General
Last-Mile Delivery for Quick Commerce: Why Sub-30-Minute Promises Break the Density Model in 2026
Key Takeaways
- Quick commerce last-mile delivery is not standard last-mile delivery accelerated. It is last-mile delivery with its main cost lever removed.
- McKinsey finds raising drops per stop from one to five cuts labor and vehicle costs by more than half. A ten-minute promise structurally forbids that batching.
- Once density is unavailable, the economics move to four levers: proximity, rider utilization, pre-positioning against predicted demand, and assignment quality.
- Idle time becomes the dominant cost line rather than travel time. In sub-30-minute models a rider’s shift is mostly waiting, and utilization is the number that decides profitability.
- Batching does not disappear entirely, it becomes a promise trade. Pairing two orders extends one customer’s wait, so it is a commercial decision rather than a routing one.
- Locus reasons across 250+ real-world constraints, allocating and re-allocating at the cadence sub-hour promises require.
The direct answer
Quick commerce last-mile delivery fails when operators treat it as standard last-mile delivery run faster. It is a different problem, because the promise window removes the lever that makes conventional last mile affordable.
That lever is density. McKinsey puts the last mile at 60% to 70% of total parcel delivery cost and found that raising the number of parcels dropped per stop from one to five cuts labor and vehicle cost by more than 50%. Every mature last-mile operation is built to accumulate orders until a route is dense enough to be economic.
A ten-minute promise forbids accumulation. There is no window in which to wait for a second compatible order, so the rider carries one order, travels to one address, and returns. The single most powerful cost lever in last-mile delivery is not merely weakened. It is unavailable by design.
This is why sub-hour operators can improve routing, pay and technology and still lose money per order. They are optimizing around a constraint rather than addressing it. What works instead is recovering economics from four places that do not require a batching window: proximity, rider utilization, pre-positioned capacity and assignment quality.
Locus, the world’s first agentic Transportation Management System, is built for decisions at that cadence. Its Digital Supply Chain Officer (DiSCO) framework reasons across 250+ real-world constraints and has orchestrated more than 1.5 billion deliveries for 360+ enterprise customers across 30+ countries at 99.99% uptime. Locus has been recognized by Gartner for seven consecutive years, including the 2026 Gartner Hype Cycle for Supply Chain Execution and Logistics Technologies, is a Leader in Transportation Management Systems in the QKS Group SPARK Matrix, and ranked #1 in Route Planning on G2’s 2026 Best Software Awards.
The four promise regimes, and what changes at each
Promise width is not a service setting. It determines which operating model is even available.
| Promise window | Batching available | Dominant cost | What the operator is really optimizing |
|---|---|---|---|
| Next day or same day | Extensive, hours of accumulation | Cost per drop | Route density and stop sequencing |
| Two to four hours | Partial, several orders per trip | Cost per drop, with slot constraints | Slot capacity and batch compatibility |
| Twenty to thirty minutes | Marginal, occasional pairing | Rider hours | Utilization and dispatch cadence |
| Under ten minutes | Effectively none | Idle rider hours and fixed store cost | Proximity and demand prediction |
The jump that breaks models is between rows two and three. Above roughly forty minutes an operator is still running a recognizable last-mile business. Below thirty it is running a staffing and real estate business with a delivery component attached.
Also Read: Delivery Under 2 Hours: How Quick Commerce Leaders Can Scale Fulfillment
Why quick commerce last-mile delivery is a proximity problem first
If travel time must fit inside ten minutes including handover, the serviceable radius is small, usually a few kilometers in dense urban geography and less where congestion is severe. Everything else follows from that.
It means coverage is bought with fixed cost rather than earned with routing. Each dark store covers a small catchment, so growth means more stores rather than longer routes, and every store carries rent, staffing and inventory whether or not demand materializes. It also means the store network, not the algorithm, sets the ceiling on what the last mile can achieve. No dispatch system recovers a badly placed catchment.
Congestion decides how much radius a promise actually buys, and it varies enormously by market. INRIX found US drivers lost 49 hours to congestion in 2025, but Chicago at 112 hours and New York at 102 against a national average of 49. The same ten-minute promise describes a materially smaller service area in Chicago than in a mid-size metro, which is why catchment radius has to be set per market and per time of day rather than as a network standard.
The instructive part of the operating record is where profitability actually came from. Reporting on the category notes that Blinkit’s route to EBITDA profitability came from density and average order value optimization at the cluster level rather than from opening stores faster, while Swiggy’s Instamart took nine years to break even. Cluster-level density, not store count, was the variable.
Rider utilization is the metric that decides profitability
In conventional last-mile delivery, the planner’s enemy is travel time. In sub-30-minute delivery it is idle time, and this inversion is the single most common thing operators get wrong.
