ExamplesOnline storesLaunch your first storeFix the store you haveGrow a store that is sellingWebsites, AED 3,500 flatClinicsVetsRestaurantsGaragesLaw firmsPricingBlog
21 September 2026 · 8 min read
We timed 1,876 deliveries out of our own store across eleven markets. The slow part was never the warehouse, and it was never the driver.
For about a year I answered the question "how long does delivery take?" with a guess. A good guess, I thought. We ship out of Dubai to eleven markets, I read the complaints, I had a feel for it.
Then in August we pulled the event logs. Not the monthly summary a courier sends you, the per-order log: the timestamp on every scan, from the moment an order lands to the moment someone signs for it. 2,103 logs, 1,876 of them complete enough to use, January through early August.
We could only do this because SwftBox, who move our parcels, expose that log, and our live stock, through their API. Most couriers in this region will send you a delivery rate and a monthly invoice and nothing in between. If your provider cannot give you a per-order timeline, you cannot run the analysis below at all, and that on its own is worth knowing before you sign with anybody.
Almost everything I believed about where the time went was wrong. Not slightly wrong. Wrong in the sense that I had been trying to fix the wrong half of the journey.
Here is the median, market by market. Median, not average, because one order stuck in a depot for a month drags an average somewhere no customer has ever been.
zamani, Jan to Aug 2026
Median time from order placed to delivered
Now the part that reorganised my thinking. In all eleven markets, the time from the order landing to the parcel reaching a depot is between 9.7 and 14.7 hours. In all eleven, the final run from the depot to somebody's door is between 3.3 and 6.9 hours.
Read that again with the chart above it. The two ends of the journey are roughly identical whether the parcel is going to Sharjah or to Johannesburg. Everything between 0.9 days and 13.4 days is the middle leg, the part where the box is sitting in a vehicle or a warehouse waiting for a truck or a plane that has nothing to do with us.
The clearest version of this is Malaysia. Its depot handling is the fastest of any market we ship to, faster than the UAE's. It is also our second slowest country, at eight days. Those two facts sit in the same row of the same spreadsheet, and until we had the per-order log there was no way to see both at once.
Domestic is where we have actual control, so we cut it finer. This is the customer's clock, from checkout to doorstep.
UAE only, 242 orders
Where a UAE delivery day actually goes
I had spent months thinking about the courier. The courier was fine. The courier was always fine. Six of every ten hours a UAE customer waits are hours the order spends sitting with us, before it has been handed over at all.
The one lane I was actually ready to complain about went the same way. Saudi cash-on-delivery looked much slower than Saudi prepaid, and on the courier's own clock the two are identical: 4.05 days against 4.11. The whole gap was an order-hold rule we had written ourselves and then forgotten about. I was one email away from raising it with our provider as their problem.
That is an uncomfortable thing to find in your own data, and it is also the best possible news, because it is the half we can change without renegotiating anything with anybody.
Then we split domestic orders by the hour they were placed, Dubai time. I did not expect this to show anything.
UAE orders by time of day placed
The 1pm problem
An order placed at noon is in someone's hands in about eight hours. The same order placed at 2pm takes twenty six. Nothing about the customer changed. They crossed an internal dispatch cutoff that exists nowhere on our website, in any email, or in any promise we have ever made.
We are not the only store with a cutoff. I would bet most stores in this country have one. I would also bet most of them have never looked at what it does, and none of them tell anybody.
"It is the addresses." UAE addresses are famously loose. No postcodes worth using, landmarks instead of street numbers. Obviously that was slowing us down.
It was not. The relationship runs backwards. Malaysia and South Africa, our two slowest markets, have the most complete address data we hold: near enough every order has a postcode and a second line. The UAE, our fastest market by a distance, has short unusable addresses on 27% of orders and a postcode on about one in a hundred. Bahrain has the sparsest data of anywhere and is our second fastest market.
There is one real exception and I will concede it rather than bury it. In Jordan, the handful of orders with a genuinely sparse address took 8.9 days against 5.6 for the rest of the lane. That explains Jordan's worst tail. It does not explain Jordan's middle, and it does not generalise: in the US and UK, sparse addresses came out marginally faster.
"It is Ramadan." Ramadan did slow the UAE and Saudi, by under a day each. In the same window Jordan and Malaysia both got faster. Whatever is happening in those two lanes is not seasonal and it is not religious, it is structural.
