State of Online Shopping in Pakistan 2026: Cash on Delivery, Returns & Buyer Behaviour

The first published study of cash-on-delivery failure and return-to-origin rates in Pakistan, and one of the largest first-party datasets on Pakistani cash-on-delivery behaviour ever published: ≈170,000 orders placed over twelve months and ≈127,000 shipped-parcel outcomes, on returns, geography and the 13× commitment gap.

By OrderNation Research · Published 10 June 2026 · Data windows: June 2025 – May 2026 (orders placed) and July 2025 – June 2026 (shipped-parcel outcomes)

Executive summary

As of June 2026, Pakistan shops online with cash. This study draws on one of the largest first-party datasets on Pakistani cash-on-delivery (COD) behaviour ever published: about 170,000 orders placed over twelve months, and about 127,000 shipped-parcel outcomes. Across the twelve months from June 2025 to May 2026 (n≈170,000 placed orders), 94.5% of orders were placed cash-on-delivery, and fewer than 2 in 100 (1.6%) were prepaid digitally.

That cash habit has a measurable cost. A COD order is more than 13 times more likely to be cancelled than a prepaid one (37.3% vs 2.8%, measured across the same twelve months), and the gap held in every individual month: COD cancellation ran roughly 30–42% while prepaid never exceeded 5.9%, and the monthly ratio never fell below 6×. Of every 100 COD orders placed, roughly 63 ship and only about 49 end with cash collected. Roughly 1 in 5 shipped parcels is returned to origin, with one rupee in every five of shipped order value (19.4%) travelling back over twelve months, a rate that held at 19–22% in every settled month.

This failure is not random. It tracks purchase intent and geography, and it is almost invisible to standard e-commerce analytics. The findings carry direct implications for retailers, couriers, payment providers and the policy push towards a cashless Pakistan.

Cash is still king: COD dominance and what it costs

Pakistan's e-commerce growth story is, underneath, a cash story. Across ≈170,000 orders placed in the twelve months from June 2025 to May 2026 at a single large Pakistani online retailer, 94.5% were placed cash-on-delivery. Digital prepayment (card, wallet or bank transfer) accounted for just 1.6% of orders. A further 4.0% of orders carried no recorded payment gateway and are excluded from the split.

A decade into Pakistan's digital-wallet expansion, fewer than 2 in 100 online orders at this scale are paid for before the rider knocks. That is the defining economic constraint of the sector, not a checkout-design quirk. Every downstream number in this study (cancellations, returns, courier economics, even what "conversion rate" means) flows from a single fact: in Pakistani e-commerce, the sale is not complete when the order is placed. It is complete when cash changes hands at the doorstep.

The cost of that gap is the subject of the rest of this report: orders that cancel before dispatch, parcels that travel hundreds of kilometres and come back, and an analytics blind spot that hides the whole problem from the dashboards the industry steers by.

One honest caveat up front: COD share is partially shaped by each store's own checkout design and incentives, and this dataset is one retailer's. The national prepaid share may differ, but no published national dataset exists at this resolution, which is precisely why we are publishing ours.

The trust gap: returns and RTO as a COD-mechanism phenomenon

When a COD parcel ships, the buyer has committed nothing. That produces a failure mode prepaid markets rarely see: return-to-origin (RTO). The parcel travels to the buyer's city, the delivery fails (a refusal, an unreachable phone, second thoughts), and it travels back.

Measured across ≈127,000 shipped parcels between July 2025 and June 2026, roughly 1 in 5 shipped parcels was returned to origin: roughly one rupee in every five of shipped order value (19.4%) travelled to a buyer's city and came back. Blended delivery success sat at 78.4% (≈99,000 parcels delivered).

The rate barely moves across the year. In every one of the eleven fully settled months from July 2025 through May 2026 (monthly shipped volumes ranged from roughly 9,200 to 13,000 parcels), RTO landed between 19% and 22%, with delivery success at 78–81%. This is the structural, steady-state cost of running a payment mechanism in which the buyer can walk away at the door, not a bad month, a flash sale gone wrong or a courier strike.

