Does AI-powered product recommendation (upsell/cross-sell apps) actually lift AOV meaningfully for a mid-market catalog, or is that overstated?
Short answer
AI product recommendation apps have no public average order value (AOV) lift range for mid-market Shopify catalogs (Deploi search, October 2026), so any AOV percentage stays unproven until your own holdout test confirms it. In one agency test, Rebuy beat Shopify's free recommendations by 23% on conversion but 2% on AOV over four days (Blend Commerce, August 2026).
- Agency test against native
- Conversion up 23%, AOV up 2%Rebuy against Shopify's own recommendations on Stone Creek Coffee's product pages: 3,452 visitors over four days, ended early (Blend Commerce, August 2026).Source: Blend Commerce (Increase PDP Revenue with Rebuy Product Recommendations)
- Independent field experiment
- Conversion up 5.9%; no AOV figureA randomized test on a top North American retailer's website with 184,375 users (Lee and Hosanagar, 2016 Web Conference proceedings).Source: IW3C2 (When do Recommender Systems Work the Best? The Moderating Effects of Product Attributes and Consumer Reviews on Recommender Performance)
- Rebuy entry price
- From $25 a monthOne package at the pricing page's default of 500 orders a month; Platform One shows $534 a month, billed monthly, at the same volume (Rebuy pricing page, October 2026).Source: Rebuy (Rebuy Engine Pricing)
- Native baseline
- Free related productsGenerated automatically, plus up to 10 picked by hand; they skip sold-out items and items already in the visitor's cart (Shopify Help Center, October 2026).Source: Shopify Help Center (Customize product recommendations with Shopify Search & Discovery)
Deploi’s call
Recommendation lift is real and routinely overstated, and the gap is a measurement artifact rather than a vendor conspiracy. Buy the engine if you want the cart, checkout and post-purchase surfaces and someone to own them. Do not buy it on a case-study percentage.
BuyBuy a recommendation engine when
- You will run three or more surfaces, such as cart, checkout and post-purchase; Deploi's Rebuy versus native record scores that breadth BUY (October 2026).
- The vendor supports a holdout test on your own traffic before renewal.
- Someone on your team owns the recommendation slots, their exclusions and their placement.
CustomizeTune Shopify's native recommendations instead while
- You run one or two placements, which Deploi's Rebuy versus native record scores CUSTOMIZE.
- Nobody has measured a month of revenue from Shopify's native recommendations yet, so a fee would have no baseline to beat.
- Your catalog is thin: a few hundred active products, or a single price point.
Basis: Rebuy vs. Native Recommendations: Do You Need the Full Suite?, our Build vs. Buy decision record, scored DEPENDS.
Why don't published lift numbers prove an order value gain?
Published lift numbers don't prove an order value gain, because none gives a mid-market Shopify catalog an average order value (AOV) range. Recommendations are one of the three machines sold as merchandising AI, and their public evidence comes in three kinds:
| Source | Whose stores | Sample | What moved |
|---|---|---|---|
| Academic field experiment, 2016 | A top North American retailer's website | 184,375 users, randomized | Conversion up 5.9%; AOV not reported |
| Blend Commerce, an agency selling Rebuy integration, August 2026 | Stone Creek Coffee's Shopify product pages, Rebuy against Shopify's own recommendations | 3,452 visitors over four days; ended early | Conversion up 23%, AOV up 2% |
| Hello Retail vendor guide, April 2026 | Not stated | Not stated | "76% better conversion" |
Deploi searched in October 2026 and found no independent holdout test on a mid-market Shopify catalog. Deploi's AI merchandising record gives the same caution: lift claims come from vendor case studies, so insist on a holdout test on your own traffic before renewal (October 2026).
Which measurement problems inflate recommendation lift?
Four measurement problems inflate recommendation lift, and each one makes a widget look better than the business result:
- Widget attribution counts orders that were already happening. If a shopper clicks a recommendation for a product they meant to buy, the widget books the revenue, so attributed revenue is not incremental revenue.
- Average order value (AOV) is the wrong denominator. An engine that adds a cheap accessory to some orders can raise attach rate and lower AOV at the same time. Revenue per session reflects the business.
- Mix shifts get read as lift. Launch a recommendation engine in September, compare it to August, and you have measured the season.
- The baseline is usually nothing, not free. Shopify generates related products automatically, lets you add up to 10 by hand, and skips sold-out items and items already in the visitor's cart (Shopify Help Center, October 2026). The honest baseline is Shopify's own recommendations, configured.
Deploi's Rebuy versus native record names the conflict of interest: "inside the suite, the dashboard that sells the lift also grades it" (October 2026).
What has to be true for recommendations to lift anything?
