Other Marketplaces and Their Ranking Systems
Rank systems differ in what they measure, how visible they are and how they can be tracked. What transfers from one marketplace to another and what does not.
The methods on this site were developed against one marketplace's sales rank. Most of the reasoning transfers; the specifics do not.
What transfers
Rank is relative. True everywhere there is an ordering.
Recency weighting and decay. Present in most systems, with different parameters.
The need for a control basket. Universal.
Context capture — price, availability, reviews. Universal.
Estimation requires calibration and travels badly across systems.
Seasonality confounds everything. Universal.
What does not
Conversion tables. Absolutely marketplace-specific and category-specific. A table from one marketplace applied to another is meaningless.
Update cadence.
Category structures.
Whether rank is published at all. Several marketplaces do not expose a numeric rank, only badges or list positions.
Variant handling.
Identifier schemes.
The categories of system
Published numeric sales rank. The system this site mostly describes. Rich, trackable, and the basis of most rank analytics.
Bestseller lists without numbers. A top-N list, refreshed periodically. Position in the list is observable; anything below N is not. Useful for tracking the top of a category and blind to everything else.
Badges only. "Bestseller" or "popular" markers with no ordering. Binary signal, low resolution, and still informative as a state change.
Search position only. No sales rank at all; visibility must be measured by querying. Different measurement problem entirely, covered separately.
Internal seller metrics. Available to the seller, not to observers.
Practical differences worth knowing
Smaller marketplaces have shallower catalogues, so a given rank represents fewer units and rank moves more on single sales. Noise dominates sooner.
Regional marketplaces of the same operator frequently share a rank system but not a catalogue, so ranks are not comparable across them.
Marketplaces with heavy third-party competition on shared listings complicate seller-level attribution, because the listing's rank aggregates all sellers.
Marketplaces with strong first-party retail may treat their own products differently in ranking, which is rarely documented.
Building a multi-marketplace view
Track each separately. Do not merge rank series across marketplaces.
Compare direction and relative position, never absolute rank.
Use a per-marketplace basket so market-wide movement is controlled within each.
Where a unit view is needed, calibrate per marketplace or state that the comparison is directional only.
Record the marketplace on every observation, which sounds obvious and is the most common omission in a dataset that started with one marketplace and grew.
The comparison people want and cannot have
"Which marketplace is this product doing better on?"
Rank cannot answer it. Rank 5,000 on a marketplace with two million products and rank 5,000 on one with forty million represent very different sales.
What can answer it: your own sales data per marketplace, obviously. For competitors, nothing observable.
What is available: relative position within each marketplace against a local basket, which supports "they are stronger relative to their local competition here than there" — a different and weaker claim, and an honest one.
Where to start on an unfamiliar marketplace
Find out what ranking signals are exposed and at what granularity.
Establish the noise floor for a few products before analysing anything.
Build a local basket.
Do not import assumptions from the marketplace you know. The decay rate, the update cadence, the category depth and the conversion are all different, and the first three are discoverable by observation within a few weeks.
A checklist for an unfamiliar marketplace
Ten questions to answer before analysing anywhere new.
Is a numeric sales rank published, or only lists and badges?
Overall, category, or both?
How deep does the published ranking go?
What is the effective update cadence, measured rather than assumed?
How large is the catalogue, roughly?
Are listings shared between sellers?
How are variants handled?
What identifier scheme is used?
Is there an official API, and what does it cover?
Do the terms permit what you intend to do?
Answer all ten before writing any analysis. Importing assumptions from a familiar marketplace is the standard way an international expansion produces confident wrong conclusions in its first quarter.