Last-mile and e-commerce logistics

In short

Last-mile logistics is the final movement of goods from the nearest stocking point — fulfilment centre, dark store or delivery hub — to the customer's door, and it is the costliest, slowest and most failure-prone leg of the whole chain, typically 30-50% of total shipment cost. E-commerce logistics is the end-to-end system built around it: order management, pick-pack, line-haul, sortation, delivery, cash-on-delivery collection and returns. It matters because the customer judges the entire brand on this one leg, and because India's parcel volumes, COD habit and dense pin-code geography make it unusually hard. It is examined in Warehousing and Transport, Supply Chain Operations and Service Operations papers, and is the core interview subject for Flipkart, Amazon India, Delhivery, Blue Dart, Ecom Express, XpressBees, DTDC, Blinkit, Zepto, Swiggy Instamart, BigBasket, Shadowfax and Porter.

Explained

The e-commerce fulfilment chain, stage by stage

• Order capture: the marketplace or D2C site receives the order; an OMS (order management system) validates payment mode, address, serviceability and inventory, then allots it to a node.
• Node allotment / order splitting: the OMS picks the warehouse nearest to the delivery pin code with stock; a multi-item order may split into several shipments from different nodes.
• Pick, pack, manifest: items are picked (batch, zone or wave picking), packed to the carrier's size and weight rules, labelled with an AWB (air waybill) number, and manifested to a courier.
• First mile: pickup from the seller or fulfilment centre to the origin hub; for marketplace sellers this is a scheduled courier pickup.
• Line haul / mid mile: hub-to-hub movement by road (mostly), air (for far or high-value shipments) or rail, with sortation at a mother hub and regional sort centres.
• Last mile: the delivery centre (DC) sorts shipments by beat, loads riders, and attempts delivery.
• Returns / reverse flow: refused, undelivered (RTO — return to origin) or customer-returned shipments travel back, are quality-checked and restocked, refurbished, liquidated or scrapped.
• COD reconciliation: cash collected by the rider is deposited and remitted to the seller, usually on a weekly cycle; this is a treasury and fraud-control process, not just a logistics one.

Last-mile delivery models

• Hub-and-spoke courier model: national line haul into DCs, riders on fixed beats. Low cost per shipment at scale, but delivery promise is 2-5 days.
• Dark store / micro-fulfilment (quick commerce): small 2,000-4,000 sq ft stores holding 2,000-6,000 SKUs inside a 2-3 km catchment, enabling 10-30 minute delivery. Cost per order is high; it works only at very high order density.
• Hyperlocal / on-demand: the store or restaurant is the stocking point; the platform supplies only the rider (Swiggy, Zomato, Dunzo-type models).
• Click-and-collect / BOPIS: customer collects from a store or kirana; removes the delivery attempt cost entirely.
• Parcel lockers and pickup points (PUDO): shipments aggregated at one secure point; high drop density, no failed attempts, but needs customer willingness.
• Crowdsourced / gig fleet: independent riders paid per delivery; flexes with demand, weaker on training and shrinkage control.
• Seller-managed / self-ship: the seller uses its own fleet or a chosen courier; the marketplace only hosts the order.
• Drop shipping: the seller never holds stock; the supplier ships directly to the customer. Low capital, poor control over service and returns.

Why the last mile is expensive — the cost drivers

• Drop density: the number of successful deliveries per km of route. Cost per shipment falls almost hyperbolically as density rises; rural routes with 8-10 drops cost many times an urban beat with 60-80.
• First Attempt Delivery Rate (FADR): every failed attempt roughly doubles the delivery cost for that shipment. Causes: customer unavailable, wrong or incomplete address, no cash for COD, phone unreachable.
• COD burden: cash handling, remittance float, and a far higher refusal/RTO rate than prepaid. Prepaid share has risen sharply with UPI, which is the single biggest structural cost improvement in Indian e-commerce logistics.
• Returns: apparel and footwear can see very high return rates (frequently quoted in the 20-40% band for fashion; it varies by seller and season), and a return costs both legs plus inspection.
• Volumetric weight: carriers charge on the greater of actual and dimensional weight, so loose packing is paid for twice.
• Address quality: unstructured Indian addresses force rider phone calls; geocoding and digital addressing (DIGIPIN, the India Post grid-based digital address code) aim to fix this.
• Urban constraints: no-entry hours for goods vehicles, gated societies, high-rise vertical travel time, parking, monsoon.
• Rider attrition and fuel: per-delivery payouts, high churn, and fuel or electricity cost per km.