A rider on a ten-minute promise cannot be dispatched in advance, cannot carry a full load, and cannot be usefully rerouted mid-trip. They wait, take one order, deliver it, and return to wait again. Across a shift, waiting is often the largest single block of paid time. That makes utilization the profitability variable, and it is not improved by better routing because there is no route to improve.
What does move it is demand shape management. Quick commerce demand is sharply peaked around meal times, weather events and weekends, and a fleet sized for peak is badly underused off-peak while a fleet sized for average fails at peak. The practical levers are staggered shift design against forecast demand, cross-utilization of riders across nearby catchments so a quiet store’s capacity serves a busy one, and flexible capacity that can be added in short increments.
Incentive design matters more here than in any other last-mile model, because rider behavior under time pressure is itself a variable. Peer-reviewed work on quick commerce operations examines the interplay of delivery fee, penalty structure and rider-platform collaborative effort, which is the formal version of a point operators learn expensively: a pay and penalty structure that rewards speed alone will produce speed at the cost of safety, accuracy and retention.
Also Read: Hyperlocal Fulfillment: Engineering Profitable 2-Hour Delivery
Batching does not vanish, it becomes a promise trade
The interesting nuance is that batching is not strictly impossible below thirty minutes. It is possible whenever two orders are close in both space and time. What changes is who pays for it.
Pairing two orders means one customer waits longer so the other can be served on the same trip. In a two-hour model that is invisible, because both orders still land inside their windows. In a fifteen-minute model it is visible immediately, and it consumes promise headroom that belongs to a specific customer.
That makes selective batching a commercial decision rather than a routing one, and it needs to be governed as such. Three things have to be defined before a system is allowed to pair orders: how much of the promise may be spent on a co-delivery, which customers or order types are never eligible, and what the operation does when a pairing starts to run late. Systems that batch purely on proximity will quietly convert a delivery promise into a cost saving without anyone deciding that trade was acceptable.
The four levers that replace density
| Lever | What it controls | The decision it requires |
|---|---|---|
| Proximity | Serviceable radius inside the promise | Where catchments sit and how many, set per market against congestion |
| Rider utilization | Paid idle share, the largest cost block | Shift design against forecast demand, plus cross-catchment sharing |
| Pre-positioning | Whether peak is met or cancelled | Capacity committed ahead of predicted demand, not after it arrives |
| Assignment quality | Promise adherence and trips per hour | Allocation on live position, catchment and remaining promise time |
None of these is a routing improvement, which is the point. In a model without routes, the four decisions above are where the margin is.
Also Read: 10 Best Hyperlocal Delivery Management Software (2026 Guide)
What to measure in quick commerce last-mile delivery
Conventional last-mile metrics mislead here, because most of them assume a route.
| Metric | Why it matters below thirty minutes |
|---|---|
| Orders per rider hour | The real productivity measure once stops per route is meaningless |
| Paid idle share | Share of paid rider time spent waiting for an order, usually the largest single cost block |
| Promise adherence by percentile | Median performance hides the tail, and the tail is what customers remember |
| Catchment coverage at promise | Share of the catchment actually reachable inside the promise, by time of day |
| Store-level contribution | Delivery and store cost against order value per store, since network averages hide loss-making catchments |
| Pairing rate and cost | How often orders are co-delivered, and how much promise time that consumed |
Store-level contribution is the one most often missing. A network can look viable in aggregate while a third of its catchments lose money on every order, and that is invisible until it is computed per store.
Also Read: Grocery Delivery Management System: What Enterprises Need
Where the model genuinely does not work
Any honest assessment of quick commerce last-mile delivery has to state the boundaries, because ignoring them is how operators lose money at scale.
Low-density catchments. Below a demand threshold the fixed cost of a store cannot be spread across enough orders, and no operational improvement fixes it. The answer is a wider promise, not better dispatch.
Large baskets. The economics assume small, frequent orders that a single rider carries on a two-wheeler or on foot. Once basket size requires a larger vehicle, the promise and the cost structure both break.
Sustained peak. Fleets can be flexed for spikes but not for permanently elevated demand at peak-only staffing levels, and the failure mode shows up as cancellation rather than lateness.
Categories where accuracy beats speed. For pharmacy, alcohol and age-restricted goods, verification time is not compressible and should not be treated as latency to remove.
It is also worth treating category forecasts with caution. Published estimates of quick commerce market size for 2026 vary by more than a factor of three depending on definitions and regional scope, so they are useful for direction and unusable for planning.
Also Read: Route Optimization Software vs Last-Mile Platform 2026
How Locus helps quick commerce operations decide at speed
Locus, the world’s first Decision-Intelligent, Agentic TMS, is relevant to this model for a specific reason: when there is no route to optimize, value shifts to the quality and cadence of allocation decisions, and that is what the DiSCO framework is built to do. It runs a continuous Sense-Decide-Execute-Learn cycle across eight specialized agents, reasoning over 250+ real-world constraints under six governance mechanisms.
