Glossary/Measurement & Attribution

• A/A Test

  An experiment showing identical versions to two groups to verify the testing tool splits traffic and measures results correctly.

• Adstock (Carryover Effect)

  An MMM transformation capturing how an ad's impact lingers and decays over the weeks after it runs, rather than ending when spend stops.

• Anomaly Detection

  Automatically flagging data points that deviate sharply from expected patterns, such as sudden traffic drops or tracking breaks.

• Attention Metrics

  Measures of how much human attention an ad actually received, beyond viewability.

• Attribution

  Assigning credit for a conversion to the marketing touchpoints that influenced it.

• Attribution Window (Lookback Window)

  The period after an ad interaction during which a conversion is credited to it.

• Average Engagement Time

  In GA4, the average time a site or app was in the foreground or focus per user or per session.

• Average Order Value (AOV)

  Total revenue divided by the number of orders over a period.

• Baseline Sales (Base Sales)

  The sales a brand would generate without current marketing activity, driven by brand equity, seasonality, distribution and pricing, as estimated in MMM.

• Bayesian A/B Testing

  An approach to experiments that reports the probability each variant is best and the expected gain, rather than a p-value.

• Bayesian MMM

  A marketing mix model that combines prior beliefs with observed data to produce probability distributions, rather than single point estimates, for channel effects.

• Blended CAC

  Total sales and marketing cost divided by all new customers acquired, including organic ones, rather than only paid-attributed customers.

• Bounce Rate

  The share of sessions where a visitor leaves without meaningful engagement; in GA4 it is the percentage of sessions that were not engaged sessions.

• Brand Lift

  The measured change in awareness, consideration, or perception caused by a campaign.

• Budget Optimization (Scenario Planning)

  Using model outputs such as MMM response curves to simulate spend scenarios and reallocate budget across channels for the best expected return.

• CAC Payback Period

  The number of months it takes for the gross profit from a new customer to recover the cost of acquiring them.

• Causal Inference

  Statistical methods used to determine whether marketing actually caused an outcome, rather than merely being correlated with it.

• Churn Rate

  The percentage of customers or revenue lost over a given period.

• Click ID (GCLID / FBCLID)

  A unique parameter platforms append to ad click URLs, such as Google's gclid or Meta's fbclid, used to tie conversions back to specific clicks.

• Click-Through Rate (CTR)

  Clicks divided by impressions, expressed as a percentage.

• Closed-Loop Measurement

  Linking ad exposure directly to verified sales data, common in retail media.

• Cohort Analysis

  Grouping customers by a shared trait, usually acquisition date, and tracking their behavior over time to compare retention, revenue and value.

• Confidence Interval

  A range of values that likely contains the true effect of a test, expressing the uncertainty around a measured lift.

• Contribution Margin

  Revenue minus variable costs such as product, shipping and payment fees, showing what each sale contributes toward fixed costs and profit.

• Conversion API (CAPI)

  A server-to-server connection that sends conversion events directly to ad platforms.

• Conversion Rate

  The percentage of visitors or clicks that complete a desired action.

• Correlation vs. Causation

  The principle that two metrics moving together does not prove one causes the other; establishing cause requires experiments or causal methods.

• Cost Per Acquisition (CPA)

  The cost to drive one conversion, such as a sale or signup.

• Cost Per Click (CPC)

  The amount an advertiser pays for each click on an ad.

• Cost Per Incremental Acquisition (CPIA)

  Ad spend divided by the number of conversions proven to be caused by the campaign, rather than all attributed conversions.

• Cost Per Mille (CPM)

  The cost of 1,000 ad impressions.

• CUPED

  Controlled-experiment Using Pre-Experiment Data, a variance-reduction technique that uses prior user behavior to reach significant test results faster.

• Custom Dimension

  A user-defined attribute registered in an analytics tool, such as GA4, so a custom event parameter or user property can be reported on.

• Customer Acquisition Cost (CAC)

  The total cost to acquire one new customer.

• Customer Lifetime Value (CLV / LTV)

  The total revenue or profit a customer is expected to generate over the relationship.

• Dark Social

  Sharing through private channels such as messaging apps, email and DMs that analytics tools typically record as direct traffic.

