MASS Analytics Glossary

A/B Testing
A controlled experiment involving two variables. Used extensively in digital to optimize messaging performance. lt is essential that all contextual factors, audience, content environment, time, etc. be perfectly matched to isolate the comparative effect of A versus B.
Addressable TV
Television platform able to serve different targeted ad content to different audience segments watching the same linear TV program on internet-enabled IVs and set-top boxes in live, playback, or VOD mode.
The prolonged or lagged effect of advertising on consumer purchase behavior. It is an essential model specification for capturing the full extent of advertising’s contribution.
A procedure or formula for solving a problem, based on conducting a sequence of specified calculations or steps. For example, a media optimizer uses an algorithm to sequentially add the next best medium to the plan.
The credit or value you assign to a given marketing touchpoint. Marketing attribution provides a quantifiable evaluation of your marketing efforts and their direct and traceable effects on revenue.
The intrinsic value and level of performance against a set of key performance indicators (KPIs) derived solely from a brand’s recognition in the marketplace. It provides an initial value from which to evaluate the impact of incremental marketing investments.
Bayesian Priors
In Bayesian statistical, a prior probability distribution, often simply called the prior, of an uncertain quantity is the probability distribution that would express one’s beliefs about this quantity before some evidence is taken into account. This enables facts taken from other sources to be imposed on a model. it also enables a model to work with data sources of different levels of granularity.
Behavioral Economics
Behavioral economics studies the effects of psychological, cognitive, emotional, cultural and social factors on the decisions of individuals and institutions and how those decisions vary from those implied by classical economic theory.
The California Consumer Privacy Act is a state statute intended to enhance privacy rights and consumer protection for residents of California, United States.
Collinearity is a situation in which the predictor, or exogenous, variables in a linear regression model are linearly related among themselves or with the intercept term, and this relation may lead to adverse effects on the estimated model parameters.
Cross-Channel Attribution
Attribution is the practice of tracking and valuing all marketing touch points consumers encounter along the path to conversion. The goal of Cross-Channel Attribution is to determine which channels and messages had the greatest impact on the decision to convert, or take the desired next step.
Cycle Time
The period required to complete one cycle of a modeling operation (ingesting new data, refitting the model, and providing an updated evaluation of campaign performance).
Data Management Platform (DMP)
A data warehouse technology that centralizes and deduplicates first, second and third-party data sources to generate audience segments that can be used for marketing and media audience targeting and creative optimization.
Data Matching
Data matching is the task of identifying, matching, and merging records that correspond to the same entities from several databases or even within one database. 
Dependent Variable
The dependent variable is the variable that is being measured or tested in an experiment. The dependent variable is dubbed dependent because it is thought to depend in some way on the variations of the independent variable.
Device Graphs
A cluster of browsers and devices that have been probabilistically connected using anonymous, non-PII data sources like ad server logs, cookie IDs, device IDs, browser fingerprints, and IP addresses from the open web, and are unified around individual, but anonymous users.
Device lDs
A device ID (device identification) is a distinctive number associated with a smartphone or similar handheld device. Device IDs are among one of the easiest ways to identify mobile users.
Device lDs
A string of numbers and letters that identifies every individual smartphone or tablet in the world. It is stored on a mobile device and can be retrieved by any app that is downloaded and installed.
Digital Video Advertising
Digital video advertising is a part of the digital advertising that displays video ads – either standalone or within online video content. This includes lots of formats designed for various environments, such as banner video, connected TV, or other OTT services.
Diminishing Returns
An economic law stating that if one input in the production of a commodity is increased while all other inputs are held fixed, a point will eventually be reached at which additions of the input yield progressively smaller, or diminishing, increases in output.
Discrete Choice
A discrete choice model specifies the probability that a person chooses a particular alternative, with the probability expressed as a function of observed variables that relate to the alternatives and the person.
The use of statistical methods using quantitative data to develop theories or test existing hypotheses in economics or finance. Econometrics relies on techniques such as regression models and null hypothesis testing. Econometrics can also be used to try to forecast future economic or financial trends.
