Ingka Group acquires Locus! Built for the real world, backed for the long run. Read here>Read the full story>
Ingka Group acquires Locus! Built for the real world, backed for the long run. Read the full story

5 Silent Signs Your TMS Is Bleeding Retail Margin Without Showing It on the Dashboard

A Saturday in mid-November. Your senior planner has been rerouting manually since 9am. A regional carrier missed its cut. A returns truck arrived at the wrong DC. Three stores swung to BOPIS-heavy in the same two-hour window.

The dashboard was green throughout.

Most retail logistics leaders live this scene monthly. The reflex is to blame carriers or peak chaos, but the actual cause is structural. The system is executing its rules correctly; those rules describe a problem shape that no longer exists.

Five signs we keep seeing across retail logistics in 2026.

Sign 01

The flattening peak

Fig. 01 Online retail spend, Nov 1 – Dec 31

Seven years ago, peak was a four-day spike. In 2025, twenty-five separate days cleared $4B.

Calendar heatmap, 2018 vs 2025 holiday spend Two horizontal calendar rows. 2018 shows four hot cells at Black Friday, Cyber Monday and Free Shipping Day. 2025 shows twenty-five warm-to-hot cells spread across both months. NOV DEC 2018 2025 4 hot days 25 hot days $4B+/day · 25 days peak isn’t a weekend anymore VOLUME cold → $4B+ day

Adobe Analytics, 2026

What it looks like. Finance flags Q4 freight variance bigger than expected. CX is overwhelmed with WISMO calls. Daily reporting shows missed delivery windows climbing through November. Planners sleep less.

What we tell ourselves. “It’s peak. Everyone hurts during peak. Carriers over-promised again.”

What it actually is. Peak isn’t a weekend anymore. Adobe’s 2026 holiday recap recorded $4B+ in daily online spend on 25 separate days last season — up from 15 days the year before. Cyber Week alone was just 17.2% of the $257.8B holiday total. Volume is now plateau, not spike. Carrier caps were negotiated against spikes. The system reaches cap on what should be a routine Tuesday and defaults overflow to a national carrier at peak surcharges.

Sign 02

The split-shipment hemorrhage

Fig. 02 $100 omnichannel basket, split across two nodes

Dashboards see one shipment. Margins see three.

Basket P&L waterfall A horizontal waterfall starting at $100 revenue, deducting picking labor twice (the second leg highlighted as duplicated), packing twice, two carrier costs, leaving a net loss of $13. $100 $75 $50 $25 $0 +$100 REVENUE −$8 PICK node 1 −$3 PACK node 1 −$8 CARRIER node 1 −$8 PICK node 2 · split −$3 PACK node 2 · split −$8 CARRIER node 2 · split −$13 NET DUPLICATED LEG · −$19

McKinsey, 2024 — grocery economics, applied to omnichannel split

What it looks like. Top-line omnichannel revenue is healthy. Total fulfillment cost as a percentage of revenue is climbing. Store managers complain about associates pulled off the floor to pick digital orders. Finance flags outbound shipping expense that does not track with order volume.

What we tell ourselves. “Omnichannel is just expensive. The e-commerce tax is the price of doing business.”

What it actually is. Some cost is real: McKinsey’s grocery unit economics show ~$8 of picking labor and ~$8 of last-mile delivery on a $100 basket. Without a delivery fee, that lands at −$13 per basket. The omnichannel parallel is sharper. When a 3-item order splits across nodes, each leg duplicates the picking, packaging, label, and carrier line. The dashboard reports a single shipment status. The P&L absorbs every duplicated line.

Learn more about the Agentic TMS purpose built for omnichannel retailers

Sign 03

Tight delivery window drift

Fig. 03 Minutes ahead/behind promised window, by stop number

Failed windows aren’t random. They compound.

Drift compared: static batching vs. dynamic re-routing Two lines plotted across stops 1 to 12. Static batching curves sharply downward past stop 5, missing windows from stop 7 onward. Dynamic re-routing oscillates within a narrow band around the on-time axis. FAILED-WINDOW THRESHOLD +20 0 −60 MINUTES VS. WINDOW STOP NUMBER 1 2 3 4 5 6 7 8 9 10 11 12 Stop 7: first failed window Stop 10: 3 windows missed · $51.60 Static batching Dynamic re-routing

Capgemini, 2025

What it looks like. Finance asks why on-time-in-window has slipped quarter-over-quarter. CX explains a third 5pm-instead-of-3pm to the same customer. Inventory writes off another temperature-failed batch. The slip pattern is the same across grocery slots, same-day pickup, and premium SLAs.

What we tell ourselves. “Urban traffic is unpredictable. Some drift is the cost of operating tight delivery windows at scale.”

