Quick commerce unit economics 2026: four dark store models, four very different P&Ls

Quick commerce unit economics 2026: four dark store models, four very different P&Ls

Quick commerce is no longer one playbook. Shipsy platform data across dark-store operators shows contribution-positive unit economics emerge at meaningfully lower peak order volumes for the best-performing models than for the weakest. The differentiator isn’t GMV — it’s pick time, delivery radius, and driver utilization working together.

The finding

Across quick-commerce operators on Shipsy — including one of Asia’s largest quick-commerce arms processing millions of deliveries per day, and other regional players across MENA, SEA, and ANZ — four distinct dark-store models are now visible in the data. Each has different economics, different defensibility, and a different tipping point to profitability. Hub-and-spoke 10-min models, micro-dark-store urban models, attached-to-retail models, and partner-supplied hyper-dense models all work — but only when their cost structure matches their demand curve. The uniform “9–15 min everywhere” model is already out of favor. The winning operators — including Flipkart Minutes, which used Shipsy to cut COD loss by 82% and RTO by 30% — are running different models in different geographies under one routing brain.

Why it’s happening

Three forces reshaped the economics between 2023 and 2026.

1. Customer expectation stratified. Customers in Tier 1 urban cores want 10–15 min; Tier 2 cities accept 20–30 min; suburban expects 45 min. Operators running one speed promise everywhere overspend on courier density in slow markets and underspend in fast ones. Shipsy’s routing layer applies per-geography speed policies.

2. Pick time became the controllable lever. Delivery time is constrained by physics; pick time is constrained by process. Operators who instrumented pick workflows — SKU slotting, pick-path optimization, batch picking — compressed pick time into a fraction of what it used to be. On a 12-minute promise, that efficiency is material.

3. Driver utilization stopped being a gig economy problem. At dense hubs, utilization at peak has climbed materially. The delta is micro-batching — grouping 2–4 orders per rider without violating speed promises. Astra handles this dynamically at sub-minute intervals.

Four distinct models emerge from the data:

• Model A — Hub-and-spoke 10-min. Dense urban, high-SKU dark stores. High fixed cost, high order density. Break-even requires high peak order throughput.
• Model B — Micro-dark-store urban. Smaller footprint, curated SKU. Lower fixed cost. Break-even at moderate peak throughput.
• Model C — Attached-to-retail. Dark-store area inside an existing store. Lowest fixed cost; inventory shared. Break-even at lower peak throughput.
• Model D — Partner-supplied hyper-dense. Dark stores in high-rise residential towers or malls. Very short delivery radius. Higher break-even but at roughly twice the revenue per hour.

What it means for quick commerce operators

The strategic question is model mix, not model choice. The operators winning at national scale run two or three models in parallel, matched to the density and wealth of each geography.

• Model-to-market fit is the core optimization. One city can profitably sustain Model A + Model D; another only Model B + C. Fighting this math is expensive.
• Unit economics are a routing question as much as a real-estate question. Once the store is built, the leverage is pick efficiency and driver utilization. That’s where AI-native orchestration compounds. Plub, for instance, runs Shipsy with 98.5% auto-allocation and has banked $168K/year in savings.
• Returns and cancellations are the silent P&L killer. Keeping cancellation well under 10% is the mark of a well-run QC operation. Beyond that, your SLA promise doesn’t match your operational reality. Flipkart Minutes’ 30% RTO decrease shows what’s possible.
• Middle-mile replenishment is back on the agenda. Stockouts in dark stores cascade into cancellations. A good replenishment routing loop fed by demand forecasting is now table stakes.

Below is the model-by-model view.

What to do about it

Audit the P&L of each dark store individually — most operators are surprised how many underperforming stores sit inside an aggregate that looks healthy. Deploy per-geography speed policies and per-model pick workflows rather than one standard everywhere. Instrument driver utilization per peak hour; it’s the most under-measured metric in QC. And treat replenishment routing as a first-class problem — stockouts are the single biggest leak in unit economics.

For a deeper view on last-mile mechanisms, read our retail omnichannel playbook. See Shipsy for quick commerce and route optimization.