• Dashboard

  A visual display that brings key metrics together in one view so teams can monitor performance at a glance.

• Data Sampling

  Analyzing a subset of data to estimate results for the whole, which speeds reports but can reduce accuracy on detailed queries.

• Data Visualization

  Presenting data as charts, graphs, and maps so patterns, trends, and outliers are quick to understand.

• Data-Driven Attribution

  Algorithmic attribution that assigns credit based on observed conversion paths.

• Deduplication (Event Deduplication)

  Removing duplicate conversion records, such as the same purchase sent by both a browser pixel and a server-side API, so it is counted once.

• Default Channel Group

  GA4's rule-based classification of traffic into channels like Organic Search, Paid Social, Email, and Direct, based on source and medium.

• Difference-in-Differences (DiD)

  A method estimating impact by comparing the change over time in a treated group against the change over the same period in a comparable control group.

• Diminishing Returns

  The point where each additional dollar spent on a channel delivers less incremental result than the dollar before it.

• Direct Traffic

  Visits analytics tools can't attribute to a referring source, including typed URLs, bookmarks, and untagged links from apps or email.

• Effective Frequency

  The number of exposures a person needs to see an ad for it to achieve its intended effect, balancing impact against wasted repetition.

• Engaged Session

  A GA4 session that lasts at least 10 seconds by default, includes a key event, or has two or more page or screen views.

• Engagement Rate

  In GA4, engaged sessions divided by total sessions; on social platforms, interactions such as likes, comments and shares divided by reach or impressions.

• Enhanced Conversions

  A Google Ads feature that sends hashed first-party customer data, such as email, with conversion tags to improve conversion matching accuracy.

• Event Match Quality (EMQ)

  Meta's 0–10 score rating how well customer information sent with server or pixel events can be matched to Meta accounts.

• Event Parameter

  Extra detail sent with a GA4 event, such as page title, item ID, or value, describing the context of that interaction.

• Event-Based Tracking

  An analytics model, used by GA4 and product analytics tools, that records each user interaction as a named event with parameters rather than relying on pageviews and sessions.

• Exit Rate

  The percentage of views of a page that were the last in a session, showing where visitors leave the site.

• First-Click Attribution

  Giving 100% of conversion credit to the first recorded touchpoint.

• Flicker Effect

  A brief flash of original content before a client-side test variant loads, which can bias results and hurt user experience.

• Frequentist Testing

  The classic hypothesis-testing approach that judges results by p-values and confidence intervals, usually requiring a fixed sample size set in advance.

• Funnel Analysis

  Measuring how many users move through each step toward a goal, such as visit, add to cart and purchase, to locate where they drop off.

• Geo Experiment (Matched Market Test)

  An experiment that varies spend across regions to measure incremental impact.

• Ghost Ads

  A lift-measurement method that records when control-group users would have seen an ad without serving it, avoiding the cost and bias of PSA ads.

• Google Analytics 4 (GA4)

  Google's event-based analytics platform that replaced Universal Analytics.

• Gross Rating Point (GRP)

  Reach multiplied by average frequency, a traditional TV planning metric.

• Guardrail Metric

  A metric monitored during an experiment to ensure a change doesn't harm important outcomes, such as revenue, latency, or unsubscribes.

• Halo Effect

  Indirect impact where marketing for one product, channel or brand lifts results for another, such as TV advertising boosting branded search.

• Heatmap

  A visual overlay showing where visitors click, move, or scroll on a page, with warmer colors marking more activity.

• Holdout Group (Control Group)

  A randomly selected group withheld from exposure to measure true impact.

• Impression

  A single instance of an ad being served or displayed.

• Incremental Reach

  The additional unique audience a channel or campaign reaches beyond people already reached by other media in the plan.

• Incremental ROAS (iROAS)

  Revenue caused by advertising, divided by ad spend, excluding sales that would have happened anyway.

• Incrementality

  The lift in an outcome caused by marketing that would not have happened otherwise.

• Key Event

  GA4's name, since 2024, for an event marked as important to the business; 'conversion' now refers to key events shared with Google Ads.

• Key Performance Indicator (KPI)

  A measurable value chosen to show how effectively a campaign, team or business is achieving its key objectives.