Route planning and the delivery beat

• Beat / route design: a DC's pin codes are cut into beats of roughly equal workload; each rider gets a sequenced manifest.
• Vehicle Routing Problem (VRP): choose routes for a fleet from a depot to serve all stops at minimum cost. Variants: CVRP (vehicle capacity), VRPTW (time windows), VRPPD (pickup and delivery — essential in India because riders collect returns on the same beat).
• Travelling Salesman Problem (TSP) is the single-vehicle case; both are NP-hard, so practice uses heuristics.
• Heuristics taught and used: nearest neighbour, Clarke-Wright savings method, sweep method, then local improvement (2-opt, or-opt); commercial engines add live traffic and historical service time.
• Dynamic routing in quick commerce: orders are assigned after the rider has left, batching 2-3 nearby orders per trip; the dispatch engine optimises rider-order matching every few seconds.
• Time windows and slots: narrow slots raise customer satisfaction and cut failed attempts but reduce route efficiency — a direct trade-off examined often.

Technology and visibility

• Last-mile delivery apps: rider app with scan-on-load, geofenced delivery confirmation, OTP-based delivery (standard for high-value and all COD in many networks), electronic proof of delivery (ePOD), digital payment at door.
• Serviceability engine: maps pin code to service type (prepaid only, COD allowed, heavy allowed, next-day) — the gate for what the website even shows as deliverable.
• Control tower and NDR management: NDR (non-delivery report) handling calls the customer, reattempts, reschedules or triggers RTO; good NDR desks are worth several percentage points of delivery success.
• Track and trace: scan events at pickup, hub-in, hub-out, DC-in, out-for-delivery, delivered; the basis of all SLA measurement.
• Automation: cross-belt and shoe sorters at sort centres, put-to-light at micro-fulfilment, electric two- and three-wheelers for the beat; drones and sidewalk robots are pilot-stage in India, not mainstream.

Reverse logistics and returns

• Return types: RTO (never delivered), customer return within the return window, replacement/exchange, warranty/repair, and end-of-life/e-waste.
• Process: return request → pickup scheduled with QC checklist at the door → reverse line haul → returns processing centre → grade A restock / grade B refurbish or outlet / liquidate / scrap.
• Controls: doorstep QC (IMEI and serial checks, tamper seals, open-box delivery) to stop return fraud; try-and-buy shifts the decision to the doorstep.
• Cost view: a return destroys the forward margin and adds the reverse cost, so reducing returns (better size charts, images, descriptions) is cheaper than optimising the reverse flow.
• E-waste and packaging rules: returns of electronics interact with the E-Waste (Management) Rules, 2022 and plastic packaging obligations under the Plastic Waste Management Rules with EPR (extended producer responsibility) — brands must report plastic packaging recycling targets, which is why paper mailers and rightsized boxes spread.

