DPD’s flagship ‘Predict’ promise - did the parcel arrive in the customer-notified 30-min / 1-hour window? Below 92% means the brand promise is breaking and customers paid premium for nothing.
At a glance
Share of DPD Predict consignments that arrived within the predicted 1-hour or 30-minute time window the customer was notified about. Predict is DPD’s flagship customer-experience product: the day before delivery, the recipient gets a 1-hour window (“your driver will arrive between 14:00 and 15:00”) via SMS / email / DPD app. The promise is concrete: arrival inside that window counts on-slot. Arrival before the window opens, after the window closes, or on a different day all count off-slot. The brand’s premium pricing on Predict-tier services is paid for this specific promise; if it breaks, customers paid extra for nothing.
| What it counts | COUNT(predict_consignments WHERE actual_delivery_time BETWEEN predicted_window_start AND predicted_window_end) / COUNT(predict_consignments WHERE status = 'DELIVERED'). The window is whichever the customer was last notified about: typically 1-hour, sometimes refined to 30-minute on the day. |
| API endpoint | DPD’s tracking API exposes the per-consignment Predict slot via predictedSlot.start and predictedSlot.end, plus the actual deliveredAt timestamp. Card matches the two. |
| Window type | DPD has progressively tightened windows: legacy 4-hour (pre-2018), 1-hour Predict (current default), 30-minute Predict (premium add-on for high-value or B2B). The card honours whichever window was issued for that consignment. |
| What counts as on-slot | Actual delivery timestamp falls inside the window inclusive of both bounds (HH:00:00 to HH:59:59 on a 1-hour slot). Driver early-arrival within 5 minutes before the window starts is off-slot by DPD’s own published definition; the card mirrors that. |
| Slot refinement | DPD sometimes refines a 1-hour window down to 30 minutes on the morning of delivery via the DPD app. The card uses the most recent window communicated to the customer; if the customer received a 14:00-15:00 SMS the night before then a refined 14:30-15:00 push the morning of, the 14:30-15:00 window is the test. |
| Service level scope | Predict-only services. Standard DPD shipments (no Predict notification issued) are not in this card’s denominator. Predict was historically opt-in but is now the default on most B2C-grade DPD services. |
| Returns / RTO | Returns and RTS consignments are excluded (the customer-side promise does not apply). |
| Time zone | UK local (GMT or BST). Window timestamps and delivery timestamps both originate from DPD’s UK depot network. |
| Time window | 30D (rolling 30 days). |
| Alert trigger | <92%. The DPD-published benchmark for healthy Predict accuracy is 94 to 96 percent; 92 percent is the floor below which customer trust degrades materially. Below 88 percent is contract-renegotiation territory. |
| Roles | owner, operations |
Calculation
Calculated automatically from your DPD data. See the At a glance summary above for what the metric tracks and the worked example below for a typical reading.Worked example
A UK premium DTC wine subscription brand based in West London, around 3,200 Predict NextDay consignments per month, 80 percent residential delivery, 20 percent office. Customers pay £6.99 for delivery on top of the subscription; the brand markets the 1-hour window as a key benefit. Reading taken at 09:00 GMT on 12 Mar 26 for the trailing 30 days (10 Feb 26 to 11 Mar 26).| Sub-segment | Consignments | On-slot | On-slot rate |
|---|---|---|---|
| Greater London (M25) | 1,890 | 1,795 | 95.0% |
| Home Counties | 740 | 691 | 93.4% |
| Rest of UK mainland | 520 | 472 | 90.8% |
| Highlands and Islands | 50 | 41 | 82.0% |
| All Predict (this card) | 3,200 | 2,999 | 93.7% |
<92% is not tripped at the aggregate level. Five things to notice:
- The headline hides geographic spread. London is at 95 percent, Highlands at 82 percent, the aggregate sits comfortably above the alert because London is 59 percent of the volume. If the brand acquired 200 new customers in Inverness next month, the aggregate could drop below 92 percent without any operational change. Always look at the per-region split alongside the headline.
- “Off-slot” is mostly “delivered earlier than promised”, not late. Of the 201 off-slot deliveries, roughly 130 (65 percent) were delivered before the window opened (the driver was running ahead of schedule and chose to deliver rather than wait). DPD’s own definition counts this off-slot, customers experience it as “the parcel arrived before the SMS said it would”. Cross-reference with On-Time Delivery Rate: the same 201 parcels are mostly on-time on OTD but off-slot on this card. The two metrics tell different stories.
- The 71 truly-late (after window) deliveries are the customer-experience loss. A wine customer who was notified “between 14:00 and 15:00” and received the parcel at 15:42 is the customer who calls customer service. Each is a goodwill-credit candidate.
- The Predict promise is only as good as the SMS / email getting through. Of customer complaints about “Predict didn’t work”, 30 to 40 percent are not actually slot-accuracy issues, the customer never received the notification (filtered as spam, wrong mobile number, opted out of marketing emails). Pair with the merchant’s email-deliverability and SMS-OTP success rates.
