Marketplace Search Rank and Share-of-Shelf Data

The buyer problem: visibility is not one rank number

Brand and ecommerce teams need to know whether shoppers can actually find their products. A single rank column cannot answer that question because marketplaces mix organic results, sponsored placements, category pages, carousels, and personalized or location-dependent inventory. The decision this article supports is how to define a repeatable visibility dataset that separates those surfaces before calculating share of shelf.

Define the observation before the metric

Paid placements must be explicit. Amazon’s official Sponsored Products FAQ says ads may appear at the top, alongside, or within shopping results and on product detail pages; its Sponsored Brands guide similarly describes placements on top of and within results. That documented mixing is why an unlabeled position is not analytically safe.

A canonical result observation

{
  "marketplace": "example-market",
  "query": "protein bar",
  "location": {
    "country": "IN",
    "postal_code": "560001"
  },
  "device": "mobile_web",
  "observed_at": "2026-08-19T09:00:00+05:30",
  "page": 1,
  "slot": 4,
  "absolute_position": 4,
  "placement_type": "sponsored_product",
  "listing_id": "SKU-4821",
  "brand": "Example Brand",
  "in_stock": true
}

Preserve duplicate appearances: a product can occupy an ad and an organic slot in the same result set. Keep the raw result snapshot or a reproducible reference alongside parsed rows. Relevant PLOTT pages include the marketplace directory, rankings and visibility data, and CPG brand monitoring use case.

Calculate rank and share of shelf transparently

Organic rank is the ordered position among organic product results after removing clearly labeled paid modules. Absolute rank is the visible slot counting every product placement. Report both where possible. Category rank should refer only to a marketplace’s stated category ordering or badge; do not infer it from a keyword query.

A simple unweighted share of shelf for brand b over the first N product slots is slots occupied by b / N. A position-weighted version is sum(w(position) for b) / sum(w(position) for all observed slots). Publish the weight function— for example 1 / log2(position + 1)—and calculate paid, organic, and combined shares separately.

Aggregate only after defining the query set. A brand’s weekly score should include the keywords, locations, collection times, and weights used. Otherwise a score can rise merely because the sample changed.

Collection and QA checklist

• Use a fixed, versioned keyword and category set with business weights.
• Collect all competitors in the requested depth, not only the buyer’s brand.
• Capture sponsored labels and non-product modules before page order is flattened.
• Resolve listings to canonical products and brands; retain unknowns.
• Repeat priority queries across agreed locations and dayparts.
• Flag partial pages, blocked runs, consent screens, and changed layouts.
• Monitor result count, unknown-placement rate, and parse completeness.

Limitations and POC validation

Results can vary with location, stock, account state, experiments, personalization, device, and ad auctions. Labels and layouts change, infinite-scroll pages can load conditionally, and an observed position is a snapshot, not a guarantee that every shopper saw it. Do not describe an automated collection as the marketplace’s official rank unless the platform itself provides that metric.

A POC should test named keywords across representative locations, mobile and desktop where relevant, and at least several collection windows. Manually audit placement classification and position counts, quantify failed or partial runs, and confirm how duplicate listings and variants are handled. Ask PLOTT for a search-results sample dataset with raw placement evidence, organic and sponsored ranks, location context, timestamps, and the exact share-of-shelf calculation.