Order Velocity: Catching the Bracketing Your Cart Reports Miss

Most bracketing analysis looks at what was ordered. The sharper signal is when. A customer who places three separate orders in one afternoon, each containing one size of the same dress, is bracketing just as surely as the customer who puts all three sizes in one cart. The single-cart bracketer is easy to spot. The split-order bracketer is invisible to every report that counts multi-unit lines per order, and they are more common than most brands think.

Why split the orders? Sometimes it is innocent: the shopper found the item again later, or a discount code arrived between orders. Sometimes it is deliberate, because shoppers have learned that multi-size orders get flagged. Either way, the economics are identical. Three orders, three shipments, two returns. The detection system that only watches the cart misses it completely.

Building the velocity signal

Order velocity tracking links orders by customer and looks for clusters: multiple orders within a short window containing the same or closely related SKUs in different sizes. The window matters. Same-day clusters are the strongest signal. Orders spread across a week are usually just repeat shopping. The sweet spot for detection is a rolling 24 to 48 hour window, which catches the deliberate splitters without flagging normal behavior.

The SKU relationship is what separates bracketing from enthusiasm. Two orders of different products in one day is a good customer. Two orders of the same product in adjacent sizes is a fitting room. Your matching logic should score size adjacency, not just product identity: same style in a medium and a large scores higher than two unrelated items, and the same style in the same size twice is usually a gift or a reorder, not bracketing.

What split-order bracketing costs

The cost is worse than single-cart bracketing, because you pay full outbound shipping on every order and you cannot consolidate the return. A three-size bracket in one cart costs you one outbound shipment and one return shipment. The same bracket split across three orders costs you three outbound shipments and up to three returns. The margin math that barely works for consolidated bracketing falls apart completely for split orders.

There is also a data cost. Your bracketing rate, the metric the whole program optimizes, is understated when split orders are not linked. Brands that add velocity-based linking routinely find their true bracketing rate is 15 to 30 percent higher than the cart-based number. That is not a rounding error. It changes which categories get fit investment and which customers get flagged.

Responding without punishing

Detection is not the same as accusation. The right response to a velocity flag is the same toolkit you use for cart-based bracketing: fit guidance on the product page, exchange-first returns, and keep-rate coaching. The velocity signal just tells you who needs it. What you should not do is cancel or hold split orders, which punishes shoppers for a checkout pattern and creates support tickets that cost more than the shipping you saved.

The one exception is clear abuse: velocity clustering combined with damage claims or refund-only returns. That combination is the refund-abuse profile wearing a different hat, and it belongs in the abuse workflow, not the bracketing workflow. Segment first, then act. The velocity flag is a sorting mechanism, not a verdict.

Getting started

You do not need new infrastructure to build this. Your order data already has customer IDs, timestamps, SKUs, and sizes. A weekly query that groups orders by customer within 48-hour windows and scores size adjacency will surface the pattern in an afternoon. Run it against last quarter's data first, before changing anything. The brands that do this usually find a cohort of customers whose bracketing was completely invisible, and a returns problem that is bigger, and more fixable, than they thought.