Logistics · Enterprise AI · National 3PL
Predictive ETAs that cut 'where is it?' calls by 92%
Static delivery windows became live, per-stop ETAs that update every few minutes and warn customers before a delay happens.
GoKafkaXGBoostMapbox
-92%
WISMO calls
Timeline
4 months
Team
4 engineers + 1 data scientist
The challenge
What was breaking
- 01ETAs were static guesses that eroded customer trust.
- 02The contact center was overwhelmed by 'where is my order?' calls.
- 03Dispatchers couldn't see delays coming until it was too late.
The approach
How we tackled it
- 01Trained a gradient model on live GPS, traffic, weather, and history.
- 02Re-scored every active stop on a rolling window.
- 03Pushed proactive delay alerts and re-route suggestions to dispatch.
The solution
What we shipped
- 01A live ETA per stop, refreshed every four minutes.
- 02Proactive customer alerts the moment a slip is predicted.
- 03A dispatch view that re-routes around emerging delays.
In pictures
A closer look
The results
-92%
WISMO calls
+19%
on-time rate
-33%
empty miles
4 min
ETA refresh
"Customers stopped calling because we tell them first. The contact center finally breathes."
Related work