Measured benchmarks for Pakistani cash on delivery
Every statistic about Pakistani e-commerce we could find traces back to a secondary aggregator with no stated methodology. These are computed from real parcels, and every figure carries the number of parcels and stores behind it so you can judge it.
More than one COD parcel in four comes back
This is the number that decides whether a Pakistani COD store makes money, and it is the one most often guessed at. 27.5% of parcels were refused — earning nothing while costing forward freight, return freight and packaging. See what an RTO actually costs.
COD settles faster than most sellers expect
The median is fast. The spread is the finding: the gap between a 2-day median and an 8-day 90th percentile is where working capital actually gets trapped, and an average alone would have hidden it. If you plan cash against the average you will be short in the weeks the tail lands.
Delivery rate varies 7.5 points across the four biggest cities
Islamabad delivers best, Rawalpindi worst — two cities that are effectively adjacent. That spread is larger than most operational changes a store can make, which is why routing by measured city performance is worth more than it sounds. See the courier scorecard.
The four numbers
- 27.5% of COD parcels returned · n=11,193 · 5 stores
- 2 days median from delivery to settlement, 8 at the 90th percentile · n=8,079
- 76.5% delivery rate in Islamabad, the best of the four largest cities
- 69.0% in Rawalpindi, the worst — a 7.5-point spread
How to read this honestly
Five stores is a small sample, and we are not going to dress it up. These figures describe the stores currently running on Kaarobar over 180 days — a real, measured population, but not a representative survey of Pakistani e-commerce. Category mix, price point and courier choice all move these numbers, and this sample does not span them evenly.
What it is: better than any figure we could find with a stated method behind it. What it is not: a national average. Treat it as a starting reference and measure your own.
What we intend to measure next
- RTO rate by category and price band — currently suppressed, because breaking 5 stores down by category produces cells that describe one merchant.
- Delivery rate by courier — same reason. It becomes publishable when enough stores run each carrier that no cell is one store's data wearing a disguise.
- The fully loaded cost of a refused parcel, including freight both ways and packaging, rather than the freight line alone.
- Seasonality — Ramadan, Eid and sale events move all of the above, and an annual average conceals more than it shows.
The method, stated in full
- Outcome comes from timestamps, not status text. A parcel counts as delivered or returned based on when it was recorded as such, never by pattern-matching a courier's status string. Every carrier words its statuses differently and changes them without notice, so a benchmark built on string matching silently measures a different thing per courier.
- Only terminal outcomes count. Parcels still moving are excluded entirely rather than counted as not-yet-delivered, which would drag every rate down by however much was in flight on the measurement date.
- Sample size always shown. Parcels and stores, on every figure. A rate without a denominator is a rumour.
- Thin cells are suppressed, not rounded. No figure is published from fewer than 3 stores or fewer than 500 parcels. Below that, an "aggregate" broken down finely enough is one merchant's data wearing a disguise — hiding the store name does not fix that, so those cells are simply not emitted.
- Aggregate only, never per store. No store is identifiable in anything on this page, and the query that produces these figures never selects a store identifier at all.
- Period stated and dated. These cover the 180 days to 30 August 2026. Pakistani e-commerce is intensely seasonal, so a figure without a period is not much use.
The method is enforced in code rather than by discipline: the suppression floor is a constant in the script that produces these numbers, so a cell that falls below it cannot be published by accident.
If you run a Pakistani COD store
- Your own numbers matter more than these. A benchmark cannot tell you whether last month was better than the month before; only your own baseline can.
- Both figures above are computed for your store individually in the app — see the courier scorecard and returns and RTO.
- If your RTO rate is materially above 27.5%, the five levers are where to start.
Questions merchants actually ask
- What is the average RTO rate in Pakistan?
- Measured across 11,193 cash-on-delivery parcels from 5 stores over the 180 days to 30 August 2026, 27.5% were returned to origin and 72.5% were delivered. That is a real measurement rather than an estimate, but it is a small sample and not a representative national survey — category, price point and courier choice all move it.
- How long does a courier take to remit COD in Pakistan?
- Across 8,079 settled parcels, the median was 2 days from delivery to settlement, the mean 3 days, and the 90th percentile 8 days. The spread matters more than the median: one parcel in ten takes at least four times the typical time, and that tail is where working capital gets trapped.
- Can I cite these figures?
- Yes, with a link, and please carry the sample size and period with them — 11,193 parcels, 5 stores, 180 days to 30 August 2026. That is what makes a statistic checkable rather than another unattributed number.
- Is any individual store identifiable in this data?
- No. Figures are pooled across stores, the query never selects a store identifier, and any cell built on fewer than 3 stores or 500 parcels is suppressed entirely rather than shown.