From Tracking Dashboards to Risk Signals: Why Retailers Are Starting to Measure Delivery in Money and Experience
Most eCommerce operations have solved the first delivery-data problem. If a customer asks where their parcel is, someone can look it up, and usually the customer can look it up themselves on a tracking page. Ten years ago that was hard. Today it is table stakes, and the question that comes up in operations reviews has moved on. It is less "where is this parcel?" and more "what are late deliveries actually costing us this week?" and "are our customers getting a good experience, and how would we know?"
Those are harder questions, and the tracking dashboard was never built to answer them. What we are seeing, across the retailers we talk to and in the public research, is delivery analytics moving from describing what happened to quantifying what is at stake. This post is about why that is happening now, what the new signals look like, and what tends to get in the way.
Visibility did not make the problems go away
It is worth starting with how much still goes wrong, because the dashboards can make it feel under control. In Descartes' 2025 Home Delivery Consumer Sentiment Study, which surveyed 8,000 consumers across Europe and North America, 66% said they had experienced a delivery problem in the first quarter of 2025, and among 18 to 35 year olds the figure was 79%. The previous year's edition of the same study found 67% had faced a delivery problem, and 63% of those took some action that had negative consequences for the retailer or the delivery company. Narvar's 2025 State of Post-Purchase report, based on 3,461 US consumers, found 74% had experienced at least one late delivery in the past year.
None of that is a visibility failure. The retailer could see those parcels. The point is that seeing a shipment is late, after it is late, is a fairly weak position to be in. It tells you about the exception. It does not tell you what the exception costs, whether it is part of a pattern, or which customers you are about to lose because of it.
Why the money question arrived
A few things have pushed delivery from an operations metric to a finance one.
The first is that customers have shown, repeatedly, that they change their behaviour after a bad delivery. In the Narvar research, 50% of consumers said they were less likely to shop with a retailer again after a late delivery, and 64% said the risk of a delivery problem sometimes or regularly stops them from ordering at all. In Descartes' 2024 study, 21% of consumers said negative delivery experiences would put them off buying online. Metapack's 2025 Ecommerce Delivery Benchmark Report, produced with Retail Economics, found 60.1% of British shoppers would switch retailers for more convenient delivery options elsewhere. The flip side is also true. In Ryder's 2025 consumer study, 96% of people who had a positive delivery experience said they were more likely to shop with that retailer again. So a late parcel is not only a cost to reship or refund. It is a probability change on the next order, and that is a revenue number even if nobody has written it down.
The second is returns, which are the largest single pool of post-purchase money most retailers do not manage as money. The National Retail Federation and Happy Returns estimated that US consumers would return $849.9 billion of merchandise in 2025, which is 15.8% of sales, with an online return rate of 19.3%. A return that has been requested but not yet processed is a refund the business owes. If your reporting cannot show how much of that liability is sitting in the queue today, you are carrying an unquantified number on the balance sheet, and finance teams have started to notice.
The third is peak, where margins get thin and exposure gets large. Deloitte's September 2026 forecast puts US holiday retail sales at $1.70 to $1.71 trillion, with e-commerce growing 7.5% to 8.4% to around $316 to $319 billion. Salesforce's 2026 holiday predictions estimate retailers will spend an additional $3 billion globally subsidising free shipping this season. When you are already paying to ship, every failed or late delivery is money spent twice, and it is usually spent in the weeks when the operation has the least slack to recover it.
The three signals that are replacing the dashboard
When we look at what the more advanced retailers are building, or asking their vendors for, it tends to come down to three signals. They are related, but they answer different questions.
Delivery risk. This is the question of which shipments are going to miss the promise, ideally before they miss it. Carrier performance sets the floor here. ShipMatrix's analysis of peak 2025 put on-time performance at 97.2% for UPS, 95.3% for FedEx and 94.1% for USPS in December, which was better than the year before, but still means somewhere between roughly one in thirty-five and one in seventeen parcels missed. Across the 23.1 billion parcels Pitney Bowes counted in the US in 2025, that is a lot of individual promises. The retailers who are ahead here are not just tracking carrier scan events. They are measuring performance against the delivery date they actually promised the customer, by carrier and by lane, and using that history to flag which open shipments are most likely to slip.
