Recommendations that prove they worked.
Affina puts AI product recommendations on your product and cart pages, learns what actually sells together in your store, and measures the added revenue against a real 10% holdout — so you see lift, not vanity metrics.
Pre-launch · no customer data collected (Protected Customer Data Level 1)Who it's for
Shopify merchants who have related products but no reliable way to surface the right pairings, and who want a higher average order value without hiring a merchandiser. If you've ever wondered whether your “you may also like” widget actually makes money, Affina is built to answer that.
The problem
Most recommendation apps show impressions and clicks, then ask you to assume they helped. They run on models pre-trained on other stores, and they never tell you whether the sale would have happened anyway. You're left guessing whether the app earns its keep.
Built for where you are
Different stores need different things from a recommendation app. Affina meets each one honestly — and never claims a number it can't measure.
Affina works from day zero — it pairs products by what they are, not by order history you don't have yet, so you can switch it on before your first sale. It's upfront about the rest: the measured-lift number arrives once enough shoppers pass through.
Affina learns from your own shoppers and proves the revenue it adds against a real 10% holdout. You keep it only if it pays for itself — and you'll know to the dollar, instead of guessing.
The holdout gives you attribution a CFO will accept. Rules control what shows where, and Affina collects no personal data — so it clears privacy review without a fight.
Affina is the one that tells you the truth: run it alongside, watch the holdout-measured lift, and decide with evidence — not a vendor's dashboard that counts every order it touched as its own.
How it works
Adaptive model per store and per placement. Nothing pre-trained on other merchants — it learns from your own shoppers.
Affina deliberately shows about one in ten shoppers no recommendations, as a control group. That's how it measures lift against reality.
Recommendations appear on the product page and the cart page. You add them by switching the block on in the theme editor — Affina never edits your theme.
While there isn't enough traffic yet, the dashboard says so. Once there is, it reports the revenue you earned beyond what the holdout would have earned anyway.
Key features
Theme-adaptive — inherits your store's fonts and colours. Never Affina branding.
Revenue added vs. a real 10% control group, per-placement breakdown. Honest 'still learning' state until there's enough data.
Learns from your shoppers. Nothing pre-trained on other merchants.
Exclude out-of-stock, skip already-purchased, hide on-sale, margin floor, price band, frequency cap, excluded tags, a never-recommend list.
Up-weight a product, or pin 'always show B with A'.
In-app assistant that reads your live setup, explains why a product is or isn't being recommended, and makes the changes for you.
Recommendation explorer, shopper-context views, catalogue coverage, order journeys, and model performance.
Widget labels auto-translate to the shopper's language — English, Spanish, French, German, Portuguese, Italian, Dutch.
Why Affina
A real holdout and incremental lift — not impressions dressed up as value.
Per-store learning. No transfer from other merchants.
Affina recommends by what's being viewed, not by who's viewing. It works for every shopper, including anonymous ones, and stores no customer personal information (Shopify Protected Customer Data Level 1).
Everything lives in Shopify Admin, in one consistent, quiet interface.
Pricing
Paid-only, with a 14-day free trial on every plan, billed through Shopify. Tiered by your store's monthly revenue — recommendations are unlimited on every plan. Within days you see Affina working — shoppers viewing, clicking, and adding recommended products, and the assisted revenue that follows. The holdout-measured lift takes a little longer to prove and lands as your traffic builds.
All plans include both placements (product + cart). No free tier.
FAQ
What does Affina do?
It puts AI product recommendations on your product and cart pages, learns what sells together in your store, and proves the added revenue against a 10% holdout.
How is it different from a 'you may also like' app?
Honest, holdout-measured lift; a model that learns your store specifically; and no customer data required.
Does it need my customers' personal data?
No. Affina recommends based on what's being viewed, not who's viewing. It stores no customer names, emails, or addresses (Protected Customer Data Level 1).
Will it edit my theme?
No. You switch on the recommendation blocks yourself from the theme editor; Affina only reads your theme to detect that they're live.
How long until I see results?
Recommendations are live immediately, and within days you see real engagement — shoppers viewing, clicking, and adding recommended products — plus the assisted revenue that follows. The holdout-measured lift takes longer to prove (it needs enough traffic to compare against the control group, usually a couple of weeks) and lands on your dashboard as it becomes real.
What does it cost?
$19, $49, or $99 per month (Starter / Pro / Scale), tiered by your store's monthly revenue, with a 14-day free trial on every plan. Recommendations are unlimited on every plan. No free tier.
Recommendations that prove they worked
Affina is feature-complete and pending Shopify App Store approval. The install link goes live when the listing is approved.
Email us about early accessQuestions: hello@lunalink.ai