AI Optimization Platform
Case studies

No case studies yet — on purpose.

We’re in beta, and we will only publish a case study when a customer agrees to be named and the results come from completed, controlled experiments. Until then, here is exactly what each case study will contain.

Our standard

What every published case study will show

  • Baseline performance before any change, with the measurement period
  • What the AI changed — parameter, template or campaign — and who approved it
  • The control group size and how users were assigned
  • Incremental impact with confidence intervals and p-values
  • Sample-ratio and contamination checks
  • The cost of the optimization (subscription and usage) and the resulting ROI
  • What didn’t work, too
Example walkthroughs

How the workflow plays out

These scenarios describe the process. They are not customer results and contain no customer data.

Illustrative scenario — not customer results
Casual puzzle game: difficulty spike
  1. 1. Advisor detects one level with completion far below its neighbours.
  2. 2. Developer exposes a difficulty multiplier bounded to ±20%.
  3. 3. Autopilot runs a 50/50 experiment with next-day retention as guardrail.
  4. 4. The case study would report completion and retention deltas with CIs — or report a null result.
Illustrative scenario — not customer results
Subscription app: trial conversion
  1. 1. Advisor quantifies trial users who never reach the core feature.
  2. 2. Experience Evolution drafts onboarding variants from approved components.
  3. 3. A developer approves one variant for a 20% rollout.
  4. 4. Measured impact is trial-to-paid conversion vs. control within a 14-day window.