Two pieces of context keep that number honest. First, it is not a Pakistani anomaly: published industry estimates for other COD-heavy e-commerce markets such as India typically put return-to-origin (RTO) rates on COD orders in the 20–30% range (Shiprocket, 2025), against an industry-wide average of 20–25% that can spike toward 40% (Shiprocket, 2025; GoKwik, 2026), versus under 2% for prepaid orders per Shipway's ShipNotes FY25 report (Financial Express B2B, 2025); this dataset's ~20% sits at the favourable end of the band. Second, the rate is not passively accepted: like most COD retailers, this store runs a pre-dispatch confirmation-call layer and offers a visible prepaid-payment discount at checkout, the industry's standard defences against doorstep walk-aways. The ~20% wall is what remains after those defences.

Two further findings sharpen the picture:

First, most COD failure happens before the parcel even ships. Across the twelve months from June 2025 to May 2026, the COD cancellation rate was 37.3%, and in the April-2026 deep-dive cohort, roughly three-quarters of those cancellations occurred pre-dispatch, at the order-confirmation stage, with only a quarter occurring after the parcel was already moving. Pre-dispatch cancellation and post-dispatch RTO are different failure events at different funnel stages; this study never sums them naively. When a shipped COD delivery does fail, it takes a median of 21.7 days to surface in the seller's books.

Second, failure tracks purchase intent, not logistics. Across the full twelve months of shipped parcels (July 2025 – June 2026), orders originating from search-intent traffic (a buyer actively looking for the product) were returned to origin 14.2% of the time. Orders originating from social-discovery traffic (a buyer who tapped an ad in a feed) failed at 22.7%, roughly 1.6 times as often (tens of thousands of shipped parcels in each major class). Across traffic-source classes, RTO ranged from 14.2% to roughly 35%, a spread of more than 2× driven by how the buyer arrived rather than which courier carried the box. Delivered rates tell the same story: 84.8% of search-intent shipments arrived versus 75.7% of social-discovery shipments (same class definitions throughout, corroborated by a fully matured 45-day cohort; see methodology).

The "returns problem" in Pakistani e-commerce is not primarily a courier problem. It is a commitment problem, and commitment is set at the moment of payment.

Prepaid changes everything: the 13× commitment gap

Prepayment is where the numbers diverge most sharply, and where the policy case is strongest.

Across the twelve months from June 2025 to May 2026 (n≈170,000 placed orders), cash-on-delivery orders were cancelled 37.3% of the time. Prepaid orders, the small minority where the buyer paid before dispatch, were cancelled 2.8% of the time (73 cancellations across ≈2,600 prepaid orders). A COD order was more than 13 times more likely to be cancelled than a prepaid one.

The gap is not a one-month or one-season artifact; it held in every one of the twelve months measured. COD cancellation ran between roughly 30% and 42% every single month, prepaid cancellation never exceeded 5.9%, and the monthly ratio never fell below 6×.

Stack the leaks and the funnel looks like this: of every 100 COD orders placed, roughly 63 survive the cancellation stage and ship, about 78% of those deliver, so only about 49 of the original 100 end with cash collected. Roughly half of COD order value never becomes revenue. (This compounds a placement-stage rate with a dispatch-stage rate measured on different date bases, and some storefront cancellations are post-dispatch markings that can overlap the RTO count, so treat it as approximate: "roughly 49–57 of 100" is the honest range.)

Three honest qualifications, which strengthen rather than weaken the finding:

1. This is a behaviour gap, not a clean causal effect. Buyers who prepay have already self-selected as committed. Forcing every COD buyer to prepay would not mechanically cut cancellations 13-fold.
2. The prepaid base is modest: prepaid was 1.6% of ≈170,000 orders, i.e. ≈2,600 prepaid orders, 73 cancellations (Wilson 95% CI on the prepaid cancel rate: 2.20–3.46%). Even at the conservative edge of the statistical intervals, COD's lower bound against prepaid's upper bound, the gap remains at least ~10.7×.
3. Both directions matter. The gap measures what prepayment signals (commitment) as much as what it causes. Either way, the payment mix, not the shopper, the product or the courier, is where Pakistani e-commerce loses its money.