The Dispatch Agent assigns orders against live rider availability, current position, catchment boundaries and remaining promise time rather than proximity alone, and re-decides continuously as conditions change. DispatchIQ applies that logic across hundreds of concurrent constraints, which is what allows selective pairing to be evaluated against the promise rather than against distance. The Capacity Agent matches rider supply to forecast demand shape, which is the direct lever on paid idle share, and supports cross-catchment utilization when one store is quiet and a neighbor is peaking. The Customer Agent owns the promise itself, so adherence is measured against what the customer was shown at order time. Configurable autonomy levels matter at this cadence, since assignment decisions are far too frequent to review individually while eligibility rules and promise trades stay under human control.
A grocery brand delivering fresh and perishable orders across more than 30 cities ran the harder version of this problem, since a missed window on fresh product means spoilage rather than a late parcel, and its network ran on contracted third-party operators rather than an owned fleet. With Locus orchestrating allocation and execution, the operation delivered 33% faster deliveries and 15% lower fulfillment cost, with manual shipping time down 25% and customer support resolution 10 to 20 times faster.
A global FMCG operation running across 10 Asian countries with 1,000+ distributors and 5,000+ riders faced the demand-shape problem at scale. On Locus it reached 3X ROI with more than 12,000 trips saved per month across $4B+ in optimized orders, reaching over 1.8M retail outlets. The trips saved figure is the relevant one, because saved trips in a high-frequency model are recovered rider hours.
In October 2025, Ingka Investments, the investment arm of Ingka Group, the world’s largest IKEA retailer, acquired Locus. Locus continues to operate independently.
Request a Locus quick commerce dispatch assessment to find which of your catchments are losing money per order.
Frequently Asked Questions (FAQs)
What is quick commerce last-mile delivery?
Quick commerce last-mile delivery is the final-leg fulfillment of small, high-frequency orders promised in roughly ten to thirty minutes, served from dark stores or micro-fulfillment centers placed inside dense urban catchments. It differs from standard last-mile delivery in that the promise window is too short to accumulate orders into a dense route, so a rider typically carries one order per trip and the economics depend on proximity, rider utilization and demand prediction rather than on route density.
Why is quick commerce last-mile delivery so expensive?
Because the promise removes batching. McKinsey finds that raising drops per stop from one to five cuts labor and vehicle cost by more than half, and a ten-minute window makes that accumulation impossible. On top of that, coverage has to be bought with fixed cost, since each dark store serves only a small radius, and riders spend a large share of paid time idle waiting for orders. Cost per order is therefore driven by store fixed cost and paid idle hours rather than by miles driven.
How do you improve last-mile delivery efficiency in quick commerce?
Work the four levers that do not need a batching window. Place and size catchments against actual demand density rather than uniformly. Manage rider utilization through staggered shifts against forecast demand and cross-catchment sharing so quiet stores lend capacity to busy ones. Pre-position capacity ahead of predictable peaks instead of reacting to them. And improve assignment quality, which means allocating on live rider position, catchment boundary and remaining promise time rather than on proximity alone.
Can quick commerce orders be batched at all?
Yes, but selectively, and it should be treated as a commercial decision rather than a routing optimization. Two orders can be co-delivered when they are close in both space and time, which means one customer’s promise time is spent serving another. Define in advance how much promise headroom may be used for a pairing, which order types are never eligible, and what happens when a pairing runs late. Systems that pair on proximity alone convert a customer promise into a cost saving without anyone approving that trade.
What metrics matter most for quick commerce delivery?
Orders per rider hour rather than stops per route, paid idle share as the largest usually-unmeasured cost block, promise adherence reported by percentile rather than as a median, catchment coverage achievable inside the promise by time of day, and store-level contribution. That last one matters most, because a network can look viable in aggregate while a substantial share of catchments lose money on every order, and only per-store computation reveals it.
When is a sub-30-minute promise the wrong model?
When catchment demand density is too low to spread store fixed cost, when basket sizes require a vehicle larger than a two-wheeler, when peak demand is sustained rather than spiky, and in categories such as pharmacy or age-restricted goods where verification time is not compressible. In each case the correct response is a wider promise window for that catchment or category, not more dispatch technology.
General
Fleet Utilization and the Cost of Idle Time: A CFO Framework for Fresh Produce Peak Shipping
Sep 1, 2026
Idle and deadhead behave differently during a harvest surge than in steady state. How to model the cost, why fleet utilization is the only lever available at peak, and what to measure.
Read more
General
How to Optimize Grocery Delivery Routes: Strategies and Software for Same-Day and On-Demand Operations in 2026
Sep 2, 2026
Grocery route optimization is structurally harder than e-commerce routing. The six constraints that break standard routing logic, what software must do about each, and how to evaluate platforms.
Read moreInsights Worth Your Time
Last-Mile Delivery for Quick Commerce: Why Sub-30-Minute Promises Break the Density Model in 2026