Complete the Look vs Frequently Bought Together: Which Drives More Fashion Revenue?

Here is the thing no founder actually loses sleep over: whether a shopper added a second item because of a "Frequently Bought Together" widget or a "Complete the Look" one. You do not care how the basket got bigger. You care that it did. Which widget did the work is our problem to solve.

So this guide is not a turf war between two widgets. It is a way to answer the only question that matters for your store: which of these will earn you more, given the catalog you have and the order data you have. Both can grow average order value. They just grow it on different kinds of stores, and picking the wrong one means leaving money on the table without ever knowing it.

In short: Frequently Bought Together learns from purchase history, so it pays off on large catalogs with lots of orders and barely functions on new or small fashion stores. Complete the Look works from style and coordination, so it earns from day one and tends to win in apparel. This post is about matching the tool to your store so you make the most money, whichever one that turns out to be.

What's in this guide

Start with the goal, then pick the tool
How Frequently Bought Together makes you money
How Complete the Look makes you money
Why the answer is usually different for fashion
The catch for new and small stores
Which one fits your store
The bottom line

Start with the goal, then pick the tool

The goal is more revenue per order. That is the whole game. A shopper landed on one product, and you want them to leave with more than one, without you cutting the price to make it happen.

Recommendations are a proven way to do that. Salesforce looked at more than 150 million shoppers and found that visits where someone clicked a recommendation were only 7% of all visits but drove 26% of revenue. McKinsey puts the wider payoff of personalization at a 10 to 15% revenue lift for companies that do it well.

You have probably also seen the line that recommendations drive 35% of everything Amazon sells. It traces to a real McKinsey article, but it is from 2013, the workings were never shown, and it is genuinely hard to measure, so treat it as the most-quoted number in the category and move on.

The point stands: showing shoppers more works. The money question is which way of showing them earns the most on your particular store. So here is what each approach is actually good at.

How Frequently Bought Together makes you money

Frequently Bought Together is the widget Amazon made famous, and it runs on a method Amazon engineers published in 2003 called item-to-item collaborative filtering. It looks at every order you have ever taken, finds the products that keep landing in the same cart, and surfaces them. Add a phone and the case and screen protector appear, because thousands of people before you bought all three together.

When it works, it works because the pattern is real. People genuinely do buy those items together, so suggesting them catches a purchase that was likely anyway and pulls it into the same order. On a big catalog with deep order history and products that pair for practical reasons, that is steady, reliable extra revenue.

The reason it sometimes makes you nothing is that it can only act on what it has already seen sell. A product with no purchase history gives it nothing to work with. A pairing nobody has bought yet stays hidden, even when the two pieces obviously belong together.

How Complete the Look makes you money

Complete the Look earns in a different way. Rather than mining what sold together, it assembles a coordinated outfit around the product on screen: the shirt the shopper is viewing, then the trousers, shoes, and bag that finish the look. It makes you money by turning a one-item visit into an outfit-sized order.

Because it runs on style and coordination, it can suggest a pairing that has never appeared in a single past order, as long as the pieces genuinely work together. It can style a product the same afternoon you upload it. And the suggestion reads like a stylist did it, which is exactly the nudge that gets a fashion shopper to add the trousers and the shoes, not just admire the shirt.

The hard part is doing it well at scale. Hand-styling every product gives lovely results, but a Forrester study commissioned for one outfitting platform found merchandisers can realistically style less than 10% of a typical catalog by hand. Closing that gap is the whole reason AI-built, brand-approved outfitting exists: a curator's eye applied to the entire catalog.

Why the answer is usually different for fashion

For most stores the two are worth testing against each other. For apparel specifically, there is good evidence the styled approach earns more, and it comes from researchers with nothing to sell you.

When a team at ASOS built an outfit recommender, they found that co-purchase is not a strong signal of compatibility, because items bought in the same order are usually not bought to be worn together. So ASOS trained their system on stylist-curated outfits instead, and those style-aware looks were approved 21% more often for womenswear and 34% more often for menswear than a version that matched product types but ignored style. A separate group from Pinterest, Stanford, and UC San Diego reached the same place, noting that visual compatibility is different from visual similarity and a system has to learn coordination from data. Trained that way, their model matched human stylists.