Indian regulatory and commercial framework

• GST and e-commerce: operators collect TCS (tax collected at source — collected, not deducted) on the net value of taxable supplies made through the platform and file GSTR-8; GST slabs since 22 September 2025 are 5%, 18% and a 40% demerit rate. Courier/logistics service is itself a taxable supply.
• E-way bill: required under GST rules for movement of goods above the threshold value (₹50,000 for most inter-state movement, with state-specific intra-state thresholds); last-mile parcels are usually covered by a consolidated document for the line haul, and rules differ for e-commerce operators, so check the current CBIC notification rather than assuming.
• Consumer Protection (E-Commerce) Rules, 2020 under the Consumer Protection Act, 2019: mandatory seller details, country of origin, no unfair cancellation charges, grievance officer and defined response timelines.
• Motor Vehicles Act, 1988 (as amended 2019) and state rules: commercial goods permits, no-entry timings, and the legal grey area of white-plate two-wheelers used for delivery — several states require commercial registration.
• Labour Codes from 21 November 2025 (Code on Wages, Industrial Relations, Social Security, OSH&WC) replaced the Factories, Minimum Wages, Industrial Disputes and BOCW Acts; the Social Security Code brings gig and platform workers into aggregator-funded social security — directly relevant to rider cost. Implementation of the aggregator contribution is being phased in, so quote the principle, not a rate.
• National Logistics Policy, 2022 and PM GatiShakti; ONDC opens last-mile and fulfilment as unbundled network services; FASTag and e-commerce drone BVLOS rules sit with DGCA.
• Legal metrology (pre-packaged commodity declarations) must appear on the e-commerce listing, not only on the pack.

Key points

• Last mile - the final leg from the nearest node to the customer; 30-50% of shipment cost and the leg the customer actually experiences.
• Drop density - deliveries per km or per beat; the single strongest lever on cost per shipment.
• FADR (first attempt delivery rate) - share delivered on the first visit; each failure roughly doubles that shipment's delivery cost.
• NDR - non-delivery report; the workflow of calling, rescheduling or returning an undelivered shipment.
• RTO - return to origin; an undelivered shipment travelling back, costing both legs with zero revenue.
• COD - cash on delivery; higher refusal, cash handling and remittance float, falling as UPI prepaid share rises.
• Volumetric (dimensional) weight - chargeable weight = max(actual, L×B×H ÷ divisor); loose packing is paid for twice.
• Dark store - small urban micro-fulfilment node for 10-30 minute quick commerce in a 2-3 km catchment.
• Hub-and-spoke - line haul into sort centres and delivery centres; cheap per parcel, slower promise.
• PUDO / parcel locker - aggregated collection point; zero failed attempts, high drop density.
• VRP / TSP - NP-hard routing problems; solved in practice by savings, sweep and 2-opt heuristics plus live traffic.
• VRPTW and VRPPD - time-window and simultaneous pickup-delivery variants; India needs VRPPD because returns ride the same beat.
• Reverse logistics - the return flow and its grading into restock, refurbish, liquidate or scrap.
• Serviceability engine - pin-code-level rules deciding COD, heavy, and delivery promise shown on the site.
• Order splitting - one customer order becoming several shipments from different nodes; raises cost, hurts experience.
• Cost per shipment (CPS) - the industry's headline metric; last-mile CPS = total DC cost ÷ delivered shipments.
• SLA / promise date - the committed delivery date; breach drives penalties on 3PL contracts.
• EPR under Plastic Waste Management Rules - packaging recycling obligation that pushed rightsizing and paper mailers.

Formulas and examples

Example 1 — Cost per shipment and the effect of failed attempts A Pune delivery centre handles 4,000 shipments a day. Monthly cost (30 days): rider payouts ₹18,00,000, DC rent and utilities ₹3,60,000, line-haul allocation ₹10,80,000, fixed staff ₹2,40,000 — total ₹34,80,000. Of 1,20,000 shipments dispatched in the month, 1,08,000 were delivered and 12,000 went RTO.

• CPS on delivered shipments = 34,80,000 ÷ 1,08,000 = ₹32.22
• CPS if measured on dispatched = 34,80,000 ÷ 1,20,000 = ₹29.00 (why the denominator must be stated)
• RTO rate = 12,000 ÷ 1,20,000 × 100 = 10%
• If NDR calling lifts delivery to 1,14,000 with the same cost, CPS = 34,80,000 ÷ 1,14,000 = ₹30.53, a saving of ₹1.69 per shipment (5.2%) with no new spend.

