Cainiao extends Mexico express network from 20 states to nationwide with 38 delivery stations

The Analysis

A parcel that misses a cut-off by ten minutes does not arrive ten minutes late. It arrives one full cycle late. That is the whole game in last-mile, and it is the reason I read a network announcement as a set of clocks rather than as a map.

Here is what Cainiao put out. Its Mexico express network has gone from 20 states to nationwide coverage. Local pickup is live in Mexico City and the State of Mexico. Key parts of the capital region are quoted at two-day delivery, other major cities at three. Dispatch, route planning, address recognition and image recognition run on AI. Since entering Mexico in 2024 the company has built a 20,000 square metre sorting centre beside Felipe Angeles International Airport, plus 38 consolidation and delivery stations, and more automation equipment is going in. It works with Inditex, Mercado Libre, Temu and TikTok, and it sells customs clearance plus last-mile delivery as one service.

Nationwide is a coverage word. Two-day is a clock word. They are not the same sentence, and only one of them tells you what happens to your parcel on a Tuesday afternoon. The clocks are what I care about, so let me take them apart and put hours on every one of them.

Coverage answers a different question from speed. Nationwide coverage tells you a parcel can physically reach an address somewhere in the country. It does not tell you how many hours that takes, which clock it starts on, or what happens when the first attempt fails. A network can be nationwide and still be slow in the only state you ship to. Read the coverage sentence and the speed sentence separately, and do not let the first one flatter the second.

Two-day delivery is 48 hours. Three-day is 72. Those are budgets, not a description of a van driving for two days. A budget has to be spent across five segments, and if you do not know how many hours each segment eats, you cannot tell which one is stealing your promise. The segments are: order to pickup cut-off, pickup to sorting centre, sorting and outbound, customs release, and station to door. Ask for the hour count on each. If nobody will give it to you, time ten of your own parcels and use your own numbers.

Let me put a full hour model on the table, clearly labelled as mine. Assume a daily pickup cut-off at 18:00 local. Assume pickup to the sorting centre is eight hours. Assume sorting and outbound is four hours. Assume customs release takes twelve hours. Assume sorting centre to delivery station is six hours. Assume station to the opening of a delivery window is fourteen hours. Assume the delivery window runs six hours to completion.

Run it. Ordered at 17:50, the parcel enters the network at 18:00, reaches the sorting centre at 02:00, leaves it at 06:00, clears at 18:00, reaches the station at midnight, the delivery window opens at 14:00 the next day, and it completes at 20:00. That is 50 hours from order to door. The promise said 48. You are two hours over on a perfectly clean run with no failures anywhere.

The three-day figure deserves the same treatment. Seventy-two hours is 48 plus 24, and that extra 24 hours is not slack, it is the intercity leg. Assume long-haul from the sorting centre to a station in another major city takes between 10 and 18 hours depending on distance and on whether the parcel catches a direct run or gets cross-docked on the way. Assume station to door is another 6 to 14 hours. That leaves somewhere between 40 and 56 hours for pickup, sorting and clearance combined. The 72-hour promise is not more generous than the 48-hour one. It is the same budget with a fixed transport tax taken out first, and the segment that gets squeezed is customs, not transport. If you ship outside the capital region, customs is the first segment you measure.

Now order at 18:10. Twenty minutes later on the wall clock, and the parcel does not enter the network until 18:00 the following day. Everything downstream shifts by 24 hours and completion lands at roughly 74 hours. Twenty minutes cost you a full cycle. This is the single most useful thing to understand about a two-day promise: the cut-off is a cliff, not a slope. Once that window shuts, it is over.

So the first job is to own the cut-off. Find out the exact pickup cut-off for your account, whether it moves at weekends, whether it moves on Mexican public holidays, and whether it is measured at pickup scan or at the moment the parcel reaches a station. Then set your own last order time two hours earlier than whatever they tell you. Those two hours cost you nothing in customer experience, because customers respond to the promise you printed, not to the carrier's internal clock.

Then find out whether customs hours sit inside the 48 or outside it. This matters more than anything else in the model, and it is the question sellers forget to ask precisely because the announcement sells clearance and last mile as one product. If clearance is inside the promise, my twelve-hour assumption is part of your budget. If it is outside, add it: a 48-hour promise with clearance running on a separate clock is a 54 to 84 hour reality depending on how clean your paperwork is. Get the answer in writing before you print a delivery date on a product page.

