The viral spike that taught the wrong lesson
Forty thousand signups in a weekend. Six weeks later, retention was the same as before the spike. The lesson wasn't what the chart said.
A team I worked with had a tweet go semi-viral on a Saturday. Forty thousand signups by Sunday night. The dashboard looked like a hockey stick discovered religion. Monday morning the team was already talking about which feature had done it and how to do it again.
Six weeks later, weekly active users were within four percent of where they had been before the spike.
The forty thousand had behaved exactly the way a crowd that arrived because of a tweet behaves. They signed up, looked around, did not complete onboarding, and quietly never returned. The product had not changed. The audience had not changed. What had changed was the team's confidence in a story about what worked — a story that was, as far as the underlying behavior was concerned, almost entirely wrong.
This is the most expensive thing a viral spike can do. The spike rewrites the team's intuition about which lever produced which outcome, and almost every rewrite is in the wrong direction. The next quarter of roadmap decisions gets made in service of reproducing the spike. The work to fix actual retention — boring, slow, unsexy — gets quietly deprioritized because the company is busy chasing the costume.
Product-market fit is not visible at the top of the spike. It is visible six weeks later, in the shape of the line where the spike used to be. If retention closed back to baseline, the product did not move; the audience did, for reasons unrelated to anything the company controls. If retention stepped up by even a small amount, something happened, and the question becomes what made the step, and is it reproducible.
The team I started with eventually went back and looked at the four percent. Three quarters of it was a single cohort of signups who all came from one specific reply thread on the tweet — a thread that described a use case the team had never marketed. That use case became the next quarter's positioning. It was the only thing in the spike that was actually a signal. Everything else was weather.
The honest read of a viral moment is humbling. The peak is noise. The baseline shift is the data. Most of the time, there is no baseline shift, and the lesson is to keep doing what you were doing before the chart misled you.
Don't optimize for the spike. Optimize for what's left after the spike forgets you.