Score reasoning

The Score Reasoning panel sits next to the AI brief on the product detail page. It’s a focused explanation of why the SellerScore is what it is — which sub-scores drove the composite up, which dragged it down, and how confident Claude is in the result.

This is different from the AI brief in tone:

  • The AI brief is product-strategy advice (“here’s how to position this”).
  • The score reasoning is score-mechanics explanation (“here’s why the composite is 72”).

In this guide:

  • What the reasoning includes
  • How it’s generated
  • When it’s most useful
  • The cache

What the reasoning includes

A typical reasoning paragraph covers:

  1. The headline. “This product scored 72 — driven primarily by strong demand (84) and acceptable margin (68), held back by elevated competition (54) and a flat trend.”
  2. The drivers. Which sub-scores moved the composite most.
  3. The drags. Which sub-scores held it back.
  4. Confidence. “Data quality is high — only the cross-platform sub-score is a placeholder, which doesn’t materially change the composite.”
  5. What to verify. “Worth verifying the COGS estimate before committing — your actual supplier price could move margin from 68 to 80 or to 50.”

It’s structured for scanning — you can pick out the drivers, drags, and confidence in 15 seconds.

How it’s generated

The reasoning generator gets:

  • All six sub-scores and the weights used.
  • The data-quality flags and warnings.
  • The composite SellerScore.
  • A system prompt asking for a structured “drivers / drags / confidence / verify” explanation.

Claude returns a short paragraph. Hilal stores it alongside the search snapshot.

When it’s most useful

When you need to defend a recommendation

You’re presenting research to a partner or client; the reasoning is your defensible answer to “why this product?”

When the SellerScore surprises you

A product you expected to score low scored high (or vice versa). The reasoning tells you which sub-score is the surprise driver — usually that’s where to focus your independent verification.

When data quality is mixed

The reasoning makes the data-quality story concrete. Instead of “yellow badge,” you see “demand fell back to BSR + Trends because SFR wasn’t available; that adds ±10 to the demand score’s confidence interval.” That’s actionable.

When it’s less useful

  • For products with clean, dominant signals. A 90 SellerScore with all-green sub-scores doesn’t need much reasoning.
  • For very low-scored products. A 25 SellerScore is usually obvious from the chip strip — no reasoning needed.

The cache

Like the AI brief, score reasoning is cached. Same 48-hour cache, same (asin, marketplace) key, same shared-across-users economics. Most popular products have reasoning already cached; niche products get reasoning on-demand (a few seconds).

Regenerating

Score reasoning regenerates whenever the product is re-searched (new search → new reasoning). If you specifically want fresh reasoning without re-searching, top-right of the reasoning panel: Regenerate (small AI call, not a search credit).

Related articles