> ## Documentation Index
> Fetch the complete documentation index at: https://www.meilisearch.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Personalize ecommerce search

> End-to-end example of implementing personalized search for an ecommerce store.

This guide walks through a complete ecommerce personalization implementation. You will set up an [embedder](/capabilities/hybrid_search/overview) with personalization, collect user signals, build user profiles, and send personalized search requests that return different results for different shoppers.

## Step 1: Set up your product index

Make sure your product index contains rich, descriptive documents. The more relevant fields your documents have, the better personalization can re-rank results:

<CodeGroup>
  ```json theme={null}
  [
    {
      "id": 1001,
      "title": "Samsung Galaxy Buds Pro",
      "category": "Electronics",
      "brand": "Samsung",
      "price": 149.99,
      "description": "Premium wireless earbuds with active noise cancellation."
    },
    {
      "id": 1002,
      "title": "Sony WH-1000XM5",
      "category": "Electronics",
      "brand": "Sony",
      "price": 349.99,
      "description": "Industry-leading noise canceling over-ear headphones."
    },
    {
      "id": 1003,
      "title": "JBL Go 3",
      "category": "Electronics",
      "brand": "JBL",
      "price": 39.99,
      "description": "Compact portable Bluetooth speaker with bold sound."
    }
  ]
  ```
</CodeGroup>

## Step 2: Collect user signals

Track user interactions on your ecommerce site. The most useful signals for personalization include:

| Signal            | Example                                 | Weight |
| ----------------- | --------------------------------------- | ------ |
| Purchases         | Bought 3 Samsung products               | High   |
| Cart additions    | Added Sony headphones to cart           | Medium |
| Product views     | Viewed 15 electronics items this week   | Medium |
| Category browsing | Spent 10 minutes in "Audio" category    | Low    |
| Search history    | Searched for "wireless earbuds" 3 times | Low    |

Store these signals in your user database. You do not need to send raw event data to Meilisearch. Instead, you aggregate these signals into a profile string on your backend.

<Tip>
  If you use Meilisearch analytics, you can track clicks and conversions with the [events API](/capabilities/analytics/how_to/track_click_events) and use that data to build richer user profiles.
</Tip>

## Step 3: Build a user profile string

Transform aggregated signals into a profile string. Focus on positive, affirmative statements:

<CodeGroup>
  ```javascript theme={null}
  function buildShopperProfile(user) {
    const parts = [];

    // Purchase patterns
    if (user.topCategories?.length > 0) {
      parts.push(`Frequently buys ${user.topCategories.join(' and ')}.`);
    }

    // Brand preferences
    if (user.favoriteBrands?.length > 0) {
      parts.push(`Prefers ${user.favoriteBrands.join(', ')}.`);
    }

    // Price sensitivity
    if (user.avgOrderValue < 50) {
      parts.push('Budget-conscious shopper.');
    } else if (user.avgOrderValue < 200) {
      parts.push('Mid-range budget.');
    } else {
      parts.push('Prefers premium products.');
    }

    // Recent activity
    if (user.recentSearches?.length > 0) {
      parts.push(
        `Recently searched for ${user.recentSearches.slice(0, 3).join(', ')}.`
      );
    }

    return parts.join(' ');
  }
  ```
</CodeGroup>

Example output: `"Frequently buys electronics. Prefers Samsung, Sony. Budget-conscious shopper. Recently searched for wireless earbuds, portable speakers."`

## Step 4: Send personalized search requests

Pass the user profile string in the `personalize` search parameter:

<CodeGroup>
  ```bash theme={null}
  curl \
    -X POST 'MEILISEARCH_URL/indexes/products/search' \
    -H 'Content-Type: application/json' \
    --data-binary '{
      "q": "headphones",
      "personalize": {
        "userContext": "Frequently buys electronics. Prefers Samsung, Sony. Budget-conscious shopper. Recently searched for wireless earbuds, portable speakers."
      }
    }'
  ```
</CodeGroup>

## Step 5: Compare results for different profiles

The same search query returns different result rankings for different user profiles. Here is how results for "headphones" might differ:

### Budget-conscious electronics buyer

**Profile**: `"Frequently buys electronics. Prefers Samsung. Budget-conscious shopper."`

| Rank | Product                 | Price    |
| ---- | ----------------------- | -------- |
| 1    | Samsung Galaxy Buds Pro | \$149.99 |
| 2    | JBL Go 3                | \$39.99  |
| 3    | Sony WH-1000XM5         | \$349.99 |

### Premium audio enthusiast

**Profile**: `"Prefers premium products. Audiophile, values sound quality above all. Prefers Sony and Bose."`

| Rank | Product                 | Price    |
| ---- | ----------------------- | -------- |
| 1    | Sony WH-1000XM5         | \$349.99 |
| 2    | Samsung Galaxy Buds Pro | \$149.99 |
| 3    | JBL Go 3                | \$39.99  |

The underlying search results are the same, but personalization re-ranks them based on relevance to each user's profile.

## Tips for effective ecommerce personalization

* **Update profiles regularly.** Recalculate the user profile string after each session or purchase to keep it current.
* **Use affirmative language.** Write "prefers budget options" instead of "avoids expensive products." The re-ranking model responds better to positive signals.
* **Keep context concise.** One to three sentences is ideal. There is no hard maximum length, but longer strings increase latency and cost without improving results.
* **Test with real users.** Compare click-through rates and conversion rates between personalized and non-personalized search to measure impact. Use [analytics](/capabilities/analytics/overview) to track these metrics.
* **Start with high-confidence signals.** Purchases and cart additions are stronger indicators than page views or browse time.

## Next steps

<CardGroup cols={2}>
  <Card title="Getting started with personalization" href="/capabilities/personalization/getting_started/personalized_search">
    Enable personalization and perform your first personalized search
  </Card>

  <Card title="Track click events" href="/capabilities/analytics/how_to/track_click_events">
    Use analytics events to collect user signals for personalization
  </Card>

  <Card title="Generate user context" href="/capabilities/personalization/how_to/generate_user_context">
    Strategies for building user context from different data sources
  </Card>
</CardGroup>