Example 2 — Rider productivity and fleet size A rider works 480 minutes, spends 60 minutes on loading and reconciliation, so 420 productive minutes. Average travel 4 minutes between stops, service 3.5 minutes per stop.

• Time per stop = 4 + 3.5 = 7.5 min → stops per rider = 420 ÷ 7.5 = 56 stops/day
• At FADR 85%, successful deliveries per rider = 56 × 0.85 = 47.6 ≈ 48
• For 4,000 deliveries a day: riders = 4,000 ÷ 47.6 = 84.03 → 85 riders (fleet size always rounds up)
• Raise FADR to 92% (OTP + slot confirmation): deliveries per rider = 56 × 0.92 = 51.5 → riders = 4,000 ÷ 51.5 = 77.6 → 78 riders. Saving 7 riders; at ₹18,000 per rider per month that is ₹1,26,000 a month.

Example 3 — Chargeable weight and packing A 1.1 kg kurta set is shipped in a 40 × 30 × 25 cm box.

• Volumetric weight = (40 × 30 × 25) ÷ 5,000 = 30,000 ÷ 5,000 = 6 kg
• Chargeable weight = max(1.1, 6) = 6 kg. At ₹40 per kg, freight = ₹240.
• Rightsize to a 35 × 25 × 5 cm poly mailer: volumetric = 4,375 ÷ 5,000 = 0.875 kg → chargeable = max(1.1, 0.875) = 1.1 kg, rounded to the usual 0.5 kg slab = 1.5 kg → freight = ₹60. Saving ₹180 per shipment, i.e. 75%.

Example 4 — Clarke-Wright savings, first merge Depot 0, stops A, B, C. Distances (km): 0-A 12, 0-B 10, 0-C 14, A-B 5, B-C 7, A-C 15.

• S(A,B) = 12 + 10 − 5 = 17; S(B,C) = 10 + 14 − 7 = 17; S(A,C) = 12 + 14 − 15 = 11
• Separate out-and-back routes total = 2(12 + 10 + 14) = 72 km
• Merge on the best saving, say A-B, then extend with B-C → route 0-A-B-C-0 = 12 + 5 + 7 + 14 = 38 km
• Distance saved = 72 − 38 = 34 km (= 17 + 17, the two savings used). Drop density rises from 3 ÷ 72 = 0.042 to 3 ÷ 38 = 0.079 deliveries/km.

Example 5 — Does a return destroy the margin? A ₹1,200 shoe: product cost ₹700, forward freight ₹70, packing ₹15, payment gateway 2% = ₹24.

• Delivered margin = 1,200 − 700 − 70 − 15 − 24 = ₹391
• A return costs reverse pickup ₹60, reverse line haul ₹50, QC and restock ₹25, plus the forward ₹70 + packing ₹15 already spent = ₹220 lost per return
• At a 30% return rate on 100 units: 70 delivered × 391 = ₹27,370; 30 returns × (−220) = −₹6,600 → net ₹20,770, i.e. ₹207.70 per unit shipped against ₹391 if nothing returned — a 47% margin erosion.
• Cutting returns to 20%: 80 × 391 − 20 × 220 = 31,280 − 4,400 = ₹26,880 → ₹268.80 per unit shipped.

Where it is used

• E-commerce marketplaces and D2C brands: Flipkart Ekart, Amazon Transportation Services, Meesho, Myntra, Nykaa — node allotment, serviceability, NDR desks and returns processing centres.
• 3PL and express courier: Delhivery, Blue Dart, Ecom Express, XpressBees, DTDC, Shadowfax, India Post (the widest rural last mile in the country, often the only serviceable option for remote pin codes).
• Quick commerce and hyperlocal: Blinkit, Zepto, Swiggy Instamart, BigBasket — dark store siting, dynamic rider dispatch, batching.
• Pharma and cold-chain last mile: PharmEasy, Apollo 24|7, lab sample pickup; temperature control on the beat.
• Job roles: last-mile operations executive, hub/DC in-charge, network planner, route and capacity planner, NDR/customer-experience analyst, reverse-logistics executive, rider-fleet and gig-supply manager, logistics business analyst.
• Beyond retail: food delivery, LPG and water distribution, newspaper and milk beats (the oldest high-density Indian last mile), e-governance document delivery and ONDC-based unbundled fulfilment.

