Katana’s study of 375 product brands puts the median annual stockout loss at roughly $21,000, with the hardest-hit ten percent losing over $268,000. Every brand measured already ran inventory software, which makes $21,000 a floor, not a sales pitch.
Every one of the 375 brands in this study was already running inventory software when it lost the money. That is the finding, and it is not the one the headline number is pointing at.
A brand runs out of its bestseller about 14 times a year. Each gap lasts roughly two days. Nobody calls a meeting about two days. But stack those gaps across twelve months, and the typical brand has spent about thirty-four days, more than a month, with its single most important product unavailable to buy.
That figure comes from Katana’s study of 375 product brands, published this summer. It is a genuinely useful piece of research and work.
The study modeled 12 months of inventory and sales data for 375 US product brands, all of which were Katana customers, selling across Shopify, Amazon, wholesale channels, and their own storefronts between June 2025 and May 2026. For each brand, it identified the top-selling products, found the windows when those products were unavailable, and estimated how much they would have sold at their prior sales rate during those windows.
Three things about that design matter more than the output.
First, the losses are modeled rather than observed. Katana says so plainly, and repeats the word estimated throughout. A stockout leaves no transaction record behind, so there is nothing to measure directly. That is not a flaw, it is the only way to do this, but it means the figures are a well-reasoned projection rather than a bank statement.
Second, they excluded seasonal products, the limited runs and holiday items that are supposed to sell out. Counting those as losses would have inflated the total, and leaving them out is the more honest call. The $27.5 million total across all 375 brands reflects that adjustment.
Third, and this is the one nobody will pick up: every brand in the sample was already paying for inventory software. These are not spreadsheet businesses. They had connected channels, live stock counts, and a system telling them what was on hand. They lost the money anyway. I got into the adjacent version of this with Katana’s COO on the podcast earlier this year, and the framing then was mostly about brands running blind. This data says something more uncomfortable than that.
The median brand lost about $21,000, but the distribution is so skewed that the median describes almost nobody. The top quartile lost around $82,900 each. The hardest-hit tenth lost over $268,000. Katana’s own read is that 69% of all measured losses fall on just 10% of brands.
The main driver of the spread is volume. A two-day gap costs a $500K brand a few hundred dollars and costs a $20M brand five figures, because more revenue runs through each best seller per day. That part is intuitive.
In practical terms, you should stop using the median as a benchmark and calculate your own. Take your top five products, find your average daily revenue for each over the last ninety days, and multiply by the number of days each one showed as unavailable. That is your number. For many readers, it will come in well under $21,000, and the honest conclusion is that this is a problem worth twenty minutes a quarter rather than a platform migration. For others it will be a genuinely alarming figure, and those readers should keep reading.
Two brands at identical revenue can have wildly different stockout exposure depending on how concentrated their catalog is. The report’s category examples make this clearer than any general principle would.
The apparel number is low because a shopper who cannot get the medium in olive will often take the large, or the navy. Demand spreads across variants and substitution absorbs most of the loss. Beauty has no such cushion. When a hero product goes dark there is no adjacent SKU to catch the customer, so a stockout problem concentrated in three or four items reaches a quarter of a million dollars.
Food and beverage sits in between for a reason worth naming separately. The loss there is not really the missed order, it is the interrupted habit. A subscriber who cannot reorder their coffee this week reorders it somewhere else, and that damage does not show up in a stockout model at all. Katana’s methodology cannot capture it, which means the food and beverage figure is probably conservative.
So before you benchmark yourself against a revenue peer, ask what share of your revenue sits in your top three SKUs. If it is over forty percent, you are structurally closer to the beauty profile regardless of what you sell.
Stockouts are not randomly distributed across a catalog; well over half of them hit products that had already run out earlier that same year. The same items run out, get restocked, and run out again, roughly fourteen times annually for a typical brand’s top seller.
That single pattern is the most actionable thing in the report, and it is a diagnosis rather than a statistic. If a SKU repeatedly runs out, the reorder point for that SKU is set too low relative to how fast it actually sells. Not the forecasting model, not the demand signal, the reorder point. It is a number someone typed in once, probably eighteen months ago, based on a sales velocity the product has since outgrown.
