Two minutes, no signup

What is a quietly falling rate costing you?

Numbers you already know. You will get a rough size of the problem, whether the slide is getting steeper, and an honest answer about whether Shadow can help — including if it can't.

If that trend simply continued · 30 days

—

fewer activations than if the rate simply held, taking it from 58% to about 56.3%.

At the value you gave, that is roughly 15,520 a month — your assumption, not our estimate.

Measured over the last 45 days rather than the full 90, it comes out between 388 and 465.

It is getting steeper. The last 45 days fell faster than the 45 before, at about 0.47pp a week. A decline that is accelerating is a different problem from one that is levelling off, and the difference is invisible in a single before-and-after number.

This is your own trend continued in a straight line for 30 days — arithmetic, not a forecast, and sensitive to the three numbers you typed. Shadow fits 21 days of real data instead of three readings, carries a confidence band and a worst case, and publishes a backtest showing how often that band actually held. Where a cohort is too thin to forecast honestly, it says so instead of guessing.

Shadow would work for you

~1,732 users / cohort / week6mo history

Your cohorts are large enough to forecast, you have enough history to backtest against, and your events are somewhere we can read.

You clear every threshold Shadow gates on. A pilot takes 3–5 working days and about four hours of your team's time.