What it actually is. Tight-window economics are unforgiving. Last-mile delivery now runs 53% of total logistics cost per Capgemini and McKinsey aggregations. Inefficient routing degrades route efficiency by 37% on average; the cost per failed delivery, counting redelivery and service overhead, is $17.20. Static batching against zip-code density cannot recalculate live traffic and dwell time on the fly. Each failed window compounds against the next.

Sign 04

The latency-cost chasm

Fig. 04 Cost to recover, indexed, by hours since deviation

Detection delay, not deviation, drives recovery cost.

Decision-latency cost curve Recovery cost rises exponentially with hours-since-deviation. Three zones: cheap-fix at 0 to 1 hour, spot-premium at 4 to 8 hours, and expedite-plus-write-off at 12 to 24 hours. CHEAP-FIX ZONE SPOT-PREMIUM EXPEDITE + WRITE-OFF 10× COST TO RECOVER (×) 0h 4h 8h 12h 16h 20h 24h HOURS SINCE DEVIATION TYPICAL TODAY · 8h manual exception queue AGENTIC DETECTION −28% response time −19% recovery cycle

Gartner Resilience Benchmark, 2025

What it looks like. Walk into the control tower. Your most experienced planner has a dozen browser tabs open, cross-referencing carrier portals against a spreadsheet, manually tracking delayed page-containers. Constant triage. Junior planners cannot cover.

What we tell ourselves. “Logistics is unpredictable. Our seasoned planners know how to hustle.”

What it actually is. The hero complex is genuine, and largely a symptom of architectural latency. By the time a planner discovers a deviation, cheap recovery options are gone. Gartner finds 14% of supply chain leaders still rely heavily on technology-assisted manual processes; the rest sit on mainstream platforms that funnel exceptions through human queues. A 2025 Gartner Resilience Benchmark finds AI-detected deviations reach response 28% faster, with recovery cycles shortened by 19% versus dashboard-reliant exception management. Gartner expects 60% of disruptions to resolve without human intervention by 2031. The inverse holds today: the planner is the system’s safety net.

You’ve got equal parts antiquated, legacy capabilities, automation software, trying to harmonize with modern, sophisticated automation, and you’re trying to harmonize the integration of those two worlds.

Sean Barbour · Senior Vice President of Supply Chain, Macy’s · via NRF

Sign 05

The hemorrhage of returns routing

Fig. 05 % of original retail value retained, by days since return

Every day a returned item is in transit, its margin shrinks.

Resale value retained: dynamic vs. static returns routing Two curves over 45 days. Dynamic routing holds 90 to 95 percent retained value through day 15 and gradually declines. Static routing drops sharply around day 21 and reaches a 40 to 50 percent floor after a markdown around day 35. 100% 75% 50% 25% 0% VALUE RETAINED 0 9 18 27 36 45 DAYS SINCE RETURN Day 7: dynamic restock Day 21: static arrives at DC DAY 35 · MARKDOWN 40% 70% fails to resell at full value 70% 42% Dynamic returns routing Static (DC consolidation → markdown)

Radial, 2024 / NRF + Happy Returns, 2024–2025

What it looks like. Finance flags inbound parcel costs overrunning budget every quarter. Customer service explains another delayed refund. Merchandising watches a returned-but-stranded SKU miss its selling window. The CFO calls reverse logistics the black hole of the P&L.

What we tell ourselves. “Returns are a cost center. The flow is unpredictable.”

What it actually is. The scale tells the story. NRF and Happy Returns put 2024 retail returns at $890B, or 16.9% of total sales. Online return rates project to 19.3% in 2025. Processing a $100 returned item costs about $26.50 on average across Optoro and OpenSend’s 2024–25 aggregations. The terminal cost is worse: Radial’s 2024 returns analysis finds only 30% of returned merchandise is resold at full value. Most of that destruction is locked in by the routing decision. A static system that defaults returns to the primary DC creates a multi-week lag before items become resaleable. By the time they are, the sell-through window has closed.

Notes & sources

Notes

Not every TMS deployment is an architectural failure. Highly predictable, full-truckload B2B lanes between owned distribution centers run on rules-based logic with adequate cost performance. The architectural failure surfaces specifically where retail volatility, fragmentation, and speed demands collide.

The transition is also imperfect. Gartner finds 56% of chief supply chain officers cite legacy-system integration as a major obstacle to scaling AI in their supply chains. Layering an agentic wrapper on a fragmented, siloed data architecture compounds the problem rather than fixing it. The architectural shift is real. So is the cost of doing it badly.

Nachiket Murthy

Nachiket Murthy

Product Marketing Manager

Nachiket leads Product Marketing at Locus, bringing over seven years of experience across financial analysis, corporate strategy, governance, and investor relations. With a multidisciplinary lens and strong analytical rigor, he shapes sharp narratives that connect business priorities with market perspectives.