Where these definitions come from. Amazon does not publish the DSP scorecard's metric definitions. They live in the partner portal, behind a login, and they change. Everything below is attributed to the third-party source it came from, with the date that source was published or last updated. Your own portal is the reference. This page is not.
That warning is not boilerplate. As you will see, the published sources disagree with each other about something as basic as how many performance tiers exist.
What the scorecard actually decides
The scorecard is a weekly grade. What makes it matter is not the number but what the number is attached to: route allocation. Delivery stations hand out routes on performance, so a bad week is not a bad week — it is a smaller next quarter.
That is the part most guides underplay. An uncovered route on a Monday costs you the revenue of that route once, and it costs you standing in a ranking that decides how much work you get offered for months. The two costs are not the same size, and the second one compounds.
It is also why the scorecard is a retention problem disguised as a metrics problem. Almost every line on it is produced by a driver doing or not doing something at a doorstep, and drivers who have been with you a year do those things without being asked.
The metrics, and who says so
Three sources, read on 25 September 2026. I have named each one against each metric rather than blending them into one authoritative-sounding list, because blending them is exactly how a wrong definition becomes received wisdom.
| Metric | What it measures | Source |
|---|---|---|
| DCR — Delivery Completion Rate | "the percentage of assigned packages your team successfully delivers in a given day" | Nizod, 12 Apr 2026 |
| DCR — Delivery Completion Rate | "packages delivered as a percentage of attempted" | Hera, updated May 2026 |
| DNR — Delivered Not Received | "when a customer reports that their package was marked delivered but never arrived" | Nizod, 12 Apr 2026 |
| POD — Photo on Delivery | "whether your drivers are taking required delivery photos" | Nizod, 12 Apr 2026 |
| Contact Compliance | driver check-ins and route adherence | Hera, updated May 2026 |
| Customer Escalations | "complaints routed directly to Amazon about your delivery team" — rude behaviour, unsafe driving, damaged property, package mishandling | Nizod, 12 Apr 2026 |
| Customer Feedback | survey responses tied to deliveries | Hera, updated May 2026 |
| FICO / Safe Driving | "the aggregate driving safety performance of your fleet" — speeding, harsh braking, distracted driving, seatbelt compliance | Nizod, 12 Apr 2026 |
| Safe Driving | telematics on braking, acceleration, distracted driving | Hera, updated May 2026 |
Notice the first two rows. Delivered as a percentage of assigned and delivered as a percentage of attempted are not the same measurement, and the difference is every package a driver never got to. Two sources, both current, both published by companies selling to DSPs, describing the same three-letter metric differently.
I am not going to adjudicate between them. I am going to point at them, because knowing that the public definitions are inconsistent is more useful than being handed a confident wrong one.
No thresholds appear below. No public source states the numerical cut-off for any tier, and neither will this page. If you see a guide asserting that DCR must be above some specific figure, ask where it got it.
The tiers, where the sources disagree outright
Nizod, in April 2026, names six: Fantastic Plus, Fantastic, Great, Good, Fair, At Risk. It also states a $5,000 bonus each time Fantastic Plus is achieved.
Hera, updated May 2026, names four: Fantastic, Great, Fair, Poor.
Both are recent. Both are written by vendors who work with DSPs daily. They describe different scales.
The likely explanations are that the scale changed between the two dates, or that the two describe different programmes — and that second one matters enormously here, because almost all of this material is American. A dollar bonus is a straightforward tell. If you operate in the UK, treat every tier name, every bonus and every incentive in the online guides as unverified for your contract until you have seen it in your own portal.
That is the single most useful thing on this page, and it is the reason a British operator should not take US scorecard content at face value.
What moves each metric, in operational terms
Stripping the acronyms away, the scorecard rewards four things.
Packages actually delivered. DCR falls when routes go out short-handed, when a van is off the road, or when a driver runs out of day. Every one of those is a scheduling and fleet problem before it is a driver problem — which is why a depot that knows on Thursday that Saturday is a driver short scores better than one that finds out on Saturday.
Evidence at the doorstep. POD is a habit, and habits are set in the first week. A driver who was shown how to photograph a delivery on day one does it in month six; one who was left to work it out photographs the pavement. This is an onboarding metric wearing a delivery metric's clothes.
Customers not complaining. Escalations and feedback are the hardest to move because they arrive late and anonymously. The lever that exists is turnover: a driver in their third week is more likely to generate a complaint than one in their second year, and turnover is a thing you can act on.
Driving behaviour. Telematics scores — Nizod attributes FICO generation to Netradyne — come from the vehicle, not from paperwork. They correlate with insurance cost and they are the one part of the scorecard where the reading is continuous rather than weekly.
The week is the unit, and that is the opening
The scorecard closes weekly. Operators tend to read that as a deadline, which is the pessimistic reading. The useful one is that a week is long enough to save.
A route that went out short on Monday is already a fact by Monday evening. But the grade is not settled until the week is — so Wednesday is still early enough to matter, if on Wednesday you know where you stand. Most depots do not, because the scorecard arrives after the fact and nothing inside the week tells them the same story sooner.
The things that would tell you are unglamorous and already in the building: how many routes were covered against how many were offered, which drivers are approaching hours you would rather not push, which vans are booked in and when they come back. None of that needs a report. It needs the schedule and the fleet record to be the same record, so the question is a lookup rather than a reconstruction.
The British part nobody's scorecard guide mentions
Here is a failure mode that is invisible in American material and expensive here: a driver who cannot legally go out.
A Right to Work check that was never completed, a DBS certificate nobody chased, a licence check the driver never gave their code for — each of those is an administrative gap, and each of them ends the same way, with a person in the depot who cannot take a route. That lands on the scorecard as a delivery completion problem, and it gets managed as one, and the briefing about delivery completion changes nothing because the cause was a document.
The compliance floor and the performance grade look like separate concerns. They are the same concern seen a fortnight apart. Our page on running a DSP in the UK covers the checks themselves; what matters for the scorecard is that their expiry dates are attached to the person and visible before the week they bite.
The mistake that costs the most
Reading the DSP-level number and stopping there.
The DSP grade is an aggregate, and aggregates hide the shape of the problem. Five drivers at a comfortable standard and one having a genuinely bad month average out to something unremarkable — and the unremarkable average gets managed with a team-wide briefing that helps nobody, while the driver who needed a conversation never has one.
The useful question is never "what was our DCR". It is "which routes, on which days, with whom" — which means you need the assignment record to answer it. If who drove which van on which day lives in someone's memory, the scorecard tells you that something went wrong last week and nothing about where.
What to do with this page
Three things, in order.
Open your own portal and write down the metric names and definitions as they appear there, with today's date. That document, not this one and not any other guide, is what your managers should work from.
Then check whether you can answer, for last week, which driver ran which route in which van. If you cannot, the scorecard will keep being a report card rather than a diagnostic.
Then look at your turnover. Most scorecard problems that survive more than a month are retention problems that nobody has named yet.
Sources
- Nizod, How to Improve Your Amazon DSP Scorecard: A Complete Guide, published 12 April 2026.
- Hera, What Is an Amazon DSP? A Guide for Owners, last updated May 2026.
- Amazon's UK Delivery Service Partner pages — logistics.amazon.co.uk — consulted 25 September 2026, which publish programme and financial information but no scorecard metric definitions.
All three consulted on 25 September 2026. None of them is Amazon's documentation of the scorecard, because Amazon does not publish one.
How WaveGo helps with this
Amazon DSP software for UK operators — five seats free, no card and no contract.