delivery · Customer outcome
Delivery app B2C & B2B
The starting point, work, and retained outcome record from an anonymized Lemon AI engagement.
Evidence note: The customer identity is withheld. The operating context, actions, and outcome figures are retained from Lemon AI’s previously published case record and were approved for this migration. Results describe this engagement and are not a promise of identical performance for another app or market.
Operating context
Starting point
- The team could not reliably find users likely to place frequent, high-value orders
Goal
- Raise ROAS
- Cut CAC
What Lemon AI did
- Trained model to predict users with higher order frequency and order value
- Built a prediction model for couriers
- Focused campaigns on the top 15% and 50% of users by D90 LTV
- Reworked the page for couriers seeking full-time work
- Expanded into three new countries
Measured results
| Outcome | Result |
|---|---|
| CAC | -34.7% |
| Employment conversion | +48.8% |
| LTV on Day 90 | Android +57.7% · iOS +31.7% |
| Order frequency on Day 90 | Android +23.8% · iOS +23.8% |
How to interpret the result
The outcome should be read in the scope shown above: its platforms, region, acquisition sources, budget band, objectives, and measurement period. The result does not isolate a universal causal effect and should not be compared with another case without aligning its metric definition and time horizon.
Read the measurement and forecasting methodology for source, attribution, aggregation, reconstruction, and prediction limitations.