That is why this number belongs in the digital-payments policy conversation. The State Bank's cashless agenda is usually argued from documentation, tax and financial-inclusion angles. This dataset adds a commercial one: every percentage point of order volume that migrates from cash to prepaid moves orders into a population that has historically cancelled at about 2.8%, against 37.3% for the cash population. Committed buyers self-select into prepayment, so a migrated order would not mechanically inherit the full gap, but even if only part of the 13× difference is causal, the addressable waste per migrated order is large. That waste (confirmation call centres, two-way courier journeys, repacking, weeks of working capital in limbo) is a direct, quantifiable cost of cash dominance.

The geography of delivery: city and courier patterns

Delivery success in Pakistan is not one number; it is a map.

Across ≈35,000 settled two-courier orders spanning 50 cities (January–April 2026, within a ≈45,600-settled-order base), city-level COD delivery success ranged from above 90% on the best-served lanes to roughly 50% at the periphery.

• The best lanes clear 90%: Rahimyar Khan 90.2% (n=205, on one courier), with smaller lanes such as Wah 92.3% (n=52) and Lalamusa 90.9% (n=55) in the same band.
• The big cities cluster tightly: Karachi ~80% on either courier (80.6% / 79.5%, difference not statistically significant), Islamabad ~83% (83.2% / 82.6%, not significant), Lahore 78.5–86.0% depending on courier.
• The hardest geography is the north and the frontier: in Gilgit, delivery succeeded just 52.0% and 47.4% on the two couriers (n=161 combined), so roughly half of COD parcels never reached the buyer, whichever courier carried them. In Bannu, Manshera, Kotli and Haripur, at least one courier fell below 60%. Small samples at these extremes mean wide intervals, but the direction is consistent.

Does courier choice matter? Less than the industry assumes. Of 50 city-level head-to-heads between the two couriers that carried the largest share of the store's volume, only 7 showed a statistically significant difference (two-proportion z-tests). Where it mattered, it mattered a lot, with significant gaps ranging from 7.5 to 20.5 percentage points, but in the typical city the two couriers were statistically indistinguishable.

Two analytical warnings the data itself taught us: province-level courier league tables are misleading (a single mega-city's volume can flip a whole province's apparent "winner", a textbook Simpson's-paradox artifact we observed in Punjab), and these routing comparisons are observational, not randomised, so even the significant gaps partly reflect which parcels each courier was given.

The deeper pattern echoes the trust-gap section: in the highest-failure cities, no courier fixes the problem, because the problem being shipped is an unconfirmed cash order, not a badly driven van.

Demand itself is also concentrated: in the most recent 90 days, the top four cities (Lahore, Karachi, Islamabad and Rawalpindi) accounted for ~39.6% of all orders (≈17,000 of ≈42,900), and the top 25 city rows for roughly 57.9% of orders and 65.0% of sales value (one junk placeholder city excluded, consistent with the data-hygiene rule in the methodology). The big-city basket is also slightly larger: order values across the largest cities cluster within a narrow ~±5% band; Quetta runs about 20% below that norm, and Gilgit's average basket runs roughly 45–50% below it, about half the big-city basket.

When and how Pakistan shops: trend, seasonality and behaviour

Underneath the COD mechanics, the demand picture is steady and quietly shifting upmarket.

Volume is stable; baskets grew. Over the eleven full months from July 2025 to May 2026 the store processed ≈157,000 orders. Monthly order volume moved within a band of roughly 12,200–16,100 orders with no trend break. What changed was the basket: from December 2025 onward, average order value stepped up ~27.6% versus the July–November 2025 baseline and stayed there. With order volume roughly flat, the entire revenue lift of the past six months was basket-driven rather than volume-driven. The cause is not determinable from this data (price, mix, shipping-fee and seasonal effects are all candidates), and we make no claim about it.

