Product Recommendations

The recommendation engine provides three algorithms: co-purchase frequency, product similarity scoring, and personalized recommendations based on order history. All algorithms filter out inactive and out-of-stock products.

Architecture

Recommendation Types

Frequently Bought Together

getFrequentlyBoughtTogether(productId, limit) analyzes the last 100 orders containing a product and ranks co-purchased products by occurrence frequency.

Algorithm:

1. Find distinct orders containing the target product (last 100)
2. Group other products in those orders by productId
3. Count occurrences and sort descending
4. Fetch product details, filtering to active + in-stock
5. Normalize scores (0-1) based on max occurrence count

const coOccurrences = await prisma.orderItem.groupBy({
  by: ['productId'],
  where: {
    orderId: { in: orderIds },
    productId: { not: productId },
  },
  _count: { orderId: true },
  orderBy: { _count: { orderId: 'desc' } },
  take: limit * 2,
})

You May Also Like

getYouMayAlsoLike(productId, limit) scores candidate products by similarity to the source product:

Products are sorted by total score descending.

Personalized Recommendations

getPersonalizedRecommendations(userId, limit) builds a preference profile from the user's last 10 paid orders:

1. Extract all purchased product IDs, categories, and heat levels
2. Rank categories and heat levels by frequency
3. Find active, in-stock products matching the top category or heat level
4. Exclude already-purchased products
5. Return with a fixed confidence score of 0.8

const topCategory = Array.from(categories.entries())
  .sort((a, b) => b[1] - a[1])[0]?.[0]
const topHeatLevel = Array.from(heatLevels.entries())
  .sort((a, b) => b[1] - a[1])[0]?.[0]

RecommendedProduct Type

All three algorithms return the same type:

interface RecommendedProduct {
  id: string
  name: string
  slug: string
  price: number
  featuredImage: string | null
  heatLevel: string | null
  sku: string
  inventory: number
  score: number  // 0-1 confidence score
}

UI Components

Product Recommendations

The ProductRecommendations component renders a grid of recommended products below the product detail page, using getFrequentlyBoughtTogether() and getYouMayAlsoLike().

Recently Viewed

The RecentlyViewed component (lib/store/recently-viewed.ts) uses a separate Zustand store to track and display products the user has visited in the current session.