Split-Order Bracketing: When Shoppers Place Two Orders to Dodge Multi-Size Flags

The order that looks like two customers

Your bracketing dashboard flags orders that contain two sizes of the same style. It shows a clean trend line, a manageable bracket rate, and a keep-rate score you can defend in a meeting. Then the warehouse notices something odd: two orders, placed eleven minutes apart, from the same address, each containing one size of the same dress. Neither order triggered a single flag. Together, they are textbook bracketing.

Split-order bracketing is the practice of dividing a multi-size order across two or more separate checkouts. The shopper gets the same try-on-at-home experience, but your detection logic sees two ordinary single-size orders. Every brand that flags bracketing at the order level, and never joins the data at the customer level, is undercounting its real bracket rate. In some apparel categories the gap between order-level and customer-level bracketing is ten points or more.

Why shoppers split

Some of it is deliberate evasion. Shoppers talk, and the advice circulating in deal forums is explicit: if a store warns you about ordering multiple sizes, place separate orders. A small number of habitual bracketers do this routinely, the same way serial returners rotate accounts. But most split-order bracketing is accidental. A shopper adds a medium, hesitates, checks out, then talks themselves into trying the large too and places a second order. Mobile checkout flows make this frictionless: saved payment details and one-tap checkout mean a second order takes fifteen seconds.

The accidental kind matters just as much as the deliberate kind, because it carries the same cost. Two shipments, two return labels, twice the warehouse handling, and a customer record that says this person is a well-behaved single-size buyer. Your keep-rate scores are wrong for both groups, and your fit interventions target the wrong people.

The patterns that expose it

Split orders leave fingerprints that are easy to spot once you join on the customer instead of the order. The strongest signal is time clustering: two orders for the same style in different sizes within 24 hours, shipping to the same address. A looser but still useful signal is the payment token: same card or wallet, same style family, different sizes, inside a week. Email matching catches the obvious cases; device and payment-token matching catches the ones where the shopper used guest checkout for one order and an account for the other.

The second pattern is return coupling. Split orders return together. If two orders from the same address both generate return labels within days of each other, and one size comes back from each, you are looking at a single bracketing decision that happened to use two checkouts. Returns data is actually cleaner than order data for this, because the shopper's intent is fully revealed by what they send back.

What to do about it

Start with measurement. Recompute your bracket rate at the customer-week level: any customer who ordered two sizes of the same style within seven days counts as a bracketer, regardless of how many orders it took. That number is your real bracket rate, and it is the one your interventions should move. The gap between the order-level number and the customer-level number tells you how much bracketing is hiding in plain sight.

For intervention, the same playbook applies, just keyed to the customer instead of the order. Fit guidance at the second checkout, not the first: when a returning customer starts a second order containing a style they already ordered in another size, that is the moment for a size nudge. Post-purchase, combine the shipments when you can. Two parcels to the same address on the same day is a cost you chose, not a cost the shopper forced on you.

The deeper fix is identity resolution, which sounds expensive and mostly is not. You do not need a customer data platform to join orders on email plus shipping address. Most brands can build the split-order view with a weekly query against their order export. The brands that do it discover the same thing every time: their bracketing problem was bigger than the dashboard said, and their best customers were not who they thought they were.