← PROJECTSretention-oracle / story

SYSTEM: Retention Oracle

LIVE
↗ OPEN

Why I Built This

Most retention tools require 365 days of data to tell you what your 365-day retention is. Useless. This one takes Day 1, Day 7, Day 30 — the data you actually have — and extrapolates the curve using power law regression. The verdict is deliberately blunt. You want the truth early, not a comfortable lie at the end of a year.

Background

Retention curves in consumer apps reliably follow a power law: R(t) = a × t^(−b). Fit that to Day 1, Day 7, and Day 30 data — the three points you always have — and you can project the full curve forward. The Oracle does exactly this, shows the curve, and delivers a Gonzo verdict based on where Day 30 sits and how the decay behaves.

It is not a replacement for long-term measurement. It is a signal: is this cohort clearly terrible, clearly healthy, or inconclusive? That question has an answer at Day 30. You do not need to wait for Day 365.

Tech Stack

Next.jsTypeScriptPower Law RegressionRecharts
STATUS:AVAILABLE FOR CONSULTINGTIME: BUDAPEST [--:--]

We use Google Analytics to understand how you interact with our site. This helps us improve your experience. No personal data is collected without your consent.

Read our full privacy policy