Guide to Use AI for Product Recommendation and Upsell


Shipped
Last Updated: 8 hours ago

This guide explains how to create SOPs that help Cavalry AI recommend the right products to customers and suggest complementary items as upsells. With the right setup, AI can either:

  1. Natural Catalog Search – Gather information, search your catalog, and recommend the most relevant product(s), while suggesting add-ons.

  2. Guided Recommendation – Follow a structured quiz-like flow with predefined mappings to recommend specific products and targeted upsells.

Both approaches ensure customers receive confident, personalized recommendations while also driving higher average order value (AOV).


When to Use

  • You want AI to leverage your catalog search results and product metadata.

  • If the product description is descriptive enough, the AI can use it (along with any exposed product tags or metadata) to provide relevant recommendations to the customer.

SOP Structure

Step 1: Identify Intent

  • Detect when a customer is asking for a recommendation.

Step 2: Ask Clarifying Questions
If the request is too general, prompt the AI to ask details such as:

  • “What’s your main goal (e.g., hydration, anti-aging, soothing)?”

  • “Do you prefer capsules, powder, or liquid form?”

  • “Any ingredient preferences or restrictions?”

Step 3: Catalog Search

  • Instruct AI to search your product catalog with the customer’s keywords.

  • AI can also see any exposed product tags in the search results (e.g., “vegan,” “oily skin,” “travel size”) and use them to prioritize or refine its recommendation.

  • Include fallback instructions if no relevant products are found.

Step 4: Recommend Products

  • If one clear match: confidently recommend that product.

  • If multiple matches: suggest 2–3 options and explain differences.

  • Always link the product name to the storefront URL so customers can easily view or purchase.

Step 5: Tone and Upsells

  • Confident but friendly phrasing:

    • “Based on what you shared, I’d recommend our Hydration Serum—it’s great for dry skin.”

    • “We have two options that could work: Product A and Product B. If you prefer [format], I’d go with [Product B].”

  • Upsell example:

    • “For best results, many of our customers like to pair this with Night Repair Cream.”


Approach 2: Guided Recommendation (Quiz Style)

When to Use

  • You want a controlled flow with guaranteed results.

  • Your products map cleanly to specific needs or preferences.

⚠️ Note: Since the AI is not conducting a knowledge search in this approach, you must include the exact product names and links in the answer matrix for the AI to reference directly.

SOP Structure

Step 1: Build the Answer Matrix
Define how answers map to products. Example:

Question

Possible Answer

Recommended Product

Skin type?

Dry

Hydration Serum

Skin type?

Oily

Balancing Gel

Concern?

Fine lines

Firming Cream

Step 2: Script the Question Flow
Create a conversational quiz-like path:

  • “What’s your skin type?”

  • “What’s your top concern?”

  • “Do you prefer fragrance-free or scented?”

Step 3: Match Answers to Products

  • Instruct AI to follow the answer matrix and provide the mapped product(s).

  • Example:

    • If customer selects “Oily + Fine lines” → Recommend “Balancing Gel” + “Firming Cream.”

Step 4: Add Specific Recommendations & Upsells

  • Always recommend the Hero Product as a first option.

  • Conditional upsells:

Primary Product

Upsell Condition

Upsell Product

Hydration Serum

Dry Skin

Night Repair Cream

Balancing Gel

Oily Skin

Oil-Free Cleanser

  • Example AI phrasing:

    • “Thanks for sharing! Based on your answers, the Hydration Serum is the perfect fit for you.”

    • “For even better results, I recommend pairing it with the Night Repair Cream.”

Step 5: Tone

  • Keep it light and conversational:

    • “Got it! One last question before I give you my top pick…”

    • “Here’s my recommendation for you—plus a bonus option if you’d like to enhance your routine.”


Tips for Both Approaches

  • Product Names & Links: Always provide exact product names and storefront links in recommendations.

  • Tags in Results: AI can reference exposed product tags from catalog search results to refine responses.

  • Fallbacks: Always include instructions for when no exact product fits.

  • Review & Update: Keep SOPs updated as your catalog evolves.


If you need help creating a Product Recommendation SOP or building an answer matrix for guided recommendations, contact support@cavalry.ai.


Was this article helpful?