Your upsell isn't a product problem - it's a placement problem.

The fix isn't more products.

Upselling and cross-selling drive 10-30% of ecommerce revenue, and 35% of Amazon's purchases come from its recommendation engine. Yet across the 1,200+ pages Levri has analysed, the average upsell converts at just 4.1% (cross-sells worse at 3.2%) - and roughly half of stores have no offer at all between add-to-cart and order confirmation. The other half put it in the wrong place: a "You may also like" grid at the foot of the product page, below the scroll, unrelated to the cart.

Most of the gap is placement and UX (user experience), not new products or a personalisation engine. The single highest-converting surface is the confirmation page - the subject of thank-you page optimisation.

The six patterns that move AOV (Avg. order value).

1. No offer between cart and confirmation - the entire mid-funnel left empty.
2. Irrelevant cross-sells - generic grids instead of cart-aware suggestions.
3. Too many choices - an eight-product block that creates decision fatigue.
4. The offer on the wrong page - placement drives 3-4x the difference.
5. No one-click accept - re-entering payment drops conversion 78%.
6. A shrunken mobile layout - a desktop grid squeezed onto a phone.

#1 - No offer between add-to-cart and confirmation.

The most common miss. Stores invest months in photography, copy, and checkout - then leave the mid-funnel empty. The median site has exactly one upsell surface, the product page, which is the lowest-converting placement in the funnel.

What we see on scans

• Zero recommendation blocks between cart and confirmation.
• The product-page "You may also like" as the only upsell surface.
• A thank-you page showing only an order summary.

What to ship

• Add a complementary-product block to the cart drawer or page.
• Add a one-click post-purchase offer on the confirmation page.
• Even a simple "Add [complementary item] for $X" in the cart beats a blank page.

Typical lift: +8% to +15% AOV from zero offers to one cart add-on plus one post-purchase. Impact: high - net-new revenue from existing traffic.

#2 - Irrelevant cross-sells in the cart.

52% of desktop sites show cart cross-sells that are generic or based only on what other customers bought - not what's in the cart. The problem isn't having recommendations; it's having generic ones. Someone buying running shoes wants socks or insoles, not a handbag.

What we see on scans

• Fixed grids of four or six products regardless of cart contents.
• "Bestsellers" blocks instead of cart-specific suggestions.
• Alternatives (competitors to the cart item) shown alongside complements.
• "You may also like" with no contextual framing.

What to ship

• Replace fixed grids with dynamic, cart-aware suggestions.
• Show only complements, never alternatives to what's already chosen.
• Use contextual labels ("Goes with your [product]").
• Two highly relevant items beat six generic ones; prioritise compatibility items (chargers, refills).

Typical lift: +12% to +22% cross-sell acceptance from generic to cart-relevant. Impact: high - relevance is the single biggest lever.

#3 - Too many choices in the offer block.

More options feel generous and convert worse. An eight-product cart block creates decision fatigue - the customer ignores it or scrolls past to the checkout button. The data converges on one complementary item, priced under 30-50% of cart value; pre-configured bundles beat build-your-own. This is the post-cart version of the three-CTA (Call to action) problem.

What we see on scans

• Eight or more products in the recommendation block.
• Build-your-own bundle builders with five-plus steps.
• Multiple competing offers on one page (upgrade + add-on + bundle).
• An upsell priced higher than the primary item.

What to ship

• Limit to one or two offers per page.
• Pre-configure bundles by outcome ("The starter kit"), not SKU (Stock Keeping Unit — an individual product code).
• Show the saving clearly - 10-20% on bundles is the sweet spot.
• Keep the upsell at 25-40% of the primary product; name bundles by benefit.

Typical lift: +4% to +8% on bundled offers, +15% to +20% AOV. Impact: medium-high - simplification drives acceptance and speed to checkout.

#4 - The offer sits on the wrong page.

Placement drives more variance than offer quality. The same recommendation converts at 3-5% on the product page, 5-8% in the cart, 8-15% at checkout, and 10-16% post-purchase. The confirmation page wins by commitment escalation: a $12 add to an existing order feels trivial; the same $12 on the product page is a separate decision.

What we see on scans

• Upsell only on the product page - the lowest-converting spot.
• No post-purchase offer page.
• A cart drawer with no recommendation content.
• A checkout with no add-on (not even gift-wrap or protection).

What to ship

• Add a post-purchase confirmation offer - the highest-converting placement (see thank-you page optimisation).
• Use the cart drawer as a cross-sell surface with one complement.
• Test a low-friction checkout add-on (gift wrap, protection - warranties convert 12-18%).
• Stack two or three placements across the funnel, not all in one spot.

Typical lift: +10% to +16% acceptance on post-purchase vs +3-5% on product page. Impact: high - moving the offer can triple acceptance with zero new content.

#5 - No one-click acceptance post-purchase.

Post-purchase offers live or die on one mechanic: accepting without re-entering payment. Requiring re-entry drops conversion ~78% - the difference between a 14% take rate and 3%. With one-click, stores see ~10-20% AOV lift.

What we see on scans

• A post-purchase offer that redirects to a new checkout.
• An "Add to order" button that asks for card details again.
• The offer buried below order summary and tracking.
• A multi-step accept flow.

What to ship

• One-click accept - payment on file, single tap to add.
• Place the offer above the order summary.
• Show one item only; carousels convert worse.
• Price under 30-50% of cart value; a gentle visible time limit ("Add within 15 minutes").

Typical lift: +9% to +20% AOV for one-click post-purchase. Impact: high - the make-or-break mechanic for post-purchase revenue.

#6 - Mobile gets a shrunken desktop layout.

Mobile is 65% of traffic and converts ~42% below desktop, and that gap widens on upsell surfaces. A four-product grid built for a 1440px monitor becomes an unreadable, slow mess on a 390px screen. Mobile-specific upsell layouts convert ~28% better than responsive shrink. (Speed matters here too - see page speed optimisation.)

What we see on scans

• Recommendation grids that scroll horizontally on mobile.
• Cross-sells pushed below the fold by the cart summary.
• Tiny "Add" tap targets.
• Product images too small to evaluate.

What to ship

• Design a mobile-first upsell - stack vertically, one product per row, large tap targets.
• Use the cart drawer instead of a full cart page on mobile.
• Show the recommendation above the checkout button.
• Limit to one recommendation on mobile - a single large card; lazy-load its image.

Typical lift: +20% to +28% mobile upsell acceptance. Impact: medium-high - with 65% of traffic on mobile, a modest lift compounds.

How Levri spots all six in 60 seconds.

Levri detects missing upsell surfaces, irrelevant recommendation blocks, poor placement, missing one-click mechanics, and mobile layout issues - then ranks each by estimated revenue impact.

You paste the URL, and you get your fixes - a ranked list, each issue priced in $/mo, with a written hypothesis, a variant-B suggestion, and an expected lift range. No install, no tracking script.

Fix these first.

In the order we'd ship:

1. Add a post-purchase one-click offer (highest acceptance, fastest).
2. Replace irrelevant cart cross-sells with cart-aware complements.
3. Simplify to one or two offers per page.
4. Add a cart-drawer cross-sell on mobile.
5. Audit mobile tap targets and placement.
6. Stack two or three funnel placements and measure incremental AOV.

The highest-converting surface is the confirmation page (thank-you page optimisation), and the offer should never fight the primary action (product page conversion). Ship three and measure by Friday; if AOV doesn't move, the offer's in the wrong place, which is exactly what Levri is built to diagnose.