Revenue exposure. This is the question of how much order value is currently tied to something going wrong: shipments running late against the promise, deliveries that failed today, orders stuck before they ship, and returns waiting to be refunded. The cost side has been quantified before. Loqate's Fixing Failed Deliveries research, which is from 2021 and a little dated now, put the failure rate of first-time US deliveries at 8% with an average cost of $17.20 per failed order. What has changed is that retailers want the number daily, in their own currency, on their own orders, and they want it organised around what can still be recovered rather than what has already been lost. There is a subtle but important design point in this: an order that is both late and has a return pending has to be counted once, in whichever bucket is worse, otherwise the total is not a number anyone will trust in a finance meeting.
Experience scoring. This is the question of what the customer actually experienced, expressed as something you can trend and compare. Forrester's 2025 consumer data found 80% of US online adults consider an order confirmation with a specific delivery or pickup date important, and about three quarters value status update notifications, which tells you the experience is about more than the parcel arriving. A workable score combines what happened on the shipment, meaning speed against promise, whether the first attempt succeeded, what the customer was told along the way, and how any return went, with what the customer said afterwards. The important discipline, which not everyone applies, is checking the score against real customer ratings, so it is tested rather than asserted.
What tends to get in the way
In our experience the obstacle is usually not the analytics themselves, but the state of the data underneath them.
The signals above only work if orders, shipments, carrier events, fulfillment status, returns, notifications and customer feedback live in one model and refer to the same order. In most retailers they live in four or five systems, so the first six months of any "delivery intelligence" project is a stitching exercise, and a home-grown analytics layer starts with zero history and the joining problem unsolved. That is not a reason not to do it, but it is worth going in with open eyes about where the effort goes.
The second obstacle is that the investment is moving faster than the strategy. A Gartner survey published in August 2026 found 67% of supply chain digital investment is now allocated to AI, yet more than half of chief supply chain officers were uncertain about the return. An earlier Gartner survey, from June 2025, found only 23% of supply chain organisations had a formal AI strategy at all. The honest reading is that a lot of money is being spent on intelligence without a clear answer to what decision it is supposed to improve. Delivery risk, revenue exposure and experience scores are useful precisely because each one has an obvious owner and an obvious action.
The third is that these signals increasingly need to be readable by software, not just people. Gartner predicts that by 2031, 60% of supply chain disruptions will be resolved without human intervention. Whether that timeline holds or not, the direction is clear. A revenue-at-risk number that lives in a slide is something a person acts on eventually. The same number exposed to rules, and to AI agents, is something that gets acted on today. That puts a premium on a governed, normalised data layer rather than a prettier dashboard.
If you are starting now
A few practical suggestions, from what we have seen work. Start from the delivery promise, not the carrier scan, because everything else is measured against it. Get all the post-purchase signals into one model before you build any report on top, even if the first version is narrow, and make sure each order is only counted once across whatever buckets you define. Calibrate any experience score against what customers actually said, and be prepared to adjust the weights. And build it so that the same numbers a person reads can also trigger a rule or be handed to an agent, because that is where the value compounds.
At Carriyo we have been working on exactly these questions for our own customers, and the reason we could is that the platform already holds orders, shipments, fulfillment, returns, notifications and feedback in one model. On Friday we will write about the two reports we have built on top of it, Revenue at Risk and CX Score, and when they open to customers. If you would like to talk through how this applies to your operation before then, our team is happy to have that conversation at carriyo.com/contact.