Exogenous Factors
Also called external influences, they are factors that occur entirely beyond the marketers’ control but exert influence on the way advertising in a particular category behaves. These factors include: weather, consumer confidence, regulations, etc.
Exposed vs. Unexposed
A core component of measuring ad effectiveness that is s comparing performance metrics associated with those exposed to an ad – including number and time of exposures – versus those unexposed.
External Influences
Also known as exogenous factors, External influences occur entirely beyond the company’s control, but exert influence on the way advertising in a particular category behaves. These factors include changes in legislation, the environment, consumer confidence, etc.
First-Party Data
The data that your company has collected directly from your customers, site visitors, and social media followers. “First-party” refers to the party that collected the data firsthand to use for re-targeting.
The General Data Protection Regulation is a regulation in EU law on data protection and privacy in the European Union and the European Economic Area. It also addresses the transfer of personal data outside the EU and EEA areas.
Game Theory
The study of mathematical models of strategic interaction among rational decision-makers. It is used to formulate the alternative strategy to compete with one another and in the same sense, it is an essential tool for the decision-making process.
Granularity helps define the level of detail used in modeling situations or straight decision-making processes. The greater the granulation, the deeper the level of detail. Granularity is usually used to characterize the scale or level of detail in a set of data.
Hierarchical Bayesian
Bayesian hierarchical modeling is a statistical model written in multiple levels (hierarchical form) that estimates the parameters of the posterior distribution using the Bayesian method. The sub-models combine to form the hierarchical model.
Holdout Samples
Sample of observations withheld from the model fitting process. Model predictive validity can be estimated by its ability to predict the data, Sometimes the holdout is chosen for convenience, but a mixture of random and designed holdouts (to provide a set of specific situations) is preferred.
A measure of the number of times a media advertisement or marketing message is served. In offline media, impressions are a measure of the number of times an ad or message may have been seen.
A desired outcome (revenue, sales, leads, brand engagement, etc.) gained from a marketing activity that would not have been generated without that marketing activity.
Independent Variables
The variables that, in combination, predict the dependent, or outcome, variable (e.g., sales). They represent the causal factors that drive the outcome variable.
Information Privacy
In general, privacy is the right to be let alone, or freedom from interference or intrusion. Information privacy is the right to have some control over how your personal information is collected and used.
Key Performance Indicator (KPI)
A measurable value that demonstrates how effectively a company is achieving key business objectives. Organizations use KPIs at multiple levels to evaluate their success at reaching targets. High-level KPIs may focus on the overall performance of the business, while low-level KPIs may focus on processes in departments such as sales, marketing, HR, support, and others.
Linear Television
Traditional television programming that follows a schedule. Typically includes live viewing and time-shifting via DVR and VOD that includes the same ad load as the live telecast.
Logit Models
Better known as Logistic Regression is a binomial regression model, it is a special case of a generalized linear model that is widely used in machine learning to associate a vector of random variables to a binomial random variable.
Long-term Effects
Cumulative effect of advertising on consumers’ brand choice behavior, lasting over several years. Measures of loyalty to a brand or consideration set of brands. lt can also reflect customer lifetime values. Lacks consistent definition and, in some cases, long-term effects of digital advertising are measured on a “next quarter” basis.
Lookback Windows
The time span during which advertising is analyzed prior to a conversion. The period of time the model “looks back” at the ad exposures that may have contributed to a conversion.
lP Address
A unique string of numbers that identifies each connected device (e.g., smartphones, tablets, television sets) using the internet protocol to communicate over the internet.
Machine Learning
The use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyze and draw inferences from patterns in data.
Market Level ANCOVA
An analysis of covariance that assumes that the relationship between the independent variable and the dependent variable must be linear in nature. It evaluates whether the means of a dependent variable are equal across levels of a categorical independent variable.
Marketing Mix Models
Marketing mix modeling is statistical analysis such as multivariate regressions on sales and marketing time series data to estimate the impact of various marketing tactics on sales and then forecast the impact of future sets of tactics.
Media Interactions and Halos
A customers’ bias towards certain products due to their favorable experience with the other products made by the same manufacturer. The halo effect is driven by brand equity and accounted for in MMM.
