Most last-mile dashboards in retail logistics are written in the past tense. They tell you what last week’s on-time rate was, what the cost per mile was, and who hit their SLAs. The trouble is that the operation they describe is still running, and most of the decisions that actually move it have to happen before any of those numbers are final.
In a practical week, that means handling whatever the network is doing in real time: a carrier that’s gone soft in the Northeast, a promotion that doubled Phoenix volume overnight, a regional partner that missed three SLAs in a row before anyone caught it. The team that handles those moments well usually has a sharper read on a smaller set of numbers all the way through the week.
That difference matters more in retail every quarter. Last-mile delivery now accounts for roughly 53% of total shipping cost, and at that ratio, sharper triage translates straight into margin.
Six metrics keep coming up in the operations that have figured this out, even though none of them are particularly new.
Delivery Promise Accuracy
Did the order show up when you said it would? McKinsey’s research doesn’t mince words: customers care more about reliability than speed.
Promise Accuracy = Orders Delivered Inside the Promised Window ÷ Orders With a Delivery Promise
Ironically, most retailers don’t track this metric well. Stord’s 2025 mystery shop found that just 34% of orders arrived as promised, and 40% of brands skip the promise entirely, keeping checkout vague because they can’t reliably hit a hard window.
Track this by delivery provider and by market. When Phoenix keeps missing your two-hour promise while Atlanta hits it clean, the move is to widen the Phoenix window. Pushing the delivery provider for faster runs rarely works.
Cost Per Delivered Order
Cost per delivered order is the fully loaded number to get one package to a customer’s door, with everything that hides in other line items folded back in.
Cost Per Delivered Order = Total Last-Mile Cost ÷ Successfully Completed Deliveries
That total has to include reattempts, refunds, support time on WISMO calls, and the return trip when an order comes back.
Get it right by slicing the metric by market, basket size, and provider. That’s how you can scale properly by finding out same-day works in one market and fails consistently in another one. Connected delivery APIs are how teams get the data clean enough to act on it.
First-Attempt Delivery Success Rate
Every order that needs a second visit costs you twice in driver time, fuel, and customer trust.
First-Attempt Delivery Success = Deliveries Completed on the First Try ÷ Total Attempts
The industry has gotten better here. Parcel Perform’s Q2 2025 data showed first-attempt success at 99.22%, on-time at 98.5%, and the issue ratio falling from 7.09% to 4.3%. Better routing, tracking, and recipient prep all helped.
When the metric slips, the carrier isn’t usually the cause. The address came in malformed, the customer never saw the notification, or the apartment had no instructions for an unattended drop. In dense urban markets, the pattern shows up by building type, parking, and time of day.
Most fixes happen upstream of the driver.
Exception Rate and Time to Recovery
Things go wrong on every route. What matters is how fast you find out and fix it.
Exception Rate = Delayed, Canceled, Reassigned, Failed, Damaged, or Unassigned Orders ÷ Total Orders
Time to Recovery = Time From Exception Detection to Resolution
Networks have gotten more complex. In-house fleets, marketplace couriers, regional carriers, and store fulfillment all run together. Deloitte’s 2026 outlook found just 30% of retailers use AI for supply chain visibility today, climbing to a projected 41% within a year.
Reporting exceptions after the fact is also too slow to save the order. Real exception management runs live and automated: the system categorizes issues as they surface, auto-escalates the easy ones, and runs recovery playbooks without a human in the loop. That’s the layer workflow automation takes care of.
Customer Visibility and WISMO Deflection
A delivery can be technically perfect and still feel terrible if the customer has no idea where it is. That’s WISMO (“where is my order?”), the largest single source of support tickets in retail logistics.
McKinsey’s research also found that about half of consumers actively track an order to confirm it’s on schedule. Descartes’ 2025 study is even sharper: 79% of consumers under 35 had a delivery issue in a single three-month window, and only 11% reported consistent satisfaction.
Solve this by tracking ETA accuracy and WISMO tickets per 100 orders. Most teams overcommunicate generic updates and under-communicate the ones that actually matter, like the package running 90 minutes late or the one already at the door.
Returns Cycle Time
The job isn’t done at the doorstep. Returns are part of the same delivery promise the customer judged at checkout, and they’re getting bigger every year.
Returns Cycle Time = Time From Return Request to Pickup, Inspection, Refund, Exchange, or Restock
NRF and Happy Returns projected $849.9 billion in U.S. retail returns in 2025. That’s 19.3% of online sales coming back, with 82% of consumers calling free returns important when they shop online. Reverse logistics affects fleet capacity, store ops, inventory, and repeat-purchase rate at the same time.
Track pickup success, cycle time by return method, refund speed, and restock rate. The shape of a return often determines whether the customer comes back, sometimes more than the original delivery did.
How Burq Puts These Metrics to Work
The framework you just read is the one we live inside at Burq, and it’s the spine of how we built the platform.
Burq sits between your storefront and your delivery network as a single orchestration layer. Pulse AI handles the live work dashboards can’t: scoring delivery promises at checkout, picking the cheapest viable provider for each order, catching exceptions in flight, running the customer messaging that keeps WISMO down, and orchestrating returns on the same rails as outbound. Every metric in this piece becomes something the system handles automatically, across both your in-house drivers and your full provider network.
If that sounds like the operation you’re trying to build, drop us a line.
