Your Upsells Raise AOV and Can Still Lose Money

Your Upsells Raise AOV and Can Still Lose Money

Attach rate is up. AOV is up. The upsell report is green across the board — and margin quietly went down.

How? Because attachment is not contribution. An upsell can add $18 to the basket while adding $4 (or less) in margin, once you count what it actually costs.

Where the margin goes

Three places AOV-focused upselling leaks:

1. Cheap add-ons with real handling cost. A $12 add-on at 50% margin looks like $6 of profit — until pick-pack, extra packaging, and a higher return probability leave $2 or less.
2. Heavy and variant-heavy SKUs. Oversized or fragile attachments carry shipping and damage rates a pure-digital "recommended for you" model never sees.
3. Returns on stretched baskets. Shoppers pulled up by a threshold or bundle return the filler. AOV counted on day one; the refund lands on day twenty.

Average checkout attach commonly runs in the high single digits, with top operators much higher — but the headline numbers are gross. Net of handling and returns, many "winning" slots are break-even or worse.

The rules that actually govern slots

• Keep the upsell a fraction of the basket. A common working rule: the add-on stays well under 40% of the subtotal, ideally under a quarter. Past that you're not nudging — you're re-planning their purchase, and returns follow.
• Cheap SKUs need volume economics. A low-margin add-on only works at high take rates and low handling. Price the pick-pack in before celebrating the attach.
• Curate mercilessly. Small catalogs should be hand-curated; the algorithm's third suggestion is usually where the margin deaths live.

How to measure it

Not take rate. Not attached AOV.

Profit per visitor to the slot = (treated-slot contribution − control, net of refunds and extra handling) ÷ eligible visitors.

Run a holdout: hide the slot for a slice of traffic and compare. The slots that survive are the ones that create margin, not baskets. Post-purchase slots deserve the same discipline: report incremental margin per eligible order, net of the upsell's own refunds — never per acceptor.

What wins

• Complementary SKUs with high margin and low return rates (care products for devices, refills for hardware).
• Threshold plays where the upsell is the reason to cross — the add-on earns the free shipping it enables.
• One job per slot: drawer shows threshold progress, checkout stays clean, post-purchase takes one shot. Never three offers fighting over the same basket.

The takeaway

AOV is what the shopper pays. Margin is what you keep. They're only the same number if you never discount, never mishandle, and never see a return.

Score every recommendation slot on profit per visitor — and kill the attachments that raise the basket while lowering what the basket is worth.