January was the biggest sales month; July the biggest order month. Best month by sales value: January 2026; by order count: July 2025 (≈16,100 orders).

The week has a rhythm. In the most recent 30 days (≈15,600 orders), order counts varied only mildly across weekdays (the busiest day ran ~42% above the quietest). Basket size is the real weekly story: the week's biggest sales day carried an average basket 83% larger than its smallest, making it the #1 revenue day despite ranking only third on order count. Big-basket buying clusters on specific days.

Traffic is murkier than orders. Of 1.21 million sessions in 30 days, 54.9% carried no referrer at all, and behavioural analytics suggest up to half of all sessions show non-human (bot-like) patterns, a behavioural-signal estimate rather than an audited census. Two practical consequences follow: raw "conversion rates" in Pakistani e-commerce are close to meaningless without bot correction, and week-on-week conversion "improvements" are often just the bots leaving.

A quarter of commerce is dark. Mirroring the murky sessions, ~27% of shipped orders carried no attributable traffic source at all: commerce arriving through WhatsApp threads, chat, repeat visits and word of mouth that no attribution system sees.

What this means: implications

For retailers

The biggest leak in a Pakistani e-commerce P&L is the gap between "order placed" and "cash collected", not traffic or conversion: roughly half of every 100 COD order-rupees in this dataset (approximate two-stage funnel). The two attack points are visible in the funnel: the pre-dispatch confirmation stage (where ~three-quarters of COD failure happens) and prepaid adoption (the 13× gap). Friction should be targeted by intent. Search-intent buyers already deliver near the practical ceiling (~85%), so blanket COD friction destroys good orders to fix bad ones; the failure pool is concentrated in social-discovery impulse orders.

For couriers

In 43 of 50 cities, courier choice made no statistically detectable difference to delivery success. Performance anxiety should give way to portfolio focus on the handful of cities where capability genuinely diverges (gaps of 7.5–20.5pp), and on frontier geographies like Gilgit, the one city where neither courier clears 60%, plus the further handful of cities where at least one courier falls below 60%. The growth product couriers should sell is confirmed-order quality tooling, not speed: verification, address resolution, prepaid integration at the doorstep.

For payment providers and fintechs

The 13× gap is the business case. The marginal prepaid order is worth far more than its payment fee: it joins a population that has historically cancelled at ~2.8%, versus ~37% for cash orders, and even if only part of that 13× behaviour gap is causal rather than self-selection, the addressable failure cost per migrated order is large. Wallet and BNPL players courting e-commerce merchants should price and pitch against the failure cost they remove, not the transaction fee they charge.

For policymakers

Pakistan's cashless agenda is usually argued on documentation and inclusion grounds. This dataset adds an efficiency argument: cash dominance imposes a measurable physical tax (two-way parcel journeys, call-centre confirmation layers, weeks-long settlement lags, a median of 21.7 days for a failed delivery to surface) on a sector the country wants to grow. Policies that make small-ticket digital payment effortless (small-ticket baskets dominate this dataset) attack the waste at its root.

The national scale of that tax is large. Applied across Pakistan's e-commerce market, US$5.4 billion (~PKR 1.5 trillion) in 2024 revenue per the U.S. International Trade Administration's Pakistan Country Commercial Guide (trade.gov, updated March 2026), a ~20% COD failure rate implies on the order of PKR 16–36 billion in dead logistics cost annually, assuming ~90% of order value is cash-on-delivery, an average order value of PKR 3,000–5,000, and PKR 300–400 in round-trip courier cost per failed parcel. That is a clearly-labelled extrapolation rather than a measurement; this study quantifies the per-100-orders mechanics behind it. Even counting only digitally-paid orders, the State Bank of Pakistan's Annual Payment Systems Review FY25 records PKR 912 billion in e-commerce transactions, and the COD majority sits on top of that figure. No published national COD-share or RTO benchmark exists for Pakistan at this resolution; this study is offered as a first benchmark, not a census.

