Glossary
One page per term, defined the way the industry actually uses it — including the ones two teams use to mean opposite things.
- Assisted conversion A touchpoint that appeared somewhere in the journey without being the last one. It records presence, not contribution — the two get read as the same thing constantly.
- Attribution model The rule that decides how credit for a conversion is divided among the touchpoints before it. Every model is a choice about what to reward, not a discovery about what happened.
- Attribution window The period after an ad interaction during which a later conversion is still credited to that interaction. Once it closes, the conversion is counted as coming from somewhere else.
- Click ID A unique identifier an ad platform appends to the landing page URL when someone clicks an ad, so a later conversion can be tied back to that exact click.
- Consent mode A mechanism that tells tags whether a visitor consented, so they adjust behaviour instead of firing or not firing. Consent is the permission; consent mode is the plumbing.
- Conversion An outcome you decided to count. The definition is yours, which is why two teams reporting "conversions" are often counting different things and comparing regardless.
- Conversion rate Conversions divided by something. The numerator is usually agreed and the denominator rarely is, which is why the same performance produces different rates in different tools.
- Cookie window How long a tracking cookie stays valid after it is set. Often confused with the attribution window, which is a reporting rule rather than a storage limit.
- Customer lifetime value What a customer is worth across the whole relationship rather than one purchase. Usually a prediction, often reported as though it were a measurement.
- Data clean room A controlled environment where two parties analyse their combined data without either seeing the other's raw records. Aggregate answers come out; row-level data does not.
- Deduplication Recognising that two records describe the same event, so one purchase reported twice is counted once. The routine failure mode of sending events from the browser and the server.
- First-party data Information you collected directly from your own customers through your own properties. Distinct from data bought or obtained through an intermediary.
- Hashing Converting a value into a fixed-length string that cannot be reversed, so two systems can check whether they hold the same value without either sending the value itself.
- Holdout test Deliberately withholding advertising from a comparable group so the difference in outcomes shows what the advertising actually caused. The method behind an incrementality claim.
- Identity resolution Deciding that separate records — a click, a session, an order, an email — belong to the same person. Everything downstream inherits whatever this gets wrong.
- Impression An ad being served. Not an ad being viewable, and not an ad being seen — three different things routinely reported under one word.
- Incrementality Whether a marketing activity actually caused conversions that would not have happened without it. The question attribution reports cannot answer.
- Last-click attribution Giving a conversion entirely to the final click before it. Still the most common model, and the one that most reliably misprices everything upstream.
- Marketing mix modelling A statistical model that estimates each channel's contribution from aggregate spend and outcome data over time, without tracking individuals. Often called MMM.
- Match rate The share of records you sent that a platform could tie to a known person. The single number that tells you whether an identity setup is working, and the first one to check when results look thin.
- Modeled conversions Conversions an ad platform estimates statistically rather than observes directly, used to fill gaps where consent, tracking prevention or privacy limits block measurement.
- PII Personally identifiable information — data that identifies a specific person. The category is narrower than most regulation, which is why "we removed the PII" is rarely the whole answer.
- Server-side tagging Sending marketing and analytics events from your own server instead of from the visitor's browser, so a request the browser might block is made somewhere it cannot be.
- Session A group of one visitor's activity treated as a single visit. The grouping is a rule each tool applies, not an event, which is why two tools rarely agree on the count.
- Signal loss The industry term for measurement data that no longer arrives — through consent refusal, browser restrictions and platform limits. It describes a change in visibility, not a change in behaviour.
- Statistical significance A statement about how likely a difference this large would be if there were no real effect. It is not a measure of how large, how important, or how certain the effect is.
- Tracking prevention Browser features that limit how long identifiers survive and which requests are allowed, applied by default to everyone. Separate from consent, and not something the visitor opted into.
- UTM parameters Tags you add to a URL to describe where a visit came from. They label the traffic for your own analytics; they do not identify the click to the ad platform.
- View-through conversion A conversion credited to an ad the person saw but never clicked. The weakest common attribution signal, and the one most likely to inflate a channel.
- Zero-party data Information a customer deliberately gave you about themselves — preferences, intent, stated needs — rather than something you observed. A marketing coinage, not a legal category.