• Last-Click Attribution

  Giving 100% of conversion credit to the final click before purchase.

• Lift Study

  A test comparing exposed and unexposed groups to measure an ad's causal effect.

• Linear Attribution

  A multi-touch model that gives equal conversion credit to every recorded touchpoint in the customer journey.

• LTV:CAC Ratio

  Customer lifetime value divided by acquisition cost; a core efficiency benchmark.

• Margin of Error

  The range above and below an estimate within which the true value likely falls, at a given confidence level.

• Marginal ROAS (mROAS)

  The revenue generated by the next additional dollar of spend in a channel, used to decide where to add or cut budget.

• Marketing Efficiency Ratio (MER)

  Total revenue divided by total marketing spend across all channels.

• Marketing Mix Modeling (MMM)

  Statistical modeling of aggregate sales data to estimate each channel's contribution and optimal budget.

• Marketing Qualified Lead (MQL)

  A lead whose engagement or fit indicates readiness for further nurturing or sales follow-up, according to criteria marketing and sales agree on.

• Marketing-Sourced Pipeline

  The value of sales opportunities whose origin is credited to marketing activity, a common B2B measure of marketing's revenue contribution.

• Measurement Plan

  A document defining business objectives, KPIs, the events and dimensions to track, and how data will be collected and reported.

• Media Rating Council (MRC)

  The US body that audits and accredits advertising measurement services.

• Meridian

  Google's open-source Bayesian marketing mix modeling framework, released in 2024–2025 as the successor to LightweightMMM, supporting geo-level data and experiment calibration.

• Minimum Detectable Effect (MDE)

  The smallest lift a test is designed to reliably detect, given its sample size, baseline rate, significance level and statistical power.

• Model Calibration (MMM Calibration)

  Adjusting or constraining a marketing mix model using results from lift tests or experiments so its channel estimates reflect measured incrementality.

• Modeled Conversions

  Conversions estimated by platforms using machine learning to fill gaps when consent, cookies or device tracking prevent direct observation.

• Multi-Armed Bandit

  An adaptive testing method that shifts traffic toward better-performing variants during a test, trading some statistical certainty for less wasted traffic.

• Multi-Touch Attribution (MTA)

  Attribution that distributes conversion credit across multiple touchpoints in a journey.

• Multivariate Testing (MVT)

  An experiment that tests combinations of changes to several page or ad elements at once to find the best-performing combination.

• Net Promoter Score (NPS)

  A loyalty metric from asking how likely customers are to recommend a brand on a 0–10 scale, calculated as promoters' percentage minus detractors' percentage.

• Net Revenue Retention (NRR)

  Recurring revenue kept from an existing customer group over a period, including expansion and subtracting churn and downgrades, expressed as a percentage.

• North Star Metric

  The single metric a company believes best captures the core value it delivers to customers and predicts long-term growth.

• Novelty Effect

  A temporary performance change in a test caused by users reacting to something new rather than its lasting value.

• Null Hypothesis

  The default assumption in a statistical test that there is no real difference or effect between groups.

• Offline Conversion Import (OCI)

  Uploading conversions that happen offline or in a CRM, like closed deals or store sales, back to ad platforms to measure and optimize campaigns.

• P-Value

  The probability of seeing a result at least as extreme as the observed one if there were truly no difference between test groups.

• Peeking

  Repeatedly checking test results and stopping as soon as they look significant, which inflates the false-positive rate in fixed-horizon tests.

• Position-Based Attribution (U-Shaped)

  A multi-touch model that gives most credit, commonly 40% each, to the first and last touchpoints and splits the remainder among middle interactions.

• Post-Purchase Survey

  A short survey shown after checkout that asks buyers about their discovery path or motivations, commonly used for self-reported attribution.

• Product Analytics

  Tools and practices, such as Amplitude, Mixpanel or PostHog, that analyze how users behave inside a digital product to improve activation, engagement and retention.

• Profit on Ad Spend (POAS)

  Gross profit generated by advertising divided by ad spend, a margin-aware alternative to ROAS.

• PSA Test

  A lift test where the control group is served an unrelated public service announcement instead of the brand's ad, isolating the ad's effect on behavior.