Recommendations can only lift orders when the catalog gives the model somewhere to send the shopper, so check five conditions before you shortlist a vendor.
| Condition | Why it matters | Rough threshold |
|---|---|---|
| Catalog breadth | A model needs somewhere to send the shopper | Thin below a few hundred active products |
| Co-purchase density | Patterns need repeat order evidence | Sparse catalogs produce near-random pairs |
| Price ladder | Cross-sell needs an accessory or a step up | A single price point leaves nothing to attach |
| Surface availability | Cart, checkout and post-purchase slots sit next to the buying decision | Information, shipping and payment steps need Shopify Plus; post-purchase offers run from the Basic plan up |
| Data freshness | Stale inventory feeds recommend what you cannot ship | Sync in hours, not days |
Shopify confirms the checkout split. Apps for the information, shipping and payment pages "are available only to Shopify Plus," while post-purchase apps run from the Basic plan up (Shopify Help Center, October 2026).
What do recommendation engines cost on Shopify?
Recommendation engines on Shopify run from a published $25 a month to quote-only platforms, and published prices rise with order volume. Rebuy's pricing page shows $25 a month for one package and $534 a month for its Platform One suite, billed monthly, both at its default of 500 orders a month (Rebuy pricing, October 2026). A mid-market store ships far more than 500 orders a month, so it pays a higher tier. Nosto is free to install, but "A Nosto account is required and is not free of charge" (Nosto listing, October 2026). Its listing shows no tier prices.
Breadth decides whether the fee pays, so count the surfaces you will actually run. Deploi's Rebuy versus native record scores Rebuy BUY when three or more surfaces will earn their keep, and CUSTOMIZE for one or two placements (October 2026). If your only upsell runs after payment, a page builder's post-purchase offers may cover it.
How do you get an answer for your own catalog in one quarter?
You get an answer for your own catalog in one quarter by running a holdout test before you renew, not after. Split traffic, leave the recommendation surfaces off for a set share of sessions, and compare revenue per session rather than average order value (AOV). Run the test across a full purchase cycle, because a short test can end before it settles: Blend Commerce stopped its Rebuy test after four days at an 89.5% probability on conversion (Blend Commerce, August 2026).
In our judgment, the result is often positive and usually smaller than the sales deck. A small real number beats a large attributed one, because you can budget against it.
Deploi’s point of view
Our own position, from building on Shopify. Kept apart from the sourced facts above.
- What we’ve seen
- Most disappointing launches we have looked at were not model failures; they were a recommendation row placed below the fold, or a feed with stale availability.
- Where we disagree
- The category publishes lift numbers as though they were properties of the product. Lift is a property of your catalog, your traffic and your measurement design. We would rather hand a client a holdout plan than a benchmark.
- What we check first
- Correct product data, sensible exclusions, a recommendation row that does not repeat the item already in the cart, and a slot in the right position on the page. In our judgment this unglamorous work matters more than the model.
- What this page adds
- The parent names recommendations as one of three machines sold as merchandising AI. This page lines up the public lift evidence by whose stores it measured, explains why average order value figures inflate, and gives the catalog conditions and the holdout plan that settle it for one catalog.
Decision records behind this answer
When this answer changes
- An independent holdout study on mid-market Shopify catalogs is published.
- Shopify adds a recommendation intent beyond related and complementary, or native recommendations in the cart.
- Rebuy or Nosto changes its pricing basis or entry price.
- Shopify opens the information, shipping or payment steps of checkout to apps below Plus.
- Your own holdout test completes a full purchase cycle.
Related questions
This follows from What does 'merchandising AI' actually mean on Shopify? Is it search, personalization, or something broader?
- Does GemPages' funnel builder support upsell/cross-sell offers natively, or do I still need a separate app like Rebuy?
GemPages' funnel builder supports upsells natively only after payment, with up to two offers; its funnel has no cart step and leaves checkout to Shopify (GemPages Help Center, October 2026).
Sources
Every figure on this page traces to one of these. Official and vendor pages are checked on the date shown; Deploi records are our own published decisions.
- Blend Commerce, Increase PDP Revenue with Rebuy Product Recommendations
- IW3C2, When do Recommender Systems Work the Best? The Moderating Effects of Product Attributes and Consumer Reviews on Recommender Performance
- Shopify Help Center, Customize product recommendations with Shopify Search & Discovery
- Rebuy, Rebuy Engine Pricing
- Shopify Help Center, Customizing your checkout and customer accounts with apps
- Hello Retail, Shopify B2B is on every plan. What about personalization?
- Nosto Solutions Ltd, Nosto | AI Search & Discovery on the Shopify App Store
- Deploi Build vs. Buy, Should You Build or Buy AI Merchandising & Sorting on Shopify?
- Deploi Build vs. Buy, Rebuy vs. Native Recommendations: Do You Need the Full Suite?