What this means for your revenue is simple. In apparel, the thing that makes someone buy the second and third piece is whether the items look right together, and a co-purchase engine has no way of seeing that. It knows what sold together. It does not know what looks good together, and in fashion that gap is money.

There is a returns angle too. Clothing is the most-returned online category at around 25%, and a shopper who can see how a piece sits inside a complete outfit buys with more confidence about what is showing up at their door. A bigger order that also comes back less is the version of AOV growth you actually want.

The catch for new and small stores

This is where the choice gets decided for a lot of brands, and it is worth being blunt about. Frequently Bought Together needs purchase history to function. If you do not have it, the widget is there but earning close to nothing.

Collaborative filtering, the engine underneath it, cannot recommend a product it has no data on. The literature calls this the cold-start problem, and a content or attribute-based approach sidesteps it by working from the features of the item instead of from past orders. Researchers studying fashion keep flagging cold-start as a recurring wall, and they note it bites hardest in fast-fashion contexts with short product cycles and constant new stock.

In practice: a newer brand with a few hundred SKUs, a catalog that turns over every season, or simply not many orders yet has almost nothing for Frequently Bought Together to mine. There is no co-purchase data on a dress that launched on Tuesday. A styling-based Complete the Look engine reads the look and attributes of the pieces and builds a relevant outfit anyway, so it is earning on day one while the other widget waits for data that may take months to arrive.

Which one fits your store

Here is the honest decision, framed around what makes you the most money rather than which tool is cleverer.

Your store	The one that earns more	Why
Large catalog, deep order history	Frequently Bought Together works well	Enough data for reliable co-purchase patterns
Products that pair for function (not style)	Frequently Bought Together	Co-purchase captures functional pairings accurately
Apparel, where coordination drives the basket	Complete the Look	Style is what makes shoppers add the next piece
New, small, or fast-turnover catalog	Complete the Look	Works from day one, no purchase history needed
You want fewer returns alongside higher AOV	Complete the Look	Seeing the full outfit builds buying confidence

Plenty of stores end up running both, and that is fine. Functional pairings can sit in a co-purchase widget while the outfit-building happens through Complete the Look. The mistake is defaulting to Frequently Bought Together because it is familiar, on a fashion catalog that does not yet have the data to make it pay.

The bottom line

You do not care which widget grew the order, and you should not have to. What you care about is picking the one that grows it most on your store. Frequently Bought Together earns on large catalogs with deep data and functional pairings. Complete the Look earns on apparel, on small and new catalogs, and anywhere coordination is what moves a shopper to buy the whole outfit. For a fashion brand, especially a growing one, that is usually where the extra revenue is sitting.

References

Salesforce, Personalized Product Recommendations Drive Just 7% of Visits but 26% of Revenue. Recommendation-clickers: 7% of visits, 26% of revenue.
McKinsey, The value of getting personalization right or wrong is multiplying. 10 to 15% revenue lift from personalization.
McKinsey, How retailers can keep up with consumers. Origin of the Amazon 35% figure (2013).
ASOS, Fashion Outfit Generation for E-commerce. Co-purchase is a weak compatibility signal; curated outfits approved 21%/34% more often.
Kang et al., Complete the Look: Scene-based Complementary Product Recommendation, CVPR 2019. Visual compatibility differs from visual similarity.
Forrester Consulting / Business Wire, Total Economic Impact study of an AI outfitting platform. Merchandisers manually style under 10% of catalog.
Statista, Most returned online product categories. Clothing most-returned at ~25%.
Wikipedia, Cold start (recommender systems). Content-based filtering is less prone to the new-item problem.
International Journal of Data Science and Analytics, Deep learning for new fashion product demand prediction. Cold-start is sharpest in fast-fashion contexts.