• PyMC-Marketing

  An open-source Python library from PyMC Labs offering Bayesian marketing mix models and customer lifetime value models.

• Reach and Frequency

  Reach is the number of unique people exposed; frequency is how often each saw the ad.

• Referral Traffic

  Visits that arrive by clicking a link on another website, excluding search engines and social platforms in most channel groupings.

• Regression Analysis

  A statistical method estimating how one or more variables, such as ad spend, relate to an outcome, such as sales.

• Retention Rate

  The percentage of customers who remain active or keep buying over a given period.

• Return on Ad Spend (ROAS)

  Revenue generated for every dollar spent on advertising.

• Return on Investment (ROI)

  Profit generated relative to the total cost of an investment.

• Robyn

  Meta's open-source, R-based marketing mix modeling package that uses ridge regression and automated hyperparameter search to estimate channel contributions.

• Sales Qualified Lead (SQL)

  A lead the sales team has vetted and accepted as a genuine potential buyer worth active pursuit.

• Sample Ratio Mismatch (SRM)

  When the actual traffic split between test groups differs significantly from the planned split, signaling a flaw that invalidates results.

• Sample Size Calculation

  Determining how many users, conversions or regions a test needs to detect a given effect with the desired confidence and power.

• Saturation Curve (Response Curve)

  A curve in MMM showing how outcomes respond to increasing spend in a channel, flattening as additional dollars produce smaller gains.

• Seasonality

  Predictable recurring patterns in demand or performance tied to time of year, week, or holidays, which analyses must account for.

• Self-Reported Attribution

  Asking customers directly, often via a 'How did you hear about us?' question, which channel influenced them, capturing sources like podcasts and word of mouth.

• Sequential Testing

  A testing approach with statistical rules that allow results to be monitored continuously and stopped early without inflating error rates.

• Server-Side Testing

  Running experiments by assigning variants on the server before a page loads, avoiding flicker and enabling tests of pricing or algorithms.

• Session Replay

  A reconstructed playback of a visitor's real interactions on a site, such as clicks, scrolls, and typing, used to diagnose UX problems.

• Shapley Value Attribution

  An algorithmic attribution method from cooperative game theory that credits each channel by its average marginal contribution across all channel combinations.

• Share of Search

  A brand's share of branded search volume in its category, a leading indicator of market share.

• Share of Voice (SOV)

  A brand's share of total advertising or conversation in its category.

• Signal Loss

  The decline in observable user-level data for targeting and measurement caused by privacy rules, consent choices, browser restrictions and platform changes.

• Split URL Testing

  An A/B test that sends visitors to entirely separate URLs for each variant, suited to testing major page redesigns.

• Statistical Power

  The probability that a test detects a real effect when one exists, conventionally targeted at 80%.

• Statistical Significance

  An indication that an observed difference in a test is unlikely to be explained by random chance, usually judged against a preset threshold like 95%.

• Synthetic Control

  An experimental method that builds a weighted blend of untreated regions or units to mimic the test group and estimate what would have happened without treatment.

• Time-Decay Attribution

  A multi-touch model that gives more conversion credit to touchpoints closer in time to the conversion.

• Tracking Pixel

  A small piece of code that records a page view or event for measurement or targeting.

• Triangulation (Measurement Triangulation)

  Combining MMM, multi-touch attribution and incrementality experiments so each method checks the others and produces a more reliable view of performance.

• Type I Error (False Positive)

  Wrongly concluding a test variant had a real effect when the observed difference was due to chance.

• Type II Error (False Negative)

  Failing to detect a real effect in a test, often because the sample was too small or the test underpowered.

• Unified Measurement

  Combining MMM, MTA, and experiments into one view of marketing performance.

• Unit Economics

  The revenues and costs associated with a single unit, such as one customer or order, used to judge whether growth is profitable.

• Uplift Modeling

  Predictive modeling that estimates how much a treatment like an offer or ad changes each customer's behavior, so marketers target persuadable people.

• UTM Parameters

  Tags appended to a URL to identify the source, medium, and campaign of traffic.

• Vanity Metrics

  Numbers that look impressive, like follower counts or raw impressions, but do not reliably connect to business outcomes.

• View-Through Conversion

  A conversion that happens after someone saw an ad but did not click it.