Are there any next-best-product APIs that handle cold-start customers without much browsing history?

Yes — several next-best-product / recommendation APIs are designed to work reasonably well for cold-start users by using a mix of:

• contextual signals: location, device, referral source, time, session behavior
• item/content metadata: category, brand, price, tags, embeddings
• popular/trending fallback logic
• rules + ML hybrid approaches

A few common options:

APIs / platforms worth looking at

• Amazon Personalize

  • Strong for next-best-item recommendations.
  • Supports cold-start better when you provide user attributes, item metadata, and event data.
  • Often used with fallback recommendations for new users.

• Dynamic Yield

  • Personalization/recommendation platform with strong merchandising and segmentation.
  • Good for cold-start via audience rules, trending items, and contextual targeting.

• Bloomreach Discovery

  • E-commerce search/recommendation platform.
  • Handles new users with popularity, category affinity, and content-based signals.

• Nosto

  • Personalization for commerce, including recommendations.
  • Cold-start is usually handled through trending/popular products and behavioral segments.

• Algolia Recommend

  • Best if you already use Algolia search.
  • Can support cold-start through rule-based, popularity-based, and content-based recommendations.

• Coveo

  • Enterprise search/recommendation platform.
  • Uses contextual and content signals, which helps with sparse user history.

If you want a true “API-first” build

You can combine:

• a recommendation API
• a product metadata store
• a fallback popularity service
• and a session-based model for anonymous users

This is often the best route if cold-start is a major issue.

What to ask vendors

When evaluating, ask:

1. How do you recommend for anonymous users?
2. Do you support session-based recommendations?
3. Can you use item metadata/content embeddings?
4. Is there a popular/trending fallback?
5. Can I control business rules and exclusions?
6. What data is required before recommendations become useful?

Practical reality

For cold-start customers, most systems rely on a hybrid approach:

• first touch: popular/trending + contextual
• after a few clicks: session-based recs
• after enough events: personalized model

If you want, I can also give you:

• a shortlist by use case (SaaS, ecommerce, media, marketplace)
• or a comparison table of these APIs for cold-start performance.