nShift Track shows an estimated delivery date — when the parcel is expected to arrive. Depending on the shipment, it is a single deadline or a delivery window with a start and an end.
That date can come from six different places: from the carrier, from your own agreements and systems, or — when none of those have anything — from nShift ETA, nShift's own AI prediction. Track checks all six in a fixed order and uses the first one that has a value.
Because nShift ETA fills the gaps left by the others, you can expect an estimated delivery date on effectively all shipments, and the Estimated delivery fields are dependable enough to filter, sort, and report on across your whole volume.
Where the estimated delivery date comes from
Track evaluates the six sources in this order and uses the first one that returns a value.
| Priority | Source | Where the data comes from |
|---|---|---|
| 1 | Carrier return data ETA | The carrier sends a predicted delivery time together with its status events. Not all carriers support this, and for some it must be activated per account. |
| 2 | Carrier booking ETA | Reference fields 329 / 330 (Carrier ETA Start / End), populated by carrier web services at booking or routing. ISO format only. |
| 3 | SLA / configuration ETA | Your agreed transit times, from the carrier performance configuration. |
| 4 | ETA reference | Reference fields 292 / 293 (Carrier ETA earliest / latest arrival time), populated from carrier setup, Shipping Rules, or your own import. |
| 5 | Delivery date reference | Reference fields 110 / 111 (Delivery date start / end), populated by import or entered manually. |
| 6 | nShift ETA (AI prediction) | Predicted by nShift from historical delivery patterns. Used only when priorities 1–5 all return nothing. |
Why the order is what it is
Priorities 1 and 2 are the carrier's own data, and the carrier is the most authoritative source on its own network.
Priorities 3–5 are data you chose to supply — from your SLA agreements, your ERP, or your order system. nShift does not override your data with its own, even in cases where a prediction might turn out to be closer.
nShift ETA therefore sits last. It never competes with a date that already exists; it only fills the empty slots.
If you would rather use nShift ETA than your own configured dates, stop populating the corresponding sources. As long as reference fields 292/293 or 110/111 carry a value, or an SLA configuration exists, those win, and no prediction is made for that shipment.
Knowing which source was used
Reports do not just show the date — they also tell you where it came from, in the Estimated delivery source column, shown next to Estimated delivery start and Estimated delivery end.
| Value in the column | Priority | What it means |
|---|---|---|
| Carrier tracking update | 1 | The carrier sent a delivery estimate with its tracking events. |
| Carrier booking confirmation | 2 | Returned by the carrier when the shipment was booked (reference fields 329/330). |
| Carrier SLA configuration | 3 | Calculated from your agreed transit times. |
| Customer-provided estimate | 4 | A date you supplied yourself (reference fields 292/293). |
| Shipment reference date | 5 | A delivery date on the shipment reference fields (110/111). |
| nShift AI prediction | 6 | Predicted by nShift because none of the sources above had a date. |
In reports, nShift ETA appears as nShift AI prediction. If no source produced a date, the cell is empty.
Knowing the source matters more than it first appears:
- It tells you how much weight to give the date. A carrier-provided ETA, a date from your own SLA agreement, and an AI prediction are three different kinds of claims. The source column lets you — and your customer service team — tell them apart at a glance.
- It shows you where your data has gaps. If a whole carrier or market shows nShift AI prediction on every row, that carrier is not sending return data ETA and is not covered by an SLA configuration. That is worth knowing, and it is visible without asking anyone.
- It comes with you when you export. The source is included in the report export, so you can group or count by source in a spreadsheet — for example, to see how much of your volume is carrier-confirmed versus predicted.
Not every source fills in both dates
Sources differ in how precise they are, and the source column tells you which kind you are looking at:
- Carrier SLA configuration typically gives only an Estimated delivery end — a date stamped 23:59, meaning "by the end of this day". Estimated delivery start is empty because a transit-time agreement defines a deadline, not a window.
- nShift ETA returns a genuine window with both a start and an end — for example, 08:00 to 10:00 on the same day.
An empty Estimated delivery start is therefore expected behaviour for some sources, not a missing value.
Where you will see it
Shipment view
The estimated delivery date appears in the shipment detail view. To see which of the six sources it came from, use the Estimated delivery source column in reports.
Reports
Estimated delivery end is available as a report column and reflects the best available date for each shipment. Estimated delivery start is available for teams who want the full window. Both are sortable and filterable by date range, and both can be exported.
Shipment Data API
The shipment's deliveryStart / deliveryEnd fields are populated from the resolved date, and deliveryWindowSource indicate where those dates came from — value 6 for an nShift ETA prediction, the last source in the chain. This lets your own systems treat a prediction differently from a carrier commitment.
About nShift ETA — the AI prediction
nShift ETA is nShift's own estimate of when a parcel will arrive, generated from historical delivery data rather than from a carrier message or a contractual transit time. It exists to close the gap left by the five sources above: many carriers do not send a predicted delivery date with their status events, and not every account has SLA transit times configured. Those shipments used to show nothing at all.
When the prediction is made
- A prediction is made only if priorities 1–5 have returned nothing for that shipment.
- It is requested when the first event that opens the delivery clock arrives — the first tracking event with SLA status Open, typically the carrier's first "collected" or "in transit" scan.
- Until such an event arrives, no prediction is requested. A shipment you have just created, but which the carrier has not yet started moving, will normally show no estimated delivery date.
- The prediction is made once per shipment and stored. It is not recalculated as later tracking events arrive.
The trigger is the first SLA-open event rather than the registered print date, because a parcel is not always collected on the day it is registered. Starting from the event that actually opens the delivery clock produces a more realistic window. The same trigger is used for the shipment report, the shipment view, and the Shipment Data API, so all three agree.
If the prediction service is unavailable or cannot produce a result, the field is simply left empty. No fabricated date is shown, and no error is surfaced to you.
What the prediction is based on
nShift sends the prediction model a minimal, non-personal set of shipment attributes:
- The carrier product concept (the service being used)
- Origin address — city, postcode, country
- Destination address — city, postcode, country
- The shipment date
- The shipment weight
No recipient names, street addresses, contact details or order contents are used. The model returns a delivery window — a start and an end time — rather than a single timestamp.
The model is trained on nShift platform delivery history and is retrained periodically as more data accumulates, so accuracy improves over time.
Good to know
- A prediction is not a commitment. nShift ETA is a best estimate based on how comparable shipments have historically performed. It is not a carrier service-level agreement, and it is not used to measure carrier performance — carrier performance is still measured against your configured SLA.
- No configuration needed. There is nothing to set up on your side. Shipments with no other source start receiving a predicted window automatically.
- Accuracy varies by lane. Routes and services with a lot of historical volume are predicted more confidently than rare ones.
- Existing fields are unchanged. Reference fields 110/111, 292/293 and 329/330 continue to behave exactly as before. Nothing you supply today is overwritten or reinterpreted.