Mercado Libre reported on November 4 that fulfilment space grew almost 20 percent quarter over quarter into the peak trading season, roughly double the pace added a year earlier, while its Brazilian cost per shipment fell 5 percent year over year in local currency. As of November 6, 2026, the published numbers support both the capacity figure and the unit cost figure, but they do not settle the question that decides whether either one pays back: what the stockout rate looks like by fulfilment site and by postcode once that space is full.
Capacity Grew Twice as Fast as Last Year
The comparison management put on the record is the useful part. Adding almost 20 percent of fulfilment space in a single quarter is roughly double the 10 percent added in the same quarter a year earlier, and it lands in the three months that carry the heaviest order volume of the year. Space added at that speed is not a one-off repair of a strained network; it is a decision about where stock will sit for the next twelve months, taken at the point in the calendar when the cost of being wrong is highest.
Read against the item count, the bet is easier to size. The platform sold 867 million items in the quarter, up 36 percent, and added 17 million buyers year over year, so the network is absorbing more parcels per buyer at the same time as it absorbs more buyers. Management framed the spending as buying frequency rather than buying capacity, and the sequence matters: if frequency holds, the added space is used, and if it slips, the added space becomes a fixed cost with no offsetting revenue attached to it.
Frequency Is What Pays for the Space
Items per buyer is the number that justifies the square metres. In Brazil the rate rose 16 percent, having run at a single-digit pace before the free-shipping threshold was cut in June 2025, and it held at that level even after the company fully lapped the anniversary of the change. A threshold cut buys order frequency by paying part of the shipping bill, which means the capacity addition should be read as the delivery end of a pricing decision rather than as an independent logistics project, and it will be unwound or extended on the same logic that set it.
Margin Compression Is the Cost of the Bet
The bill arrived in the same release. Operating margin fell 40 basis points sequentially to 6.3 percent, against 9.8 percent a year earlier, and shipping accounted for about 60 basis points of the sequential drag once the lower threshold, the capacity addition, higher energy prices and temporary recruitment are added together. Management said the capacity-related part should ease in the fourth quarter as volume rises while energy costs persist. That is a forecast, not a result, and the difference matters for anyone planning shelf coverage on the assumption that the network will keep absorbing volume at the current cost.
Cost per Shipment Fell While Space Grew
The counterweight is unit cost. In Brazil, cost per shipment fell 5 percent year over year on a foreign-exchange-neutral basis despite the added space and higher energy prices, and it sits 14 percent below the level recorded before the threshold was lowered. That combination is unusual, because capacity normally arrives before the volume that fills it, so unit cost tends to rise for a quarter or two before it falls. Getting both at once suggests the network is densifying inside catchments it already serves rather than opening new ones, and density is exactly the variable that changes how a missed replenishment travels through a delivery promise.
| Metric | Third quarter 2026 | What it implies for shelf availability |
|---|---|---|
| Gross merchandise volume | 23 billion dollars, up 33 percent at constant currency | More parcels through an estate that is still being built out |
| Items sold | 867 million, up 36 percent | Availability is tested per item, not per order |
| Fulfilment space | Up almost 20 percent quarter over quarter, against about 10 percent a year earlier | Capacity added ahead of peak rather than during it |
| Brazil cost per shipment | Down 5 percent year over year, 14 percent below the pre-threshold level | Density improved while space grew |
| Brazil items per buyer | Up 16 percent | Repeat demand is the payback case for the space |
| Operating margin | 6.3 percent, down 40 basis points sequentially | The build is funded from current margin |
The table also shows why a sales number cannot answer the availability question. Volume, items and space move in the same direction, and none of the three says whether a particular item was on the shelf when a particular shopper looked for it. A network that lifts items per buyer by 16 percent while adding a fifth more space can still post a worse stockout rate than a year earlier, because inside individual catchments the demand curve moves faster than the replenishment curve, and the gap between the two is what a shopper experiences as an empty listing.
Stockout Risk Moves With the Delivery Radius
Denser fulfilment changes the shape of the risk rather than removing it. When a site serves a shorter radius, a missed inbound appointment does not produce a slow delivery; it produces a hole in a smaller set of postcodes, and that hole is visible to every shopper inside the radius at the same moment. The same density that makes a one-day promise credible makes a bad week cluster, which is why the useful unit of measurement is the site and the postcode rather than the country, and why a national availability average can look stable while three catchments lose a week of sales.
Substitution Hides a Hole From the Sales Line
A stockout rarely appears as a lost sale in platform reporting, because the shopper is offered a substitute in the same page view. The order still lands, the item count still rises, and the availability problem disappears from the numbers that get published. What stays visible is the substitution path: which listings absorbed the demand, whether they were cheaper or more expensive, and whether the same seller lost the sale to another seller. Tracking that path is the only way to see a hole after the fact, and it is also the only way to tell a range decision from a supply failure.
The Site Level Audit
What the disclosure allows is narrower than a verdict and more useful than a headline. It establishes that capacity rose twice as fast as last year, that unit cost fell in the middle of the build, and that repeat purchase rather than new buyers is carrying the payback case. It does not establish availability, because availability by site is not a published metric in this market, and no amount of reading the release will produce it. That gap is where the monitoring work has to sit, and it is also where a supplier finds out whether the extra space reached the shelf that carries its own listing.
FAQ
Does 20 percent more fulfilment space mean 20 percent more stock on the shelf?
No. Space is capacity to hold and move stock, not stock itself. What converts it into availability is the replenishment schedule inside each catchment, and that schedule is not published.
Why did margin fall if cost per shipment also fell?
Because the two measure different things. Cost per shipment is a unit measure inside a dense catchment, while margin carries the threshold subsidy, the extra square metres, energy prices and temporary recruitment at group level.
What does items per buyer have to do with stockouts?
It is the demand side of the same equation. A shopper who buys more often gives the network more chances to fail, so frequency gains raise the standard that replenishment has to meet.
How should a supplier read a substitution path after a stockout?
As evidence about the category rather than the brand. If shoppers stay inside the category and pick a cheaper listing, the demand survived and the listing did not; if they leave, the category lost the sale.
