Detect and act on opportunities to compute a business score from index data

[emirbelkahia]

Context

Many clients still don't use custom ranking, even though they often already have relevant business signals in their index (e.g. sales, reviews, newness). This prevents them from aligning search results with business goals — a key lever for impact.

This inertia is rarely due to a lack of will, but rather:

• uncertainty about what data matters,
• lack of clarity on how to build a meaningful score,
• or not realizing that the data is already there.

Opportunity

The MCP node can bridge that gap by:

• Scanning the index for common business signals (e.g. sales, reviews, rating, recency, stock, etc.)
• Proposing a smart, default scoring formula based on e-commerce heuristics
• Explaining the logic used (fields + weights)
• Most importantly: proactively suggesting next steps based on findings.

This turns passive insight into active support, enabling clients to take action — or delegate the action to the MCP itself.

Proposed behavior

When called, the node should:

If business data is found:

• Compute a score per record using smart defaults (e.g. weighted average)
• Explain which fields were used and how
• Offer a recommended next action such as:
  • “Would you like me to enrich all your records with this score?”
  • “Should I create a new businessScore attribute with this formula?”
  • “Want me to draft a transformation rule you can review and deploy?”

If no usable data is found:

• Clearly highlight this as a missed opportunity
• Suggest relevant attributes that could be indexed
• Offer links to documentation or invite to contact a CSM
• Encourage action (index data, or get help)

Example prompt

Can you compute a business score from my product index and apply it? If I don’t have the data you need, tell me what’s missing and suggest next steps.

Expected output

Case 1: Data available

{
  "scoreFormula": "0.5 * sales + 0.3 * rating + 0.2 * daysSinceCreation",
  "fieldsUsed": ["sales", "rating", "createdAt"],
  "explanation": "I computed a weighted average based on available business signals. This approach is widely used in e-commerce to prioritize products not only by popularity (sales), but also by quality (rating) and freshness (recency). These signals are among the most common and effective indicators of commercial performance in digital merchandising.",
  "previewSample": [
    { "objectID": "123", "businessScore": 8.3 },
    { "objectID": "456", "businessScore": 6.1 }
  ],
  "nextAction": "Would you like me to enrich all your records with this 'businessScore'? This will make it easier to apply business-driven custom ranking."
}

Case 2: No relevant data

{
  "message": "I couldn't find any business-related fields in your index that are typically used to compute a business score (e.g. sales, rating, or newness).",
  "missedOpportunity": true,
  "suggestedFields": ["sales", "rating", "numberOfReviews", "createdAt"],
  "recommendedNextSteps": [
    "📘 Read this guide on how to add business signals to your index: [https://algolia.com/doc/guides/relevance/business-metrics/"](https://www.algolia.com/fr/ecommerce-merchandising-playbook/playBook/custom-ranking),
    "📈 Adding fields like sales or rating can help you prioritize high-performing products in search.",
    "🤝 If you have an assigned CSM, feel free to reach out for help setting this up."
  ]
}

Value

• Helps clients immediately activate custom ranking.
• Turns latent signals into actionable levers.
• Makes the MCP proactive, not just descriptive.
• Boosts trust and perceived intelligence of the platform.

Notes

• MCP should avoid overwriting unless explicitly authorized.
• To be fully usable, the action taken should probably create an Algolia transformation scheduled to repeat every day so the scores remain up to date
• Could be extended later to learn from user choices (accept/refuse actions).