Multi-Touch Attribution (MTA)
Multi-touch revenue attribution is the process of organizing, collecting, and cataloging all of the interactions that occur as an individual decides to make a purchase with your company.
Nested Modeling
A Nested Model depicts an immersive relationship between multiple linked KPIs and independent drivers. Two models are nested if one model contains all the variables of the other in addition to (an)other variable(s).
Non-Addressable Channels
Channels like broadcast TV, radio, print, out-of-home, in-store displays, etc., where marketing messages are delivered to individuals who cannot be identified at a user level.
Other Marketing Variables
Aspects of product marketing besides media and advertising that drive sales. Price, promotion, product features, in-store variables, competitive trade deals, and impact provide the full picture of marketplace pressure and consumer response. Models that do not include these factors fail to provide a holistic view and implicitly overstate the contribution of advertising.
Other Non-Marketing Data
Non-marketing data typically includes outside variables that influence category sales such as seasonality, weather, consumer confidence, fuel prices, holidays. etc.
Over-The-Top (OTT) Media
A media service offered directly to viewers via the Internet. OTT bypasses cable, broadcast, and satellite television platforms, the types of companies that traditionally act as controllers or distributors of such content. Services like Netflix, Hulu, or Disney+ are video OTT services.
Pooled Regression
A type of model that has constant coefficients, referring to both intercepts and slopes. Pooled regression is carried out when we have available time series of cross-sections i.e., data that has observations over time for several different groups or cross-sections.
Random Control Tests (RCT)
A randomized controlled trial (or randomized control Tests) is a type of scientific experiment that aims to reduce certain sources of bias when testing the effectiveness of a new statistical model or intervention.
Randomized Control Tests (RCT)
An experimental form of impact evaluation in which the population receiving the program or policy intervention is chosen at random from the eligible population, and a control group is also chosen at random from the same eligible population.
Refresh Time
Also called “Update intervals“, it is the maximum time interval between the original attribution model and updates with new data. Models are typically refreshed daily, monthly, quarterly, or run continuously. The same as cycle time.
A broad set of statistical techniques for estimating the relationships among variables. Regression helps determine how the typical value of the dependent variable (e.g. sales. conversions. etc.) changes when any one of the independent variables (Media Weight, Media Mix) is varied, while the other independent variables are held fixed.
Return On Investment (ROI)
Return on investment – a measure of profitability based on the incremental sales generated by advertising (or any other marketing factor) in relation to its cost. Used to determine the value of advertising. Specifically in financial terms, ROI represents the ratio between a marketer’s net profit from that investment and the cost of an advertising campaign (or media element within the plan).
Safe Haven Matches
Third-party services that use their proprietary device graph to match first-party advertiser data and other partners’ data. By scrubbing out anything that personally identifies end-users (personally identifiable information- PII) and replacing it with a generic but unique ID to use in multiple marketing applications.
Second-Party Data
Consumer data that is shared between trusted marketing partners. Examples include “intent to purchase” data that is shared between an airline and a hotel chain or “online shopping” data that is shared between a retailer and a manufacturer.
Subscription Video On Demand (SVOD)
A video-on-demand service that allows users to access an entire library of videos for a small recurring fee. The users have full control over the subscription and can decide when to start the program. They can also pause, fast forward, rewind, and stop the show as preferred.
TV Tune-in
The degree to which viewers tune into a particular program after exposures to network program-specific promotions.
Third-Party Data
Consumer data that is collected by an external source from the marketer that intends to use it. Third-party data is collected from a variety of websites and platforms and is then aggregated together by a third-party data provider such as a DMP.
Any media or marketing interaction to which a consumer is exposed. Touchpoints can include a wide range of interactions, from seeing a television commercial to clicking on a display ad or search result.
Unified Measurement Models
A new statistical approach for integrating strategic marketing mix and tactical digital analytics into a holistic model. Unified measurement has the potential to provide the metrics that allow marketers to optimize marketing spend in the campaign.
A measure of the accuracy and precision of modeled results There are two common and complementary approaches. The goodness of fit (MAPE: Mean Average Percent Error, or R2: percent of variance explained) describes how well the model replicates the historical data to which it was fit.