How SlumberCloud built a custom advertiser network

Slumber Cloud makes cooling sleep products for customers whose nights are disrupted by heat. But not every shopper is solving the same problem in the same way. A customer buying a cooling pillow is signaling something different than a customer buying a throw blanket - and those signals matter if you want recommendations to feel useful rather than generic.

That made Slumber Cloud a strong fit for Garner. Instead of treating every order as interchangeable, Garner's recommendation algorithm automatically uses the product in cart to shape the partner offers shown after purchase, making the experience tailored to what the customer actually bought without the Slumber Cloud team having to build or manage that logic themselves.

Generic recommendation inventory leaves money on the table

Most post-purchase ad networks optimize for fill rate, not fit. The same handful of offers (typically Capital One, Hulu+, and HelloFresh) get shown after wildly different purchases, with little regard for what the customer actually bought. For Slumber Cloud, that would have flattened the nuance in their catalog and made the experience feel more like rented ad space rather than a thoughtful extension of their brand. But most importantly, taking a non-custom approach would have left significant money on the table.

Making recommendations purchase-aware

A shopper who just purchased a cooling pillow does not need the same recommendations as someone who bought a throw blanket. One order might point toward a cooling wearable; while the another might perform better with a natural sleep supplement recommendation. Different SKUs have different gender splits, age ranges, and other demographic differences that make them more or less likely to convert on different products. Slumber Cloud needed a network that could adjust at the SKU level rather than forcing every customer through the same recommendation path.

Manual partnerships do not scale that granularly

Building that kind of partner mix by hand would have required sourcing, vetting, testing, and rotating dozens of brands continuously. The opportunity was real, but only if the network itself could do the discovery and optimization work automatically.

Garner's self learning recommendation algorithm

Garner's recommendation model started by learning the Slumber Cloud product catalog item by item, so each purchase could act as a signal. From there, it automatically identified non-competitive products that fit each customer context using proprietary engagement data collected from the post-purchase experience, plus network-wide visitor behavior which gives us a glimpse into how customers hop between brands. The network then tested more than 40 complementary brands and assembled a mix tailored specifically to Slumber Cloud's catalog and customers.

Product-level personalization

Because Garner's algorithm uses the product in cart as a signal, recommendations automatically change according to the purchase. Cooling pillow customers saw a different mix of offers than throw blanket customers, leading to a more relevant experience for shoppers and more profits for Slumber Cloud.

Automatic yield optimization

Garner didn't just find strong brand matches and leave them in place. As new engagement data came in, the model continued refining which recommendations performed best for each customer context and adjusted the mix automatically. Slumber Cloud didn't have to source new partners, swap offers, or monitor performance by hand. Yield continued improving on its own.

A custom advertiser channel

The same data that improved Slumber Cloud's post-purchase recommendations also revealed where the brand should show up as an advertiser. By looking at how shoppers engaged with different offers, Garner could identify the non-competitive niches where Slumber Cloud was most likely to convert. The result was more than a smarter recommendation layer. It was a highly tailored acquisition channel too.

The economics improved as the network learned. Slumber Cloud's incremental profit per order rose steadily across their first six months on Garner, climbing from 5 cents to 60 cents as partner selection improved and product-level personalization got sharper.

At the same time, Garner delivered 8x ROAS for Slumber Cloud as an advertiser channel. Instead of buying broad sleep traffic and hoping it converted, Slumber Cloud was appearing inside a partner network that had already been trained on which complementary brands and customer contexts performed best.

Key results

• 8x ROAS on Garner as an advertiser channel
• $0.60 of incremental profit per order by month six
• 40+ complementary brands tested across the network