Everything above is about the spread between markets. The thing I did not expect was how little movement there was inside each one.
Change in median, Q1 against Q2 and Q3
Which lanes got faster
Five lanes got faster over the same period, four moved by a third of a day or less, and one moved properly. Our domestic median went from 1.26 days in the first quarter to 0.89 from April, with a 90th percentile of two days, which is the number that actually matters: nine orders in ten, not the middle one.
That is the part of a courier relationship you cannot see from a delivery rate. A 96% success rate tells you nothing about whether the 96% arrive when you said they would, and a customer who gets the parcel two days late has a worse experience than one who gets a refund quickly. Consistency is the product. Speed is a headline.
Three things, and none of them is about speed.
The first is the tracking link. Every order carries one the customer can open themselves, and "where is my order" stops being an email we answer by logging into a dashboard. It is the single most common question any store gets. Customers tell us they like it, unprompted, which is not something people usually bother to do about a logistics feature. A shopper who can see where the parcel is does not need us to be awake to reassure them.
The second is inventory. The same API exposes stock at SKU level: what is physically in the warehouse, what is already allocated to open orders, what is damaged, and what is inbound on a reorder that has not landed yet. Those are four different numbers and most stores treat them as one, which is how you sell something you do not have. We can now check a SKU against the warehouse instead of against our own store's guess about itself.
The third is the reason this whole post exists. The event log is granular enough to tell a real failed delivery from a bookkeeping artifact. 17% of our orders carried a failed-attempt flag. Fewer than half were genuine. Without the per-order timeline we would have taken the 17% at face value, and I would have spent a meeting arguing about a number that was mostly noise.
A courier that hands you a monthly summary gives you a feeling about your shipping. One that hands you stock levels and every scan gives you an argument you can check, including the ones where you turn out to be wrong. We have now been wrong twice in this post using their own data.
In June we had a month roughly four times the size of May. I fully expected the delivery data to fall apart.
The median moved by three and a half hours. What moved properly was the same-day share, which fell from 62% to 49%, then went back up in July.
That is worth knowing if you are about to run a campaign. A warehouse under pressure does not usually break, it quietly stops doing the fast version. Your median barely moves, so your dashboard looks fine, and meanwhile half the people who would have had their order that day now get it tomorrow. Nobody complains about that. They just remember you as slower than they expected.
Three warnings, each of which cost us real time.
Your database is probably storing two timestamps in two different timezones. Ours keeps order dates in Dubai local time and courier timestamps in UTC. Subtract them raw and every delivery looks four hours faster than it was, and the quickest orders come out as negative numbers, which is at least an obvious symptom. A four hour error you cannot see is worse than one you can.
Your fulfilment date is probably not a delivery date. Ours matched the real delivery event on 99% of UAE orders and on 10% of everything else, because cross-border orders get their final status from a polling job that updates the status and leaves the date frozen at an earlier scan. When it is wrong it is wrong by about four days. Every cross-border number we had was quietly flattering us.
And a failed delivery attempt in a courier system is often not a failed attempt. 17% of our orders carried one. Under half of those were real. If you report that number without checking, you will go and have a difficult and completely unnecessary conversation with your logistics provider.
This is not really a logistics post. Delivery speed is a conversion lever and a repeat-purchase lever, and almost nobody measures it well enough to pull it.
The three things worth doing, in order: find your own cutoff and either publish it or move it, because a customer who orders at 2pm and waits a day longer for no visible reason has learned something about you. Measure the leg you control before you go and argue about the leg you do not. And stop reporting medians on their own, because the median is what survives a bad month while your best customers notice the difference.
All three need per-order timestamps you can actually read, and that is a question for whoever ships your parcels, not for your developer. Ask it before you ask about price.
We run our own stores and this is the kind of thing we find in them, usually by accident, usually later than we should have. If you have a store and you have never pulled these numbers, that is the teardown call worth having, and it is free. If you want the tracking and the fulfilment plumbing set up properly in the first place, that is what a Setup is.
Every figure above is from our own store, measured between 1 January and 6 August 2026. I would rather publish the ones that made us look slow than the ones that did not.
We open your store with you and name the three biggest leaks. Useful whether or not you hire us.