Here is the part that explains why nobody fixes it. More than half of these stockouts get resolved within a day, through a backorder or a quick production run, and the order still ships. The customer is served. Nothing appears in the accounting. And yet Katana estimates those recoverable stockouts account for about sixty five percent of total lost sales, on products the business could have supplied during the window the system said it could not.
This is the pattern I watched play out constantly during my years as a Merchant Success Manager at Shopify. Brands between $500K and $2M would come to me convinced they had a forecasting problem and wanted to know which app to buy. In most cases they had four or five SKUs with stale reorder points and no recurring habit of reviewing them. The tool was never the constraint. The twenty minute review nobody owned was the constraint.
The forgiveness window that made brief stockouts survivable is closing, because AI assistants filter for availability before they recommend anything. On your own product page, a shopper who hits an out-of-stock button can pick something similar, sign up for a restock alert, or come back Thursday. Inside a chat interface there is no shelf to browse and no adjacent product to notice. A product the assistant reads as unavailable is a product the shopper never learns existed.
The channel is not hypothetical anymore. Adobe Analytics reported that AI-sourced traffic to US retail sites grew 393 percent year over year in the first quarter of 2026, and that this traffic converted 42 percent better than non-AI sources in March, a reversal from a year earlier when it converted considerably worse. Katana’s report leans on the same Adobe figures, and adds that analysts project AI-driven shopping could reach ten to twenty percent of US ecommerce by 2030.
The report also makes a point about Google’s Shopping Graph that is worth sitting with: listings get removed from Shopping when the stock status in a brand’s feed does not match what is on the site. That is not a ranking penalty you can optimize your way out of. It is a correctness check, and you either pass it or you are absent.
Now apply the eighteen month filter, because I try to run it on anything trending. Will availability accuracy still matter in eighteen months? Yes, and more than most of what gets sold as AI readiness, because it is infrastructure rather than a surface. Ad creative works on one channel. A correct stock number works on every assistant that exists and every one that has not launched yet. That durability is the same reason I keep pushing operators toward the operator-facing half of Shopify MCP rather than the readiness checklists.
The most important sentence in this report is on the methodology page: all 375 brands were Katana customers during the measurement window. That reframes the entire finding. The $21,000 median is not what brands lose due to a lack of inventory software. It is what they lose while already having it.
I want to be careful here because this cuts both ways, and both are true.
It means the report understates the case for tooling, not overstates it. A sample of spreadsheet-run businesses would almost certainly have posted worse numbers, and Katana could have run that comparison and chose not to. Publishing a study where your own customers still lose a median $21,000 a year takes some nerve, and it reads as more credible to me than a vendor benchmark showing customers with the problem solved.
But it also means the obvious conclusion, buy inventory software, does not follow from this data. These brands bought it. They still had a best seller dark for about a month of the year. Software gives you an accurate count and a place to set a reorder point. It does not decide what that reorder point should be, and it will not tell you that a number you entered last year has quietly gone stale. That judgment stays with a human, and the humans at $500K to $2M are the same people doing customer service and negotiating with the 3PL.
Which brings me to the claim I will actually defend: for most brands under $2M, the recoverable share of this loss is a process problem wearing a software costume. Fix the review habit first. If you buy the platform without building the habit, you will own a more expensive version of the same thirty four days.
The right first move depends entirely on your stage, and for a meaningful share of readers the right move is to do almost nothing. Run the calculation from earlier, look at the number, and let it decide the size of your response.
The repeat-offender review is the whole exercise, and it takes about ninety minutes the first time. Pull every SKU that hit zero more than once in the last twelve months. For each one, compare its reorder point against its actual daily sales rate multiplied by your real supplier lead time, not the lead time on the original quote. Most of the gaps close themselves once you write those two numbers side by side.
If you are also migrating off Stocky, which shuts down at the end of this month, the counting side of this now lives in Shopify POS and the admin. I walked through the full count and reconciliation workflow that replaces it, including how the old purchase order and supplier features map across.