---
Sources
1. Descartes Systems Group. "Descartes' Annual Ecommerce Study Shows Younger Consumers Driving Online Buying" (2025 Home Delivery Consumer Sentiment Study, 8,000 consumers, Europe and North America), 14 May 2025. https://www.descartes.com/resources/news/descartes-annual-ecommerce-study-shows-younger-consumers-driving-online-buying 2. Descartes Systems Group. "Descartes' Annual Ecommerce Study Shows Online Buying Grows, 67% of Consumers Face Delivery Problems" (2024 study), 7 May 2024. https://www.descartes.com/resources/news/descartes-annual-ecommerce-study-shows-online-buying-grows-67-consumers-face 3. Narvar. "Nightmare on Main Street: Anxious Online Shoppers Want a Happy Ending" (2025 State of Post-Purchase Report, 3,461 US consumers), 19 November 2025. https://corp.narvar.com/blog/nightmare-on-main-street-anxious-online-shoppers-want-a-happy-ending 4. Metapack (Auctane) with Retail Economics. Ecommerce Delivery Benchmark Report 2025, as reported by The Retail Bulletin, 14 February 2025. https://www.theretailbulletin.com/retail-solutions/new-metapack-survey-how-consumer-expectations-are-reshaping-ecommerce-after-a-quarter-century-of-retail-transformation-14-02-2025/ (methodology: https://www.retaileconomics.co.uk/retail-insights/thought-leadership-reports/ecommerce-delivery-benchmark-report-2025-metapack-auctane-retail-economics) 5. Ryder. 2025 Ryder Consumer Study (last mile), 2025. https://www.ryder.com/en-us/insights/white-papers/last-mile/2025-ryder-last-mile-study 6. National Retail Federation and Happy Returns. "Consumers Expected to Return Nearly $850 Billion in Merchandise in 2025" (2025 Retail Returns Landscape), 15 October 2025. https://nrf.com/media-center/press-releases/consumers-expected-to-return-nearly-850-billion-in-merchandise-in-2025 7. Deloitte. "Deloitte Forecasts Holiday Retail Sales to Reach $1.70 Trillion to $1.71 Trillion," 10 September 2026. https://www.prnewswire.com/news-releases/deloitte-forecasts-holiday-retail-sales-to-reach-1-70-trillion-to-1-71-trillion-302874356.html 8. Salesforce. "Holiday Retail Predictions 2026," 20 July 2026. https://www.salesforce.com/blog/holiday-retail-predictions-2026/ 9. ShipMatrix. "Parcel Carriers' On-Time Performance in Peak of 2025," 12 January 2026. https://shipmatrix.com/wp-content/uploads/2026/01/SMx-on-parcel-carriers-OTP-in-peak-of-2025_Jan-12_2026.pdf 10. Pitney Bowes. Parcel Shipping Index 2026. https://www.pitneybowes.com/us/shipping-index.html 11. Loqate (GBG). "Fixing Failed Deliveries," 2021. https://info.loqate.com/hubfs/Loqate%202021/Fixing%20Failed%20Deliveries/Fixing%20Failed%20Deliveries%20-%20Final.pdf 12. Forrester. "US Shoppers Value Post-Purchase Notifications and Delivery Updates" (data snapshot, March 2025 Consumer Pulse Survey), updated 16 September 2025. https://www.forrester.com/report/us-shoppers-value-post-purchase-notifications-and-delivery-updates/RES181131 13. Gartner. "Gartner Survey Finds Majority of Chief Supply Chain Officers Unclear on AI Investment Returns," 5 August 2026. https://www.gartner.com/en/newsroom/press-releases/2026-08-05-gartner-survey-finds-majority-of-chief-supply-chain-officers-unclear-on-ai-investment-returns 14. Gartner. "Gartner Survey Shows Just 23% of Supply Chain Organizations Have a Formal AI Strategy," 11 June 2025. https://www.gartner.com/en/newsroom/2025-06-11-gartner-survey-shows-just-23-percent-of-supply-chain-organizations-have-a-formal-ai-strategy 15. Gartner. "Gartner Predicts 60% of Supply Chain Disruptions Will Be Resolved Without Human Intervention by 2031," 18 March 2026. https://www.gartner.com/en/newsroom/press-releases/2026-03-18-gartner-predicts-60-percent-of-supply-chain-disruptions-will-be-resolved-without-human-intervention-by-2031