Ask a merchandising director about how they manage out-of-stock inventory in their Shop the Look module in production. They will probably shudder. A jacket sells out in the UK on a Tuesday afternoon and is still sitting in full supply in the US warehouse. By Wednesday morning, the team still hasn’t finished updating stock availability in each market, and hundreds of orders go unfulfilled. Multiply that across fifty thousand SKUs and a dozen markets, the task quickly becomes unmanageable.

This challenge has existed for many years and made it challenging to scale product recommendations. It is now a problem that AI can now solve, by treating stock the way a merchandiser would in real time: a live signal that changes hour to hour.

Managing stock availability across multiple markets can be a major challenge. Keeping product information up to date in real time often takes significant manual effort.

A head-to-toe image is only as good as the availability of its items. At the scale of a single capsule collection, a merchandiser can catch small changes to availability. At the scale of a full catalogue, refreshed daily and sold across a dozen currencies and warehouses, often it’s challenging to keep up with the changes. The gap between “in stock when we tagged it” and “in stock when the shopper actually clicks” is where Shop the Look modules built on manual tagging fall over. They do not know what is actually available to sell right now.ave not known the cause. They describe Shop the Look as a feature they “had to keep an eye on.” That is the cost of running discovery on visual similarity alone.

What this costs each team

For the merchandising team, this means one of two choices: swap out-of-stock products by hand, market by market, every time a buy sheet updates, or let the shopper click through to a dead product and lose the sale. It is the challenge merchandising teams often bring to us: how can we automate the tagging process for head-to-toe content, and make sure the experience stays seamless when a product goes out of stock? 

For the trading and merchandising director, the stakes sit across the full catalogue, not just the hero shots. Why build a solution that only works for the top hundred SKUs, when you could maximise sell-through across everything you carry?

For the VP or director of ecommerce, the AOV and UPT lift the AI Shop the Look module was signed off to deliver has to hold up over time. A dashboard showing a 20 percent uplift at launch means little if half the looks are quietly broken by the following quarter.

For the brand team, availability cannot be the only rule. A replacement product has to be one the brand would actually put in that look: the right category, the right price tier, the right aesthetic. A tailored blazer swapped for whatever jacket happens to be in stock can lead to poor brand consistency.

And for the head of digital, there is often limited capacity to take on dev tickets every time a market’s stock position shifts. Fallback logic that needs engineering time to maintain, often means it does not get maintained.

How the AI Shop the Look solves for stock availability

This is where StyleGraph, the AI engine behind AI Shop the Look, does more than most teams expect from a styling tool. StyleGraph reads the catalogue the way a merchandiser does: category, price tier, colour, material and brand profile, applied alongside the merchandising rules your team sets, such as excluding sale lines, holding a minimum stock depth, or prioritising new arrivals. Stock is treated as a live input, synced from your catalogue feed and reprocessed overnight, so the system always knows what is actually sellable.

When a product in a look goes out of stock, StyleGraph does not remove it and leave a gap. It substitutes the next best match the brand would choose itself, applying the same style logic and merchandising rules as everywhere else in the catalogue. The shopper never sees a dead product or a broken look. For brands like Primark, this means hundreds of Shop the Look options per SKU - all on brand.

StyleGraph in action on Primark.com, managing next-best recommendations, stock and size availability, all in real time

Built to run the same way in every market

The same logic is what makes this work across locales. Ecomm and product teams running content across multiple regions and accounts tell us similar things: they need a platform that can handle however many locales and accounts they run, without a parallel content operation for each one. The AI Shop the Look module applies the same StyleGraph intelligence, the same merchandising rules and the same fallback logic to every locale’s own catalogue feed and stock position. A product that is sold out in the UK and available in DACH resolves correctly in both, automatically.

What it looks like in production

We see this most clearly in accounts running a full catalogue at real scale. Primark has around 70 percent of a 16,000-product range online, moving fast enough that a manual or pixel-matched fallback process would be obsolete within days of going live. Shop The Look holds because the fallback logic runs on the same intelligence as the matching itself, live in as little as two to three weeks from catalogue integration, and rechecked every single night.

The bar has moved

In summary, building Shop the Look at scale is an infrastructure and resource challenge, and it is exactly the one the AI Shop the Look module and StyleGraph were built to solve.

Check out AI Shop The Look in action here.

Sources and further reading

• How Primark Is Reinventing Product Discovery for 2026, StoryStream

• StoryStream’s AI Shop the Look module

• Most Shop the Look tools can’t scale past your top SKUs, here’s why, StoryStream

• The New Rules of Fashion Discovery, StoryStream eBook