How to improve your Amazon Cortex scorecard

Every Amazon DSP owner has the same Monday morning ritual: open Cortex, look at the weekly scorecard, and try to work out which driver, which route, or which day dragged the numbers. If your status moved from Fantastic Plus to Fantastic — or worse, from Fantastic to Great — you've got a week to figure out what broke and prove you can fix it.

The good news: scorecards almost always slip for predictable, fixable reasons. Below is the per-metric playbook we see working with UK Amazon DSPs.

What the Amazon Cortex scorecard actually measures

The Cortex DA scorecard rolls six core metrics into a weekly status banner (Fantastic Plus, Fantastic, Great, Fair, Poor). Each metric is a leading indicator of how well your fleet executed against Amazon's promise to the customer. Understanding them in plain English is the first step:

• DCR (Delivery Completion Rate) — the share of assigned stops that were actually completed. Misses come from undelivered packages, returns to station, and incomplete waves.
• DPMO (Defects Per Million Opportunities) — physical defects: damages, missing items, wrong addresses, lost packages. Counted per million stops, so even small absolute numbers move the metric.
• DNR (Delivered Not Received) — customer-reported claims that a package was marked delivered but never arrived. Often tied to specific drivers or specific addresses.
• POD (Photo on Delivery) — compliance with the photo-on-delivery requirement for drop-offs. Low POD scores almost always cluster around a small number of drivers.
• CC (Contact Compliance) — drivers using the Mentor app to contact customers when deliveries can't complete. Skipped contacts compound into DNR and DCR misses downstream.
• CDF (Customer Delivery Feedback) — direct customer feedback rolled up into a score. Soft signal but it correlates tightly with CC and POD.

Why scorecards drift, even when nothing has obviously gone wrong

Most DSPs we talk to don't have a "bad fleet" problem. They have a visibility problem. You see the weekly scorecard on Monday — but the events that produced it happened across hundreds of stops, on ten different routes, over seven days. By the time you're trying to fix it, the people involved have forgotten the specifics.

The structural fix is to look at metrics live, per driver, mid-week — not as a summary after the fact. The tactical fixes below assume you can.

DCR — lift it with route-handoff discipline

DCR drops happen on the routes that finish late, run out of time, or get returned partially complete. Two interventions:

• Track planned vs actual finish times per route. Routes that consistently finish in the red are either over-stopped or assigned to drivers who need a route swap.
• Cut "return to station with packages" events. When a driver is on track to RTS with parcels, get them on the phone before the end of the wave — almost always you can either reroute the remaining stops or get them a one-hour extension.

DPMO — fix the upstream van check, not the downstream defect

Most DPMO defects you can prevent — damaged parcels, wrong-address attempts, missing items — trace back to two events: a sloppy van load and an unchecked van. Both are upstream of dispatch.

A daily van inspection with photos (yes, every day) catches the small issues that turn into customer complaints by Friday. And a pre-departure check on package counts catches the missing-item DPMO hits before the van leaves the station. The drivers who skip these checks are usually the drivers driving your DPMO.

DNR — investigate every claim, even when it feels unfair

DNRs feel like the metric you can't control, but the patterns are obvious once you look. Claims tend to cluster around three things: specific drivers, specific addresses, and specific times of day (usually evening drops in low-light conditions). Pull the last 30 days of DNR claims, sort by driver, and the top two names will almost always be over-represented. That's where the conversation starts.

POD — make it visible to drivers daily, not weekly

POD compliance is the easiest metric to fix because it's a behaviour, not a skill. The drivers with low POD scores almost always know exactly which photos they skipped — they just don't see the consequence until your Monday meeting. Surface every driver's personal POD score back to them daily, on their phone, and the metric moves within two weeks.

CC — train it as a habit, not a one-off briefing

Contact Compliance is about drivers actually using the Mentor app's contact buttons when they can't complete a delivery. The fix is repetition: review CC misses in the morning huddle, the same way you review safety. After three weeks of consistent feedback, drivers internalise it.

CDF — the lagging indicator that confirms the rest is working

Customer Delivery Feedback rises automatically when POD and CC rise. There's no separate lever for it. If the other five metrics are improving and CDF isn't, look at handover quality — are drivers leaving packages in places customers don't expect, or rushing the doorstep interaction.

The Monday-to-Friday review rhythm

Owners who consistently hit Fantastic Plus run the same loop:

• Monday: review the previous week's scorecard with the OSM and identify the one or two metrics most at risk.
• Tuesday–Thursday: short morning huddles focused on yesterday's actual driver-level numbers, not the fleet average.
• Friday: a quick scan of the week's trajectory — escalations for any driver consistently below standard, recognition for the top performers.

The thing that breaks this loop, every time, is data. If your OSM is rebuilding metric tables in Excel every morning instead of running the huddle, the loop dies by Wednesday.

Where DSPOps fits in

DSPOps connects to Amazon Cortex and pulls every scorecard metric in live, per driver and per route. Your OSM opens the dashboard, sees the three drivers most likely to drag this week's score, and runs the huddle in two minutes. Drivers see their own metrics inside the driver portal — the ones who care about being top of the leaderboard self-correct without a meeting.

If you want to see what your DSP's scorecard looks like inside DSPOps, you can book a 20-minute demo — we'll set up your data live during the call.