The Ecomm Analyst

Most bundling decisions start with someone in marketing or merchandising looking at the catalog and saying “these would go well together.” Sometimes they’re right. Often they’re not, and the bundle sits on the site doing nothing because the underlying assumption (that those two products belong together) was based on intuition rather than data.

The cleaner way to do this is to look at what customers are already buying together. Pull up an order history for the last six months and the answers are usually right there. Some pairs come up over and over, some pairs are almost always purchased separately, and a few pairs are bought together often enough to suggest there’s a real bundle hiding in plain sight.

Until recently I did this in spreadsheets, which is exactly as miserable as it sounds. Export orders, parse line items, write a co-occurrence matrix in Sheets, sort by frequency, try to remember to update it next month. The data was always stale and the analysis was always shallow.

ThoughtMetric just shipped a Product Bundles view that does this directly. It surfaces which products are bought together, how often, and (the part I care about most) how much revenue those combinations generate. That last column matters because frequency on its own can mislead. Two cheap products bought together 200 times a month might be less interesting than two premium products bought together 40 times a month, depending on margins.

A few ways I actually use this data:

Formalize the natural bundles. When two products are already purchased together at high frequency, putting them into an official bundle with a small discount can lift the attach rate further. The risk is that you’re discounting revenue you’d already be earning, so the math has to work. If bundling a $40 and a $25 product at $60 lifts the attach rate enough to offset the $5 discount on existing pairings, it’s a win.

Surface the almost-bundles. This is the more interesting one. Products that get bought together 15-20% of the time are the candidates with the most room to grow. Those pairs aren’t habits yet, they’re hints. An official bundle, a cross-sell on the PDP, or a subject line in email can move that attach rate meaningfully higher because there’s actual headroom.

Find product gaps. When a category has no clear co-purchase partner, that sometimes means there should be one. A skincare brand I work with noticed their best-selling cleanser had no consistent companion product in the cart. The data wasn’t telling them to bundle, it was telling them they were missing a SKU.

Identify which bundles to put in email. Not every co-purchase combination deserves homepage real estate. But every meaningful one is worth a flow. A “people who bought X also bought Y” email triggered after the first purchase, populated from real co-purchase data rather than someone’s guess, consistently outperforms the generic upsell.

A note on what not to do. Don’t just rank by frequency and bundle the top five pairs. The top of the list is often dominated by accessories or low-AOV combinations that are already happening organically. The interesting bundles are usually further down the list, where there’s a real revenue opportunity that hasn’t been packaged yet.

Bundling has always been one of the highest-leverage AOV moves available. The tooling is finally catching up to the idea that customer behavior, not the merchandiser’s intuition, should drive what gets packaged together.