Address quality is the second clock and it is the one that quietly destroys promises. A Mexican address has more parts than a US one: street, exterior number, interior number, colonia, código postal, alcaldía or municipio, the two cross streets known as entre calles, a reference, and a local phone number. AI address recognition, which the announcement says the network uses, will fix formatting and will guess at structure. It will not invent a street that does not exist inside the postal code your customer typed. When colonia and código postal disagree, the parcel still goes out, the driver still cannot find it, and the clock starts again from zero.

Here is what a failed attempt costs in hours. Assume the first attempt fails on the morning wave. Assume the second attempt goes out in the next available wave, 24 hours later. Assume the driver needs a corrected address and the customer replies the following morning. That is 48 hours gone on one bad address. Your 48-hour promise has become a 96-hour promise, and the customer is measuring you against the number you advertised, not against the number the network can deliver.

Now the money. Assume 10,000 parcels a month. Assume a first-attempt success rate of 90 per cent, so 1,000 failures. Assume a re-attempt costs 1.50 dollars and a customer message costs 0.05 dollars, and assume every failed parcel needs two attempts and one message. That is 1,000 times 3.05, roughly 3,050 dollars a month. Assume 1 per cent of failures end in a return at 5 dollars each, another 500 dollars. Call it 3,550 dollars a month for bad addresses. Lift first-attempt success from 90 to 95 per cent and you remove 500 failures, which is about 1,525 dollars a month back. Those are my numbers for illustration. The structure is what to take away, and the structure says address quality is a profit and loss line, not a form design question.

Fix it at the point of capture. Make the address form structured instead of one free text box: street and number in separate fields, postal code first with colonia and municipality auto-filled from a lookup and validated against it, cross streets and a landmark reference required, and a local mobile number mandatory with a length and format check. Reject the order when postal code and colonia disagree rather than shipping it and hoping. A rejected checkout is annoying for thirty seconds. A parcel bouncing around a station for 96 hours costs you the customer.

The third clock is pickup coverage, and there is a distinction in the announcement that will bite you if you read it fast. Nationwide coverage is one thing. Local pickup is stated for Mexico City and the State of Mexico. Those are two different promises. If you ship from another state, ask directly: is my origin inside the pickup network, or inside the delivery network only? If it is delivery only, your parcel enters the network at a consolidation station, on your own transport, inside that station's receiving window, and every hour of that first leg is yours to manage and yours to lose.

Assume a station receiving window of 09:00 to 18:00. Assume no weekend receiving unless someone confirms it in writing. A parcel you hand over at 18:30 on Friday does not start its clock until 09:00 on Monday. That is 63 hours of dead time before anything moves at all, and no amount of AI route planning buys it back. Get the receiving window and the weekend rule in writing for every station you depend on, and build them into the promise you show at checkout.

The fourth clock is the linehaul into the sorting centre. The facility is 20,000 square metres next to Felipe Angeles International Airport, which is not in the middle of the capital. The run from the airport area into the city is real hours and it varies by time of day. Do not estimate it from a map. Ask for the carrier's own linehaul timings, and if you can, time ten of your own parcels from arrival scan to station scan. The most common operational mistake I see is treating arrival at the sorting centre as arrival in the city. It is not, and the gap between the two is exactly where a 48-hour promise quietly becomes a 72-hour one.

The 38 consolidation and delivery stations set the rhythm of the last mile. Station density decides how often a van leaves and how short the run is. Assume two dispatch waves a day at a station, one at 10:00 and one at 16:00. Miss the 10:00 wave and your parcel waits six hours for the next one. If the 16:00 wave fails, the next attempt is tomorrow. That is how one missed wave turns into a full extra cycle, and it is why wave times belong on your operations dashboard rather than in a carrier brochure.

Peak is the fifth clock and it is the one nobody budgets for. The announcement says more automation equipment is going in to raise sorting efficiency and peak capacity, which tells you two things: the automation is not fully in yet, and peak is a known constraint they are planning around. Assume peak volume runs at three times baseline. Assume the bottleneck is the night shift, because sorting stays labour-bound wherever automation is thin. Plan on the basis that a 48-hour promise is a baseline-week promise and not a Buen Fin week promise, and say so to your customers before the week arrives rather than after.

