Ranking

How the VIZNT Score works

The exact factors and weights behind every VIZNT recommendation — and why commission never moves them.

Weights

The default weights.

Weights shift by category — shipping matters more for a mattress than a pair of shoes — but this is the baseline.

0135%

Intent match

How closely the product matches the criteria VIZNT extracted from your request.

0220%

Feature & spec match

Whether the measurable specs clear the thresholds your request implies.

0315%

Price & value

Value at its current price relative to comparable products, and fit to your budget.

0410%

Review quality

Rating strength adjusted for how many reviews back it up.

057%

Availability

In stock now beats backordered, however good the product is.

065%

Shipping

Speed and cost of getting it to you, weighted more when you say it's urgent.

075%

Merchant quality

Return policy, warranty and reliability of the seller.

083%

Popularity

A light signal only — never enough to outrank a better fit.

Signed in // On top of the baseline
1840%

Personal fit

Signed in, this becomes one of the heaviest factors: the brands, categories and price tiers you actually click, plus the preferences you've set. Signed out, it's off entirely. When it applies, it takes that share of the score and the eight factors above are rescaled proportionally — the total is always 100%.

Baseline // WeightsTotal // 100%
Integrity

What never affects the score.

Every product, price and spec is retrieved from a connected source. AI interprets your request and explains the result — it never produces product facts. When VIZNT has no confident answer, it says so instead of guessing.

Commission rate

Merchant payouts are applied after ranking and can never reorder a result set.

Partner status

Being a VIZNT partner buys distribution, not position. Scores are blind to it.

Advertising spend

Nothing can be bought into the organic set. Sponsored slots are labelled and separate.

Invented facts

Products, prices and specs come from connected sources. AI explains; it never generates them.

Trust // ModelRetrieval // First