For analysts and journalists

Treat storefront-analytics numbers from any Pakistani retailer with suspicion: refunds here across the ≈13,800 April-2026 orders totalled a few thousand rupees, effectively zero, while one in five shipped parcels bounced. The entire failure mode is invisible to the dashboards most coverage quotes, and session-based metrics carry a bot-inflated denominator. The honest unit of measurement for PK e-commerce is the delivered, cash-collected order.

Methodology

Who and what this is

This study is built entirely on OrderNation's first-party operational data: one large Pakistani online retailer (mobile accessories–dominant catalogue, small-ticket baskets), selling nationwide since 2014. It is one of the largest first-party datasets on Pakistani cash-on-delivery behaviour ever published. It is not a national census: category mix, checkout design and customer base all shape these numbers, and other categories (high-ticket electronics, fashion) may differ. We publish it because no public Pakistani dataset exists at this resolution, and a well-documented single-retailer benchmark beats no benchmark.

Datasets used

Counts in this study are presented rounded; all rates and confidence intervals were computed on the exact underlying records.

1. Storefront order cohort (placement stage): all ≈170,000 orders placed in the twelve months from June 2025 to May 2026, read at order level from the storefront platform; pulled 10 June 2026, with outcomes (cancellations) read as of pull time. Payment mix, cancellation rates and the COD-vs-prepaid gap come from here. Orders are classified COD vs prepaid by payment-gateway token. Because cancellations accrue after placement, a cohort's cancel rate drifts up as it ages: April 2026's COD cancel rate read 31.4% at an early-June pull and 33.4% at the 10 June re-pull; the 10 June snapshot is canonical. For the same reason May 2026, the youngest month in the window, was still accruing cancellations at pull time; as a sensitivity check, excluding May entirely gives COD 38.0% vs prepaid 2.6%, a 14.8× gap, so the still-accruing month makes the headline more conservative, not less. A further ≈6,800 orders (4.0%) carried no recorded payment gateway; that unknown-gateway bucket cancels at 21.2% and is disclosed here but excluded from the COD-vs-prepaid comparison. All confidence intervals on cancellation rates are Wilson 95% score intervals.
2. Dispatch-outcome dataset (delivery stage): ≈127,000 shipped parcels, July 2025 – June 2026, from our internal order-management system. Dispatch-date basis; covers shipped orders only (pre-dispatch cancellations are excluded by construction); percentages computed on orders, not pieces. RTO = (returned + return-received) ÷ shipped. Twelve-month scale: roughly one rupee in every five of shipped order value (19.4%) returned to origin; ≈99,000 parcels delivered (78.4%). June 2026 was immature at measurement time (~31% of parcels unresolved): it is included in scale sums only and excluded from every rate claim; May 2026 was ~99% resolved and counts as settled. All settled-rate claims therefore use the eleven settled months, July 2025 – May 2026. For scale context only, the same system recorded ≈162,000 orders landing over the window before pre-dispatch cancellation; the order-management cancelled-status measure uses a different definition from the storefront cancellation measure in dataset 1, and the two are never combined in any sentence of this study.
3. Matured COD cohort (corroboration): cash-on-delivery orders dispatched 1 April – 15 May 2026, measured after outcomes matured; ≈11,300 resolved attributed orders (≈9,300 delivered, ≈2,100 returned). Retained as a corroboration check on the twelve-month intent gradient: in this fully matured cohort, search-intent RTO ran 13.5% vs 21.1% for social-discovery, consistent with the twelve-month class rates of 14.2% vs 22.7%. This window sits in a seasonal trough, so its rates are reliable but its absolute counts should not be annualised.
4. City–courier dataset: ≈45,600 settled orders January–April 2026, of which ≈35,000 were carried by the two couriers that carried the largest share of the store's volume, across 50 cities; per-city two-proportion z-tests.
5. Storefront aggregates (trend and behaviour): live analytics pulls dated 10 June 2026: 12-month monthly order series (≈157,000 orders over 11 full months), 90-day city distribution (≈42,900 orders), 30-day session and day-of-week windows.

