How Novelty Effects Rule Tech

TL;DR: New products get a novelty boost — more word-of-mouth, higher engagement, better conversion. The danger is mistaking this temporary spike for sustainable traction. Most “viral” products were just novel, and the metric collapse comes about three months in, after the team has already built half a roadmap on assumptions that no longer hold.

What it means

Chen identifies the Novelty Effect as a hidden variable that distorts early product metrics in a consistent and dangerous way. New products benefit from curiosity-driven engagement that decays over time. If you optimize your product for novelty-inflated metrics, you’ll build for a user who doesn’t exist long-term — and by the time the inflation fades, you’ve already locked in design choices that catered to the wrong audience.

This is the trap behind every “we went viral on TikTok” story that ends three months later with the team frantically trying to rebuild retention from scratch.

The argument

Novelty inflates every metric. More organic growth, higher engagement, better conversion rates — all temporarily. The risk is building your growth model on these inflated numbers and then watching everything decay at once when the curiosity expires. This connects directly to invisible-asymptotes: the novelty ceiling is an early asymptote that’s especially hard to see, because the metrics look great right up until they don’t. If you mistake novelty for product-market-fit, you’ll optimize for the wrong user, ship the wrong features, and pitch the wrong investors.

The antidote is retention analysis. Look at cohort-level retention, not aggregate metrics. If early cohorts retain well but new cohorts don’t, you may be past the novelty window. (duolingo-growth, hook-model) This is why growth-as-compass depends on understanding what’s driving retention versus what’s driving churn — aggregate growth charts hide everything that matters about whether the growth is real.

The ChatGPT question

ChatGPT is the open question Chen’s framework should make you ask. The launch was enormously novelty-inflated — millions of people who tried it once, told their friends, never came back. By 2024, however, the cohort retention had clearly stabilized at a high enough level to support hundreds of millions of weekly active users per Sam Altman’s reflections (altman-reflections). The novelty wore off and the real PMF underneath was still there. Most products that ride a novelty wave don’t have anything underneath. The interesting test of the novelty-effect framework is how long it takes for the underlying retention to assert itself, and ChatGPT happened to have enough underneath that the answer was “fast enough to matter.”