How NextToolHub ranking works, explained with examples (and what it ignores)

Why one product is above another: the weights, worked through three real situations — a search query, the deals page and the home page — plus the signals that carry zero weight, on purpose.

The How we rank page publishes our formula. This article does what a formula cannot: walks through why product A is above product B in three everyday situations, with the actual weights, and shows how a change in one signal moves a result. If you want the policy, read the page; if you want to understand it, read this.

The five signals and their weights

With a search query, a product's score is 45 % primary relevance (does the query match the title's main subject?) + 35 % relevance (matches anywhere in title, brand, category) + 20 % 'organic' quality. The organic part is itself: 35 % data quality (image, brand, category, price present and consistent), 25 % user demand (how often searches return the product and it is opened), 20 % recomputed discount, 20 % popularity (store-button clicks).

Without a query — deals, home, a category browse — there is no relevance to measure, so the organic score is the whole score. Commission rate, affiliate programme, and whether a store is one of our partners: weight zero. The signal is still computed and shown in our admin tools for transparency, but it never enters the sum.

Example 1: a search for 'mechanical keyboard'

Product A: 'Mechanical Keyboard 87-key hot-swap' — primary relevance 1.0, relevance 1.0, organic 0.55. Score = 0.45 + 0.35 + 0.11 = 0.91. Product B: 'Keyboard wrist rest for mechanical keyboards' — primary relevance 0.3 (the subject is a wrist rest), relevance 0.9, organic 0.80. Score = 0.135 + 0.315 + 0.16 = 0.61.

B has better data and more clicks, yet A wins comfortably, because relevance is 80 % of the score with a query. This is deliberate: a popular accessory should not outrank the thing you searched for.

Example 2: the deals page

Deals are gated before they are ranked. A product needs a real list price in the feed, a discount recomputed from that list price and the current price (a 6× ratio is treated as a feed error, not a bargain), data checked within the last seven days, and — if the discount is above 50 % — a manual approval. Only then does the organic score order the page.

So a '−70 %' listing with no approval simply is not there, while a −28 % listing with a good image, a known brand and steady clicks can top the page. The page is trying to show discounts you can trust, not the biggest number.

Example 3: the home page tiers

The home page is three tiers in a fixed order: Editor's Picks (chosen by a person, marked as such), then Popular (organic score with popularity and demand doing most of the work), then Recently checked (fresh data first, so new arrivals are visible before they have any clicks).

Nothing in any tier is paid. If we ever sell a placement it will sit in its own labelled slot outside these tiers — the policy is on the How we rank page and the affiliate disclosure.

What 'Verified' means in a ranking

'Verified product data' means the price and stock came from a store feed we imported and checked at the time shown, and the listing passed our link check. It raises the data-quality signal. It does not mean we vouch for the store, that the price is the lowest anywhere, or that the store is a partner — every store on the site is reached through an affiliate network (Admitad, CJ Affiliate or the eBay Partner Network) and treated the same.

Ranking is not neutral in an absolute sense: it reflects our choices about which signals matter. What we promise is that those choices are published, tested, and never include who pays us.

Method

Every number here is taken from the ranking code as of September 2026 (search-affiliate-catalog function, rankingSignals). The examples use made-up products with plausible signal values so the arithmetic is transparent. Our tests fail if the published weights drift from the code, so when the formula changes this article's date changes with it.

Verdict

With a query, relevance is 80 % of the score: the thing you searched for beats a more popular accessory.

Deals are filtered by trust rules first and ranked second; the biggest discount is not automatically on top.

Commission has zero weight, 'Verified' is about data not endorsement, and the whole formula is public and test-guarded.

Questions people ask

Can a store pay to rank higher?

No. There is no mechanism for it in the ranking code, and a test in our repository fails if a commission term is introduced. Sponsored placements, if we ever add them, are separate labelled slots.

Why is a product with more clicks below one with fewer?

Popularity is one fifth of the organic score, and the organic score is one fifth of a query search. Relevance and data quality usually decide; popularity breaks ties.

How do I see why a product is 'Verified'?

Hover or tap the Verified badge on a product page for the definition, or open the How we rank page section 'Verified product data vs official partnership'.