Why should you care?

AIPredict revenue lifts when checkout and home show relevant add-ons, not random catalog rows.

See it live — copy this example

Use a .NET console or Web API project with Microsoft.ML. Run dotnet run after pasting.

var engine = ml.Model.CreatePredictionEngine<UserProduct, ProductScore>(model);
var candidates = catalog.Where(p => !purchased.Contains(p.Id));
var ranked = candidates
    .Select(p => new { p.Id, Score = engine.Predict(new UserProduct { UserId = uid, ProductId = p.Id }).Score })
    .OrderByDescending(x => x.Score).Take(10);

What happened?

• Score user–product pairs, filter already-owned, apply stock/ margin rules, then return top-N.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Load product MF model.
2. Score candidate pairs.
3. Filter purchased and OOS.
4. Boost high-margin items +5%.
5. Cap same-category duplicates.

Remember

Score → filter → rank. Business rules last. Top-N for UI slots.

AIPredict cart upsell

Checkout suggests accessories from MF scores.

Outcome: Attach rate up without spamming unrelated SKUs.