Honesty notes and caveats

Funnel-stage honesty. Datasets 1 and 2 measure different stages on different date bases (order placement, June 2025 – May 2026, vs dispatch outcome, July 2025 – June 2026; overlapping but not identical twelve-month windows, and the two datasets also overlap order-for-order and their order counts are never summed). The "100 placed → ~63 shipped → ~49 collected" funnel compounds a placement-stage rate (COD cancellation 37.3%) with a dispatch-stage rate (78.4% of shipped parcels delivered) and is therefore approximate. Because some storefront cancellations are post-dispatch markings that can overlap the dispatch dataset's RTO count, the true figure may sit somewhat higher; the defensible range is ~49–57 of 100. Pre-dispatch cancellation and post-dispatch RTO are distinct failure events and are never summed elsewhere in this study.

Why storefront analytics under-count COD failure. Refunds in the April cohort totalled a few thousand rupees (effectively zero, with zero return objects), while the dispatch dataset shows ~20% of shipped parcels returning. COD failure does not surface as refunds: storefront platforms have no native order status for a courier return, so a failed COD delivery is recorded in the logistics system rather than as a storefront cancellation or refund. Storefront cancellation rates are therefore a floor on true failure.

External benchmark context. Published industry estimates for COD-heavy e-commerce markets such as India put return-to-origin (RTO) rates on COD orders in the 20–30% range (Shiprocket, 2025), against an industry-wide average of 20–25% that can spike toward nearly 40% (Shiprocket, 2025; GoKwik, 2026); Shipway's ShipNotes FY25 report, as covered by Financial Express B2B (2025), records roughly 26% of Indian COD orders returned versus under 2% of prepaid orders. The ~20% blended RTO measured here sits at the favourable end of that band. To our knowledge, no Pakistan-specific dataset of cash-on-delivery failure, cancellation or return-to-origin rates had been published before this study (literature and market search, June 2026) — which is why the nearest available benchmarks are Indian.

The 13× gap is a behaviour gap. Prepaid buyers self-select as committed; the gap should be read as "COD and prepaid orders behave roughly 13× differently", not "switching a buyer to prepaid causes a 13× reduction". The prepaid base is modest (≈2,600 orders, 73 cancellations; Wilson 95% CI on the prepaid cancel rate 2.20–3.46%); the conservative bound on the gap (COD's lower confidence bound divided by prepaid's upper) is still ≥10.7×, and the monthly ratio never fell below 6× in any of the twelve months measured.

Intent-gradient caveats. Traffic-source attribution covers ~73% of shipped orders (measured on the six-month dispatch window to June 2026); per-source findings describe the attributable subset. The smallest source class, the one at the top of the 14.2%-to-roughly-35% spread, is directional only (roughly 700 shipped parcels in the first half of 2026; roughly 1,000 across calendar 2025), due to sample size and a known attribution outage in late May 2026.

City–courier caveats. Routing was not randomised; couriers may receive systematically different parcels, so even significant gaps are observational. Only 7 of 50 city deltas pass significance testing; small-city extremes (e.g., Gilgit, n=161) carry wide intervals; province-level rollups are vulnerable to mix artifacts (Simpson's paradox) and are avoided. The window is four months, one season, with no replication yet.

Traffic and conversion caveats. Session data is heavily polluted: 54.9% of 30-day sessions carry no referrer, and behavioural analytics suggest up to ~50% of sessions show bot-like behaviour (a behavioural-signal estimate, not an audited bot census). We therefore publish no conversion rates.

Data hygiene. City names were case-normalised before grouping (the raw field is free text). One junk placeholder city (≈270 orders, immaterial in value) was excluded from city analysis. The billing-region field is unpopulated in Pakistani checkouts (99.99% null), so no province-level claims are made from it. Partial edge months (June 2025, June 2026) are excluded from all trend lines.

