The terms we use, explained.
Plain definitions for the words we use across the site, each with a link to where it is used.
User Barcode™
A pseudonymous fingerprint of how one customer decides: dimensions such as occasion sensitivity, budget discipline and repeat cadence. It carries no name, address or payment detail, and it cannot be reversed into a person by us. Re-identification is your key to hold.
Where it is usedProduct DNA
The structured, validated attributes of one product: the facts every engine reads. A fact corrected once improves every module that depends on it, which is why catalog work compounds.
Where it is usedProduct Intelligence Score
A 0–100 readiness number per product, with the missing fields named rather than implied. 84 means structured and comparable; the gap to 100 is a specific list of absent attributes.
Where it is usedReason
The short human-readable string returned beside the score, assembled from the exact dimensions that produced it. It cannot disagree with the decision it describes because it is built from the same inputs.
Where it is usedCounterfactual
The companion to a reason: what would have changed the answer. “If the budget ceiling had been higher, the premium tier wins.”
Where it is usedHeld back
A request that returns no recommendation because confidence fell below the floor, consent did not cover the signals, or the catalog record was incomplete.
Where it is usedConfidence floor
The minimum score at which the system is willing to answer (0.60 by default). Raising it trades coverage for certainty; lowering it does the opposite.
Where it is usedHoldout
A share of traffic deliberately shown the status quo so lift can be measured rather than asserted. Pilots run 50/50; production drops the control to 10%.
Where it is usedIncremental vs influenced
Incremental is the difference against the holdout. Influenced is every outcome a decision touched. Only the first is called lift on this site.
Where it is usedEngine
One service in the decision path with a fixed contract. Ten run in a fixed order on every request, grouped into four roles: Signal Reader, User DNA / Need State, Fit Score, and Reasoning + Routing.
Where it is usedModule
What you buy: Catalog Intelligence, Product Intelligence Score, Content & GEO Studio, AI Visibility Monitor, Conversational Commerce, Revenue Intelligence. Modules are commercial; engines are computational.
Where it is usedSurface
Where a decision is delivered: storefront, assistant, API direct, console. The same core answers all of them, which is why the answer does not change between channels.
Where it is usedTier
Launch, Expand or Enterprise. A tier sets which modules are included and the support commitment.
Where it is used