App retention curve projector
Project your Day 90 and Day 365 retention from Day 1, Day 7, and Day 30 data — with benchmark context for your app category.
Enter your Day 1, Day 7, and Day 30 retention rates. The calculator fits a power-law curve to D1 and D30, projects D90 and D365, and compares your D7 input against the model prediction.
App retention curve projector inputs and results
How the projection works
The model fits a power-law curve R(t) = R₁ × t^b to your Day 1 and Day 30 inputs, where b is the decay exponent. At Day 1, t = 1, so R(1) = R₁ × 1^b = R₁ — the curve passes through your D1 input exactly. The exponent b is solved so R(30) = your D30 input. D90 and D365 projections are then read off the same curve at t = 90 and t = 365.
Day 7 is shown as an input but not used in the fit — it's compared against the model's prediction at t = 7 to give a sense of how well the curve fits your actual data. If D7 is significantly higher than predicted, your retention curve has a steeper early drop and a more stable tail than the model assumes, which generally means D90 and D365 projections are too pessimistic.
About this tool
This tool projects long-term retention from short-term measurements using a power-law decay model fitted to your Day 1 and Day 30 retention inputs. Enter D1, D7, and D30 retention percentages — the curve is fitted to D1 and D30, and D90 and D365 retention are projected from that fit. Day 7 is shown as both an input and a model check. Benchmark ranges for games, utility, and social apps are provided for context.
Frequently asked questions
What model is used to project retention?
The projection uses a power-law decay model: R(t) = R₁ × t^b, where b is fitted so that R(30) = your Day 30 input. This model is well-established for mobile app retention because it captures the rapid early dropout followed by a stabilising long-term tail better than a simple exponential model. The fitted exponent b is always negative (retention decreases over time), and its magnitude reflects how steep the dropout curve is.
How accurate are the D90 and D365 projections?
Reasonable for planning, not precise for forecasting. The power-law model fits many real app retention curves well through Day 90, but long-term projections (D180, D365) become less reliable as behaviour changes — seasonal effects, product updates, and engagement campaigns alter the natural decay. Use the projections as a directional estimate and recalibrate as you accumulate D90 data. If your actual D7 closely matches the model's prediction, the D90 projection is likely more reliable.
What does it mean if D7 doesn't match the model prediction?
The model is fitted to D1 and D30. If your actual D7 is higher than predicted, your retention curve has a faster initial drop followed by a more stable base — common in apps with a high-engagement core audience. If D7 is lower, you're losing users faster in the first week before stabilising. Neither is inherently bad, but it tells you something about where in the funnel to focus: early onboarding (if D7 is low) or long-term engagement (if the D1–D7 gap is large).
What are typical retention benchmarks by app category?
For games: D1 25–40%, D7 10–20%, D30 5–10%. For utility and productivity apps: D1 35–50%, D7 20–35%, D30 12–25%. For social and entertainment apps: D1 45–60%, D7 30–45%, D30 18–30%. These are broad ranges — top-performing apps in each category significantly exceed these benchmarks. A useful internal benchmark is your own trend over time: consistent improvement is more meaningful than comparison to industry averages.