What we deliberately do not publish

Absolute rupee figures of any kind (revenue, GMV, returned value or basket values; all money in this study is expressed as rates, ratios or relative comparisons, and counts are presented rounded); per-order cost structures; courier and vendor identities; advertising spend; and any datum traceable to an individual customer.

About OrderNation

OrderNation is one of Pakistan's longest-running online retailers, founded in 2014 and headquartered in Lahore. The company specialises in mobile phone cases and accessories, shipping nationwide to all major cities and beyond via Pakistan's leading courier networks. Having processed orders for over a decade almost entirely on cash-on-delivery terms, OrderNation holds one of the country's deepest first-party datasets on Pakistani online buying behaviour. Research enquiries: support@ordernation.com.

Frequently asked questions

What percentage of online orders in Pakistan are cash on delivery?

94.5% of ≈170,000 orders placed between June 2025 and May 2026 in OrderNation's dataset were cash-on-delivery; digital prepayment accounted for just 1.6%. No published national census exists, but this is one of the largest first-party samples of Pakistani online payment behaviour ever published.

What is the RTO (return-to-origin) rate in Pakistani e-commerce?

Roughly 1 in 5 shipped parcels (19.4% of shipped order value) was returned to origin in OrderNation's dataset, measured across ≈127,000 shipped parcels from July 2025 to June 2026. The rate held at 19–22% in every fully settled month, indicating a structural property of the cash-on-delivery mechanism rather than a seasonal effect.

Do prepaid orders get cancelled less than COD orders?

Yes, dramatically. In OrderNation's dataset, prepaid orders were cancelled 2.8% of the time versus 37.3% for cash-on-delivery orders across the same twelve months (June 2025 – May 2026), making a COD order more than 13 times more likely to be cancelled. The gap held in every individual month, never falling below 6×.

Which cities in Pakistan have the best delivery rates?

Well-served Punjab lanes lead: Rahimyar Khan (90.2%), Wah (92.3%) and Lalamusa (90.9%) all cleared 90% COD delivery success in this dataset (January–April 2026). The big cities cluster at roughly 78–86% (Karachi ~80%, Islamabad ~83%, Lahore 78.5–86.0% depending on courier), while Gilgit sits near 50%, the widest geography gap measured.

How was this study conducted?

The study is built entirely on OrderNation's first-party operational data: ≈170,000 orders placed June 2025 – May 2026 for payment mix and cancellations, ≈127,000 shipped parcels July 2025 – June 2026 for delivery outcomes, and ≈45,600 settled orders across 50 cities for the courier analysis. Counts are presented rounded; all rates were computed on exact records, with Wilson 95% confidence intervals on cancellation rates and two-proportion z-tests on city–courier comparisons.

Why doesn't COD failure show up in standard e-commerce analytics?

Because failed COD deliveries never become refunds: refunds across ≈13,800 April-2026 orders totalled a few thousand rupees (effectively zero) even as one in five shipped parcels was returned. Storefront platforms have no native order status for a courier return, so the loss is recorded in logistics systems that standard dashboards never see.

How does Pakistan's return-to-origin rate compare with India's?

The ~20% blended RTO measured here sits at the favourable end of published benchmarks for COD-heavy markets: industry estimates for India put COD RTO at 20–30% (Shiprocket, 2025), against an industry-wide average of 20–25% that can spike toward 40% (Shiprocket, 2025; GoKwik, 2026). Prepaid orders in India return at under 2% per Shipway's ShipNotes FY25 report (Financial Express B2B, 2025), the same commitment gap this study measures in Pakistan.

Is this the first study of its kind in Pakistan?

To our knowledge, yes. Pakistan-wide e-commerce market reports exist, but no previously published study has reported order-level cash-on-delivery failure, cancellation or return-to-origin rates for Pakistan. This report, built on ≈170,000 orders placed over twelve months, is, as far as we have been able to determine, the first publicly available dataset of its kind in the country; the nearest published benchmarks come from India.