97% of Orders Are Single-Category: Why Your Cross-Sell Flows Are Built Wrong

Think about the last time you did a proper grocery shop. You walked the aisles you came for. If you were there for dinner, you hit the fresh section, the meat, maybe the pasta. You did not detour into the camping gear, the stationery, and the garden hose on the same trip. Same store, same trolley, completely different missions on different days.

Your customers shop your store the exact same way. And almost nobody builds their post-purchase flows like they know it.

I had a look at this properly over the last few months because the data underneath it keeps surprising people, including me. When you tag a whole order by its dominant category and then check what else is actually in that basket, the answer is "barely anything else". A brand sorting orders this way will often find that 97 to 99% of the revenue inside a category order comes from that one category. Buy mostly homewares, you bought homewares. Buy mostly apparel, you bought apparel. The cross-category bleed is tiny.

That single fact breaks the way most brands run cross-sell. So let's pull a typical flow apart and rebuild it.

What the basket data actually says

Here's the thing - this isn't about whether people ever buy across categories. They do, eventually. It's about timing.

When a furniture brand looked at this, they found people bought by room. In the market for the living room, they bought living room things. Nobody put a bedroom nightstand and a living room couch in the same order. The categories were real, separate missions, and they happened on separate days.

I see the same shape on ecommerce accounts again and again. A customer comes in for one job to be done. They complete that job. The order is clean. The "while you're here, grab these other four things from across our catalogue" pitch sounds smart, but the basket data says the buyer's brain isn't in that mode yet.

To put this in perspective: if your category orders are 97% pure, then a flow that fans out across your whole range is fighting the customer's actual behaviour to chase the 3% that was going to happen anyway. That's a lot of effort spent rowing upstream.

The flow most brands are running

Picture the standard post-purchase cross-sell flow in Klaviyo. Someone buys a coffee grinder. Two days later the email lands, and it pitches everything: more grinders, the espresso machine, the milk frother, the descaler, the branded mug, the subscription beans. Six categories, one email, all at once.

It feels efficient. One flow, one send, full catalogue exposure. But look at what it's asking the buyer to do.

It's asking them to context-switch six times in one scroll. It's treating "bought a grinder" as a signal that they're ready to buy anything, when the data says it's a signal they're in the market for one thing - the coffee setup - and probably not the unrelated category three aisles over.

So the email does what scattered emails do. Low click rate. The odd descaler sale that would've happened anyway. And a quiet conclusion in the back of your mind that "cross-sell just doesn't work for us". It works. The build is wrong.

A few specific problems with the everything-at-once approach:

• It splits attention instead of pointing it. Every extra category you add lowers the odds any single one gets clicked.
• It ignores the natural next mission. Some categories genuinely lead to a second mission later. Pitching all of them flattens that signal.
• It measures itself badly. When the flow underperforms, brands blame the offer or the discount, not the structure.

Rebuilding it: one new category at a time

What I'd propose is the opposite of the catalogue blast. Move people into one new category at a time, in the order their behaviour suggests, and sequence it rather than dumping it.

Volume first, behaviour second. Here's what I mean.

Step one - finish the mission they're on. If they bought the grinder, the immediate post-purchase moment is for the rest of that same job. The beans, the filters, the thing that makes the grinder better. Same category, same headspace. This is the highest-converting cross-sell you've got because you're not asking them to switch missions at all.

Step two - find the real second mission. This is the part that takes a bit of digging. Pull your returning-customer data and ask: for people whose first order was category A, what category does the second order most often land in? Not in the same basket - on the next visit, weeks or months later. You're looking for the natural path. Maybe coffee buyers come back for tea gear. Maybe homewares buyers come back for kitchen.

Step three - build a flow per path, not per product. Once you can see that first-order-in-A tends to lead to a second order in B, you set the segmentation up to match. Buy from category A, and after the first mission's wrapped, you get gently pointed toward B and only B. Buy from category C, you go down a different branch entirely toward D. You're not inventing the journey. You're building the rails under a journey that's already happening organically.

The behavioural layer sits on top of the volume layer. First you let the data tell you which second category is real for each entry point. Then you sequence the move into it, one step at a time, instead of presenting the whole map at once.

A worked example, with invented but believable numbers so you can see the shape: say a kitchenware brand's "everything" cross-sell email sits at a 1.2% click rate and a thin attached revenue number. Rebuild it so first-order grinder buyers get a beans-and-accessories sequence, then a separate, later nudge toward the one adjacent category their cohort actually moves into. I'd expect the clicks to lift simply because each message has one job, and the repeat rate to climb because you're feeding what's already happening rather than scattering against it. I'm not promising you a fixed multiple - your categories and margins decide that - but the direction is the point.

When not to copy the nine-figure brands

Now, a caveat, because this is where I see smaller brands burn time.

The big nine-figure brands run incredibly granular category structures. Separate lines of business, separate pixels, separate funnels, full basket analysis across four or five categories. It's genuinely impressive, and it's tempting to copy it wholesale.

Don't, at least not all of it. The right question is the old one: is the juice worth the squeeze?

A brand doing serious monthly volume can justify building a bespoke funnel and a custom flow branch for a category that's only ever going to be 10% of revenue, because 10% of a very big number still pays for the work. If you're earlier on, that same 10% category might not earn back the hours it takes to build it its own segmented journey. You'd get more from sequencing your two biggest categories properly than from spreading thin across six.

So my take: borrow the principle, not the org chart. The principle - people buy by mission, so move them one category at a time - is true at every size. The five-category machine is a luxury you build into, not a starting point. Start with the first layer. Get your single biggest entry category sequencing cleanly into its one most-likely second mission. Then stack the next one on top once it's earning.

That's the trap with copying scaled brands generally. You see the finished system and try to install it whole, when what got them there was building one clean layer at a time and only adding the next when the last one paid off.

Where to start this week

If you want a quick gut check on your own setup, go and tag your last few months of orders by dominant category and look at how pure they actually are. If you're sitting up near that 97% mark like most brands do, then any flow that pitches across your whole catalogue at once is working against your own data.

Then pull the one thing most cross-sell flows never look at: for your biggest first-order category, where does the second order actually land? That answer is the entire blueprint for the flow you should be running.

If you'd rather have someone map that sequencing with you - the entry categories, the real second missions, and which paths are worth their own branch versus which aren't - that's a chunk of what a Signal/Noise Audit digs into. Either way, the move is the same: stop pitching the catalogue, and start walking people down the aisle they're already heading for.