which averaged 25 days apart.
Before any of that, look up your category in the directory. Every launch there carries the views, the account size and the multiplier, so the plan starts from what companies like yours have done.
Questions we get asked
Can a B2B product grow without a referral loop?
Yes, and most do. The loop moves out of the product and into launches, where the network carries the post to people who do not follow the company. In our directory, 155 launches passed a million views and 59 of them came from accounts under 10,000 followers.
Is a launch a substitute for paid acquisition?
It is a different kind of reach. Paid reaches each person once at a fixed price. A launch's reach compounds through the feed and the accounts that quote it, at no cost per person, and it reaches the technical buyer who skips promoted posts.
Most companies we work with run both, and the launch is the part that grows the floor for next time.
What is the k factor of a launch if the company has zero followers?
Views divided by followers is undefined at zero, which is the honest answer, and the directory's smallest accounts show what happens near it. Accounts under 1,000 followers reached a median of 150 times their following. The post has to be carried by other accounts in its first hour, because there is no following to start it.