Ingest data & check accuracy
The Ingest & Scoring view is where you feed Intelligence shipment data and see how accurate its predictions turned out to be. Predictions run at ingest, and once a shipment delivers, they are graded against what actually happened.
Open Ingest & Scoring
Upload analysis CSVs
Under Upload Analysis CSVs, click Choose CSV files and drop your analysis exports, one row per scan. The shipment, its SLA promise, and the actual delivery are all read straight from the file, so you do not configure anything.
- Uploads run in the background, and large files are split into batches automatically.
- Re-uploads dedupe by content, so uploading the same window every 6 to 12 hours is safe. You will not create duplicates.
- Each file shows a status (uploading, processing, done, or error) and, on success, how many shipments and new events it added.
To push data programmatically instead, the same data goes through the ingest API. See Ingest tracking data.
Read the Prediction Scorecard
The Prediction Scorecard grades ETA and on-time predictions against actual delivery. The stat tiles summarize it:
| Tile | Means |
|---|---|
| SCORED | Shipments that have been graded |
| DELIVERED | Of those, how many delivered |
| ON TIME | Share delivered on time |
| IN BAND | Share where actual delivery fell inside the predicted confidence band |
| MEDIAN ERR | Median absolute ETA error, in hours |
The table lists each scored shipment with its promised date, first and latest ETA, actual delivery, ETA error, whether it landed in band, and whether it was on time. The view shows up to 250 rows; use Download CSV for the full set.
An empty scorecard means nothing has scored yet. Upload shipments and events, and rows appear as predictions grade themselves.