A dedicated inventory platform earns its cost when the number of places your stock count can drift exceeds what one person can hold in their head, which in practice means three or more sales channels, more than one physical location, or production that consumes raw materials. Below that threshold, you are buying capability you will not use.
If you do cross it, the field is wider than one name. Shopify’s native purchase orders, transfers, and inventory reporting now cover the basics for single-location retail. Inventory Planner and Cogsy are built for forecasting and replenishment on finished goods. Prediko targets Shopify-native brands wanting demand planning without an ERP. Cin7 and Fulfil sit higher up, closer to real ERP territory with the implementation cost that implies. Craftybase is worth a look specifically for small makers tracking materials and true cost of goods.
Katana fits a narrower slot than its own marketing implies, and the narrowness is the point. It is built for businesses that make, assemble, or kit what they sell and also move it across several channels. If you are a pure reseller with three channels and no production, several of the options above will do the job for less. If you are cutting, blending, or assembling and you need raw materials, work in progress, and finished goods reconciled against multi-channel demand, that combination is genuinely harder to find and Katana is a legitimate shortlist candidate. They also shipped an MCP connection in July that lets Claude or ChatGPT query live inventory, which is early but pointed in a direction I think holds up.
For the wider question of keeping counts aligned once you are selling in several places at once, I have covered the common multichannel inventory failure points and how to close them separately.
Whatever you land on, run the repeat-offender review first. If it turns out five stale reorder points were costing you most of the money, you will have solved the expensive part of this for the price of one focused afternoon, and you will evaluate platforms from a much stronger position.
The median product brand loses roughly $21,000 a year to stockouts on its best sellers, according to Katana’s study of 375 US brands measured between June 2025 and May 2026. That median hides an extreme spread. The top quartile lost about $82,900 and the hardest hit tenth lost more than $268,000, with sixty nine percent of all measured losses concentrated in just ten percent of brands. The figures are modeled from pre-stockout sales velocity rather than pulled from accounting records, and seasonal products were excluded. Calculate your own number instead of using the median: daily revenue per top SKU multiplied by days unavailable.
Repeat stockouts almost always mean the reorder point on that SKU is set too low for how fast the product currently sells. Well over half of all stockouts in Katana’s data hit products that had already run out earlier in the same year, and a typical brand runs out of a top seller about fourteen times annually. The usual cause is a reorder threshold entered once and never revisited while the product’s sales velocity grew past it. The fix is a recurring review: list every SKU that hit zero more than once in twelve months, then reset its reorder point against real recent demand and your actual supplier lead time.
It matters more in 2026 than it used to, because AI shopping assistants filter for availability before they recommend anything. Historically a brief stockout was absorbed: the customer picked something else, waited, or accepted a backorder, and the sale survived. Katana estimates that stockouts resolved within a day still account for about sixty five percent of total lost sales. In an AI assistant there is no adjacent product to notice and no restock alert to sign up for, so a product reading as unavailable simply does not appear in the buyer’s options. Google also removes listings from Shopping when feed stock status disagrees with the site.
Not necessarily, and this study is a useful argument against assuming so. All 375 brands in Katana’s sample were already running inventory software when they lost the money, which means tooling alone does not close the gap. Software gives you an accurate count and somewhere to store a reorder point; it does not decide what that reorder point should be or notice when it goes stale. For most brands under $2M, the recoverable portion of stockout loss is a process problem: a monthly review nobody owns. A platform starts earning its cost at three or more channels, multiple locations, or production that consumes raw materials.
The right alternative depends on whether you manufacture. For finished goods forecasting and replenishment, Inventory Planner, Cogsy, and Prediko are the common Shopify choices. For single-location retail, Shopify’s native purchase orders, transfers, and inventory reporting now cover the basics following the Stocky shutdown. Cin7 and Fulfil sit closer to full ERP with matching implementation cost. Craftybase suits small makers tracking materials and true cost of goods. Katana’s genuine differentiator is combining production planning with multi-channel stock sync, so it fits brands that make or assemble products and sell across several channels, and is oversized for pure resellers.
