Marketing Measurement Glossary

Clear definitions for the terms that matter.

A B C D G H I L M R S U

A

Attribution Model: A rule or algorithm for assigning credit for conversions to the touchpoints in a customer's journey. Attribution models range from simple rules (last-click, first-click) to data-driven models built on conversion path data. All attribution models are approximations, they assign credit based on correlation between touchpoints and conversions, not causation.
Adstock / Carryover: The lagged effect of advertising spend: exposure to an ad today can influence purchases days or weeks later. Adstock is the modeled version of this decay, in MMM, ad spend is typically transformed using an adstock function before entering the regression. The shape and duration of the decay curve is a key assumption in any MMM model.

B

Baseline (MMM): In media mix modeling, the baseline is the revenue or conversions you would generate without any advertising. It represents recurring customers, organic search, direct traffic, and word-of-mouth. Separating baseline from media-driven revenue is the core task of MMM, and the baseline estimate is often the most contested output of the model.

C

Conversion Window: The time period after an ad interaction during which a conversion is attributed to that ad. A 7-day click conversion window means that if someone clicks your ad and converts within 7 days, the conversion is attributed to that click. Longer windows capture more conversions but increase the risk of false attribution, the person may have converted regardless of the ad.

D

Data-Driven Attribution (DDA): A model that uses machine learning to assign credit to touchpoints based on observed conversion paths, rather than fixed rules. DDA compares paths that converted against paths that didn't and identifies which touchpoints were associated with higher conversion rates. Despite the name, DDA can only see tracked digital touchpoints and cannot prove causation, it identifies correlation, not incremental lift.

G

Geo Experiment: An incrementality test that uses geographic markets as the unit of randomization. Some markets run advertising as normal (treatment); others have advertising paused or reduced (control). Outcomes are compared between groups using your own data, not platform attribution. Geo experiments are the gold standard for measuring channel incrementality because they're not dependent on platform measurement.
Ghost Ads: A test methodology where the control group sees a placeholder ad (often a public service announcement) rather than no ad at all. This controls for the effect of being shown an ad, isolating whether it was the specific ad that caused the outcome, rather than the mere presence of advertising. More rigorous than simple holdout tests but logistically complex.

H

Holdout Test: An incrementality test where a random subset of the target audience is withheld from seeing ads. The conversion rate of the holdout group (who saw no ads) is compared to the exposed group to measure incremental lift. Holdout tests can be run at the user level (through platform tools) or at the geographic level (geo holdouts). The latter is more trustworthy for measuring true incrementality.

I

Incrementality: The causal impact of advertising, the lift in conversions, revenue, or other outcomes that can be attributed specifically to your advertising activity, as opposed to behavior that would have happened regardless. Incrementality is always measured relative to a counterfactual: what would have happened without the ads? Platform-reported ROAS does not measure incrementality; it measures correlation between ad exposure and conversion.
iROAS (Incremental ROAS): Return on ad spend calculated using only the incremental revenue driven by advertising, not total attributed revenue. If a campaign reported $500,000 in attributed revenue and cost $100,000, the attributed ROAS is 5x. If an incrementality test shows that only $150,000 of that revenue was truly caused by the ads, the iROAS is 1.5x. The gap between attributed ROAS and iROAS is often large, especially for retargeting.

L

Last-Click Attribution: An attribution model that gives 100% of conversion credit to the last tracked touchpoint before a purchase. It's the default in most analytics platforms and systematically overcredits bottom-of-funnel channels (branded search, retargeting) while undercrediting upper-funnel channels (display, awareness campaigns) that initiated the customer journey but don't appear as the final touchpoint.

M

Marketing Mix Modeling (MMM): A statistical technique that uses regression analysis of aggregate time-series data to estimate the contribution of each marketing channel to business outcomes. MMM doesn't require user-level tracking, making it privacy-safe and immune to cookie deprecation. It's best suited for strategic budget allocation and measuring long-run effects. It requires at least 18-24 months of data to produce reliable estimates.
Marketing Triangulation: A measurement framework that combines media mix modeling (MMM), incrementality testing, and attribution to produce a more complete and robust picture of advertising effectiveness. Named after the navigation technique of using multiple reference points to find an exact position. Where the three methods agree, conclusions are more reliable. Where they diverge, there's something worth investigating.
Media Mix: The allocation of advertising budget across channels and formats, the combination of TV, digital, out-of-home, paid social, search, and other channels used in a marketing plan. Media mix optimization is the process of finding the allocation that maximizes return on investment, typically informed by MMM and incrementality testing.
Multi-Touch Attribution (MTA): Any attribution model that distributes conversion credit across multiple touchpoints in a customer journey, rather than assigning 100% to a single touchpoint. Multi-touch models range from simple rules (linear, time-decay, position-based) to data-driven models. All multi-touch attribution faces the same fundamental limitation: it can only see trackable digital touchpoints and cannot prove causation.

R

ROAS (Return on Ad Spend): Revenue attributed to advertising divided by advertising cost. Often reported by ad platforms using their own attribution. Platform-reported ROAS is almost always higher than the true incremental ROAS because platforms attribute credit using last-click or self-serving models and don't account for organic conversions that would have happened regardless of the ad.

S

Saturation Curve: In MMM, the mathematical relationship between advertising spend and the resulting effect, specifically, the diminishing returns that occur as spend increases. The saturation curve shows the point at which additional advertising spend produces progressively smaller incremental returns. Understanding the shape of this curve is essential for budget optimization: you want to spend up to the point of good returns, not past it.

U

Unified Marketing Measurement (UMM): An approach that attempts to combine multiple measurement methodologies, typically MMM, MTA, and incrementality testing, into a single framework. UMM aims to leverage the strengths of each method while correcting for their individual weaknesses. In practice, the integration is technically complex and the results depend heavily on the quality of each underlying method. UMM is not a replacement for good measurement fundamentals.