Paid acquisition glossary
Definitions for the terms that actually change how you read ad performance data, written the way we use them internally, not textbook-generic.
Attribution window
An attribution window is the length of time a platform will credit a conversion back to an ad interaction, for example Meta's default 7-day-click, 1-day-view. Change the window and the same underlying conversions redistribute across campaigns differently, which means a reported performance swing after a window change is a measurement artifact, not a real change in results.
iOS ATT tracking loss
iOS ATT (App Tracking Transparency) tracking loss is the gap between conversions that actually happened and conversions a platform can observe, caused by iOS users declining the tracking-permission prompt Apple requires since iOS 14.5. Declined users' activity is modeled or estimated rather than directly observed, which systematically under-reports true performance on iOS traffic and inflates apparent cost-per-result.
MMM vs. last-click attribution
Last-click attribution credits a conversion entirely to the final ad interaction before it happened; MMM (Marketing Mix Modeling) instead estimates each channel's incremental contribution using aggregate spend and outcome data over time, without relying on individual user-level tracking. Last-click is simple and real-time but systematically over-credits bottom-funnel and retargeting; MMM is privacy-resilient and captures upper-funnel effects, but runs on a lag and needs real statistical rigor to trust.
Wasted spend
Wasted spend is ad spend on search queries, placements, or audience segments that have a demonstrated pattern of not converting, most commonly irrelevant search terms triggering an ad through broad or phrase match. It's identified by matching actual query-level (or placement-level) spend against a negative-keyword or exclusion audit, and reported as a range over a specific date window, never as a single lifetime figure.
Quality Score
Quality Score is Google Ads' 1-10 estimate of how relevant a keyword, ad, and landing page are to a searcher, built from expected click-through rate, ad relevance, and landing page experience. It directly affects both ad rank and cost-per-click, but it's a point-in-time diagnostic snapshot, not a metric to sum or average across a date range the way spend or clicks can be.
Conversion lag (IBNR)
Conversion lag is the delay between an ad click and the conversion event Google eventually reports for it, which means any recent date window is undercounted at the moment you look at it and gets backfilled for weeks afterward. IBNR (incurred but not reported, a term borrowed from insurance) describes exactly this: real conversions that already happened but haven't been reported yet, so a raw recent-window count always understates the true number.
Impression share and lost IS (budget vs. rank)
Impression share is the percentage of eligible auctions your ad actually appeared in; 'lost IS' splits the missing percentage into two separate causes: lost to budget (your campaign ran out of budget before the day ended) and lost to rank (your ad quality/bid combination lost the auction outright). The two require completely different fixes, so treating them as one number leads to the wrong action almost every time.
Performance Max structure
Performance Max (PMax) is Google's automated campaign type, and structurally it has no ad groups and no keywords, unlike a Search campaign. You instead build 'asset groups' (headlines, images, video, text) and let Google's system decide where to show them and what to match them against. That means you can't see which search terms or keywords actually triggered a given result, so troubleshooting has to work through broader signals like asset group performance and audience signals, not keyword-level cause and effect.
Ad-platform settings that aren't real levers
Both Google Ads and Meta Ads show you numbers and labels that look like controls but mostly aren't. Google's Ad Strength and Optimization Score are diagnostic labels describing how the system rates your setup; changing them requires you to edit something else (headlines, targeting) that then produces a new score. Genuine direct levers are things like bid amount, budget, targeting settings, and which conversion action you're optimizing for. Confusing a score for a dial is a common reason people think they've 'tried everything' when they've mostly just read a report card.
'Conversions' (why it's not one number)
Both Google Ads and Meta Ads report a metric literally labeled 'Conversions,' but it can mean different things depending on account settings and where you're looking. Google distinguishes 'all conversions' from the subset marked 'primary' (the ones bidding algorithms actually optimize toward), and Meta reports conversions broken out by individual action type, purchase, lead, add-to-cart, each counted separately. Comparing two 'conversions' figures without checking whether both are using the same definition and the same underlying action is a routine, easy-to-make error.
Cross-campaign conversion attribution (Performance Max)
A Performance Max campaign's reported conversion count is not guaranteed to reflect conversions that campaign alone drove. Google's automated bidding and attribution systems use signals and assist paths that can span multiple campaigns you're running at once, and standard per-campaign reporting doesn't fully separate out that overlap. This means two campaigns can each report conversions that were partly influenced by the other, so comparing their standalone numbers, or cutting one because its number looks weak, can be comparing figures that were never fully independent.
Automated changes attributed to "system"
Both Google Ads' Change History and Meta's equivalent change log can show edits attributed to a generic "system" or "Google Ads" actor instead of a named team member. This happens because automated bidding, auto-apply recommendations, and other platform-side automation make their own micro-adjustments to bids, budgets, or targeting on your behalf, and those changes get logged the same way a human edit would. Anyone auditing the log to reconstruct "what did we change vs what happened on its own" needs to know this distinction exists, or they'll misattribute automated moves as deliberate decisions, or the reverse.
Absolute top impression share
Absolute top impression share is the proportion of your eligible Google Search impressions that appeared in the very first ad slot, above the organic results. The share you did not get is split into two reported figures: lost to rank and lost to budget. It measures position, not profit, and a campaign can hold the top slot consistently at a cost per sale the business cannot afford.
ROAS (return on ad spend)
ROAS (return on ad spend) is revenue divided by ad spend over the same period. The figure every ad platform shows uses revenue the platform attributed to itself, which is not the same as revenue the business actually banked. A 4x platform ROAS and a 4x accounting ROAS are two different claims, and on most accounts only the first one has ever been checked.
CPA (cost per acquisition)
CPA (cost per acquisition) is ad spend divided by the number of conversions the platform counted. Both halves are softer than they look: conversions is a set of actions somebody configured, and the count for any recent period keeps rising for weeks after the click. A CPA quoted without saying which actions it counts, and how complete its window is, is not comparable to any other CPA.
Learning phase
The learning phase is the period after a new or significantly edited ad set during which delivery is still unstable and results are not yet representative. Meta describes it as ending once an ad set has had roughly 50 optimisation events within a week. Leaving it is a delivery milestone, not a verdict: an ad set can exit the learning phase and still be losing money.
Broad match
Broad match lets Google show your ad for searches it judges related in meaning to your keyword, not only ones containing it. Google uses signals beyond the keyword itself, including your landing page, the other keywords in the ad group and the user's recent activity. It buys reach at the cost of control, and the only way to know what you bought is the search terms report.
View-through conversion
A view-through conversion is one credited after someone saw your ad without clicking it, within the platform's view window, which defaults to one day on Meta. It reflects something real, since ads can influence a purchase without a click. It is also the easiest place for a platform to credit itself with a sale it did not cause, because the qualifying event is simply that an impression was served.
Modeled conversions
Modeled conversions are conversions a platform estimates statistically rather than observes, filling the gap left by consent refusals, cross-device journeys and browser restrictions. Estimating is usually the honest response to a genuinely unobservable question. The problem is presentational: the estimates arrive in the same column as directly measured conversions, with nothing marking which rows are which.
Incrementality testing
Incrementality testing measures what advertising actually caused, by withholding it from a comparable group and comparing outcomes. It answers the one question attribution cannot: how many of these conversions would have happened anyway. Geo holdouts and budget-split tests are the usual formats, and the result is often uncomfortable for whichever channel the attribution model had been flattering.