- The “rate suddenly degraded” debug case. A 4 to 8 point sudden drop in Predict accuracy is rarely DPD’s network failing; it is almost always (a) DPD adding a new sortation hub that has not yet stabilised its routing, (b) a peak-period (Q4 BFCM, Christmas, Mother’s Day) where DPD pre-emptively widens windows to 2 hours but the dispatch system is still reading the 1-hour version, or (c) a depot capacity issue at one specific UK depot (Bristol, Hinckley, and Smethwick are perennial bottlenecks). Pair with OTD by Route to localise.
Sibling cards merchants should reference together
Predict slot accuracy is a customer-experience metric. Pair it with these to read the full Predict story:| Card | Why pair it with Predict Slot Accuracy | What the combination tells you |
|---|---|---|
| On-Time Delivery Rate | OTD measures “right day”; Predict measures “right hour”. A parcel can be on-time but off-slot, or on-slot but late if the window itself was scheduled wrong. | Both above 95: healthy. OTD high, Predict low: driver-routing issue (parcel arrives early). Both low: depot or network failure. |
| Late Shipments | Late shipments are the worst kind of off-slot. | Late count rising while Predict accuracy falls means the off-slot-after-window subset is growing. |
| Exception Rate | Exceptions reroute the parcel and almost always invalidate the original Predict slot. | Exception rate climbing predicts a Predict accuracy dip 24 hours later. |
| OTD by Route | Predict accuracy is depot-network-sensitive. One slipping depot can pull the headline down. | If Predict drops 3 points and route-OTD shows two depots in red, those depots own the loss. |
| DPD OTD by Sales Channel | Per-channel breakdown. | Marketplaces (Amazon Seller-Fulfilled-Prime) penalise off-slot deliveries via feedback; D2C absorbs them via WISMO tickets. |
Cross-connector: shopify.refund_rate | Downstream impact. Off-slot deliveries on premium customers drive refund and goodwill-credit requests. | Predict drops of 4+ points typically precede a 0.3 to 0.8 point refund-rate rise at 7 days. |
| Cross-connector: customer-service ticket volume (Zendesk, Gorgias) | Most direct downstream signal. “Where is my driver” tickets correlate strongly. | A 100-bp Predict drop in a week predicts roughly 15 to 25 percent more “where is the driver” tickets. |
| Cross-connector: SMS / email deliverability rates (Klaviyo, Twilio) | The Predict promise needs the notification to land. | If notification deliverability drops, Predict feels broken even when the slot is accurate. |
Reconciling against the vendor’s own dashboard
Where to look in DPD’s own portal: DPD Customer Portal → Reports → Predict Performance. The portal exposes per-account Predict accuracy across the same 30-day window, broken down by depot. The aggregate matches the card. For account-managed customers, DPD’s QBR slides include a Predict accuracy breakdown by depot, by service tier (1-hour vs 30-minute), and by failure reason (driver early, driver late, slot mis-issued). Why our number may legitimately differ from DPD’s portal:| Reason | Direction | Why |
|---|---|---|
| Slot refinement timing | Either | DPD sometimes refines a 1-hour slot down to 30 minutes the morning of delivery; the card uses the most recent communicated slot. The portal sometimes uses the original 1-hour slot. Reconcile by ensuring both views are using “final communicated slot”. |
| Driver-early definition | Ours stricter | DPD’s portal sometimes counts driver-early-by-5-minutes as on-slot via an internal grace window. The card enforces the slot bounds exactly. Expect a 1 to 2 point gap on routes where drivers run consistently ahead. |
| Time-zone | Boundary days off | Both views are in UK local; the rolling-window boundary may differ by a day depending on the portal’s “Last 30 days” definition. |
| Webhook lag | Ours stale for “today” | Predict-window updates come through DPD’s tracking webhook; lag of 30 to 60 minutes during peak periods is normal. The card’s most-recent-day reading slightly understates accuracy until the lag clears. |
| Failed-delivery handling | Either | DPD’s portal sometimes excludes failed-delivery consignments from the Predict denominator (the slot was issued but the driver never reached the door). The card excludes RTS but includes failed-attempt-then-redelivered consignments, scoring them off-slot. |
predict_off_slot_count = predict_consignments × (1 - predict_slot_accuracy). The breakdown of off-slot into “early”, “late”, and “wrong day” is on the deeper drill-down view; the card surfaces only the rate.
Cross-connector reconciliation:
| Card | Expected relationship | Causes of legitimate divergence |
|---|---|---|
| Customer-side time-window promises on other carriers (e.g. Amazon Same-Day, Uber Direct) | Each carrier has its own slot definition. Not a like-for-like reconciliation. | Use only for portfolio-level comparison if a brand uses multiple slot-window carriers. |
shopify.refund_rate | Lagged downstream signal. | Many drivers of refund rate; Predict accuracy is one input. |
Documentation cross-reference (DPD-specific). Predict slot accuracy is unique to DPD on the UK shipping connector set; no peer card exists on Royal Mail, Evri, or APC because those carriers do not issue 1-hour customer-facing slots. The closest peer is APC’s NextDay-9am SLA, which is a hard cutoff rather than a window.
apc.apc_nextday_9am_sla(premium-tier cutoff peer, not slot-window peer)