Interview questions

• Q: Why is the last mile the most expensive leg? A: It carries small drop sizes, low drop density, high labour content per shipment and failed attempts, and none of the economies of scale that line haul enjoys.
• Q: What is FADR and how would you improve it? A: First attempt delivery rate — the share delivered on the first visit. Improve with address verification and geocoding, pre-delivery SMS or call with a slot, OTP confirmation, prepaid nudges, and beat sequencing that matches customer availability.
• Q: How does COD hurt logistics economics? A: It raises refusal and RTO, adds cash handling, security and remittance float, slows each stop, and blocks the seller's working capital until remittance.
• Q: Difference between RTO and a customer return? A: RTO never reached the customer (refused, unreachable, wrong address); a customer return was delivered and then sent back within the return window. RTO is cheaper to prevent and usually signals an address or COD problem.
• Q: How would you decide where to open a dark store? A: Map demand by order heat map, keep every point inside a 2-3 km rideable radius, check road connectivity and no-entry rules, rent per sq ft against expected orders per day, and the break-even order density where contribution covers store fixed cost.
• Q: What is volumetric weight and why does it matter? A: Chargeable weight is the greater of actual and (L×B×H ÷ divisor, commonly 5,000). Oversized packing makes you pay for air, so rightsizing cartons cuts freight directly.
• Q: Why is routing in India not just a TSP solve? A: Addresses are unstructured, traffic and no-entry timings vary by hour, riders also collect returns (pickup and delivery in one beat), orders arrive after dispatch, and service time varies hugely between a gated society and a market lane.
• Q: What trade-off do narrow delivery time slots create? A: They raise FADR and customer satisfaction but constrain the route, cutting stops per rider per day; the gain must exceed the lost productivity.
• Q: Which metrics would you put on a last-mile dashboard? A: CPS, deliveries per rider per day, FADR, attempts per delivery, SLA/on-time percentage, RTO and return rate, NDR resolution rate, cash remittance ageing, and damage or shrinkage per thousand.
• Q: How does reverse logistics differ from forward logistics? A: Flows are many-to-one, volumes unpredictable, units arrive unpacked and ungraded, value is uncertain until inspection, and the process needs QC and disposition rather than straight put-away.
• Q: Does quick commerce beat normal e-commerce on cost? A: No — cost per order is far higher because of dark store fixed cost and single-order trips; it wins on convenience and basket frequency, and only works where order density is very high.
• Q: Which Indian rules would you check before launching a delivery fleet? A: Motor Vehicles Act and state commercial-registration rules for the vehicles, the Labour Codes in force from 21 November 2025 including gig-worker social security under the Social Security Code, GST and e-way bill requirements, Consumer Protection (E-Commerce) Rules, 2020, and EPR plastic packaging obligations.

Common mistakes

• Quoting cost per shipment without saying whether the denominator is shipments dispatched, attempted or delivered — the three differ by 10% or more.
• Confusing RTO with a customer return, and so prescribing the wrong fix (address and COD controls versus sizing and product information).
• Using actual weight instead of chargeable weight in a freight calculation, or applying the air divisor to a surface tariff.
• Treating the last mile as a pure TSP and ignoring capacity, time windows, return pickups and dynamic order arrival.
• Assuming higher delivery speed always wins — a 10-minute promise is uneconomic outside very dense catchments.
• Forgetting that a return costs the forward leg too, so computing return cost as only the reverse freight.
• Writing "COD is 70% of Indian e-commerce" from an old textbook; the prepaid share has risen sharply with UPI, so give the direction of change rather than a stale percentage.

Related topics

Warehouse management · Transportation modes and freight · Cold chain logistics · Inventory management and EOQ · Supply chain network design and facility location · Logistics costing and performance metrics