The practical answer for peak is to move stock closer before the week arrives. Assume your top 20 per cent of SKUs drive 80 per cent of your volume. Get those into a Mexican local warehouse six weeks before peak and set a hard cap of 30 per cent on the share of peak volume that ships cross-border direct. Local stock turns a 48-hour international promise into a same-day or next-wave domestic one, and it takes customs and linehaul off your critical path entirely. It costs you carrying cost and a warehouse fee. Compare that against 3,550 dollars a month in address failures plus the revenue you lose when a peak shopper never comes back.

Measure three numbers every day and argue with them. Cut-off hit rate, the share of orders that enter the network at or before the cut-off, floor of 98 per cent. First-attempt delivery success, floor of 95 per cent. Median customs release time, ceiling of 12 hours. Break all three down by state and by station. Do not look at a national average. A national average lets a strong Mexico City number hide a weak number in a state where you are losing customers, which is precisely what an average is designed to do.

Exceptions deserve their own queue and their own clock, because that is where a good network still loses parcels. The announcement says image recognition is in use, which means there is probably a photo or a scan attached to each delivery event. Ask what proof of delivery is captured, how long it is retained, and how fast you can pull it when a customer says the parcel never arrived. Assume a dispute costs your team four hours if the evidence is one click away, and two full cycles of back and forth if it is not. Put one named person on the exception queue every morning with a single rule: nothing sits longer than four hours without an action taken on it. That rule is worth more than any route optimiser you can buy.

One rescue option is worth building before you need it, which is letting the customer choose a pickup point. A locker or a station collection removes the failed-attempt clock entirely, because there is no driver trying to find a colonia at 16:00 on a Thursday. It trades your last-mile risk for the customer's travel time, and in the dense parts of the capital region that trade is usually a good one. Assume a pickup point holds a parcel between 72 and 168 hours before returning it, and assume collection happens within 24 hours of the notification going out. Budget for the notification as well: a collection point nobody was told about is just a slower failure.

Before you sign anything, get the definitions in writing, because every one of them is a clock. The start point of the promise. Whether the clock pauses for customs, for weekends, for public holidays or for a failed attempt. What a business day means in Mexico and whether Saturday counts. What counts as delivered when the parcel goes to a pickup point or a locker instead of a door. What proof is captured, how long it is kept and how fast you can retrieve it. What happens on the second and third attempt, and who pays for them. And a weekly report giving hour counts by state and by station rather than one national average. None of this is exotic. All of it is the difference between a promise you can defend and a promise you end up refunding.

Set reaction times alongside the targets. On a failed attempt, contact the customer within four hours by WhatsApp or SMS with a link to confirm or correct the address, and get the second attempt out within 24 hours of the correction arriving. A parcel sitting for three days waiting for somebody in your team to notice is not a network failure. It is your failure, and it is the one you can fix this week without asking anybody for permission.

The pitfalls, in the order in which they cost hours. The start point of the promise: order time, payment completion, pickup scan, arrival scan or customs release. Five different clocks, and the spread between the first and the last can be more than a full cycle. Whether customs hours are inside the promise. Whether the two-day and three-day figures count calendar days or Mexican business days, and whether Saturday counts. What counts as delivered: door, pickup point or locker, and what proof the network keeps, given that image recognition is in use. Returns: the announcement says nothing about a reverse network, so do not assume the journey back takes the same hours as the journey out. Platform relationships with Inditex, Mercado Libre, Temu and TikTok do not mean your parcels get the same handling theirs do, because platforms have their own injection arrangements.

And one that catches everybody: free holding time at a consolidation station. Ask how many hours a parcel waits at a station before it is sent back, and how long the return journey itself takes. A parcel that reached a station and was never collected is still a failed delivery, and the customer will remember it as yours, not as the network's.

What I would do this week. Get the pickup cut-off in writing for my account, including weekends and public holidays, and pull my checkout's last-order time two hours ahead of it. Get written confirmation of whether customs hours sit inside the 48 and 72 hour promises. Rebuild the address form with postal code validation, mandatory colonia, cross streets and a local mobile number. Instrument the three metrics with state and station breakdowns and look at them every morning. Book peak stock into a Mexican warehouse six weeks out with a hard cap on cross-border share. And put a four-hour customer contact rule plus a 24-hour re-attempt rule into the exception queue, because that queue is where promises actually break.

Nationwide coverage I believe. Two-day delivery I believe too, on a clean week, with a clean address, on a parcel that made the cut-off. What I want is the hour count on every segment written down, so that when a customer asks where the parcel is I can answer in hours instead of apologising in adjectives. When something breaks, get the parcel back first. The reporting can wait until tomorrow.

↑ Back to the key points

— By Nadia Pryce