**Here is a glossary of 100 common terms used in Marketing and Media Mix Modelling (MMM), Customer Analytics and Marketing Science:**

- Acquisition – using your marketing and media budget to win new customers into your business.
- Activation – setting a campaign to go live and start delivering results.
- Addressable media – media that goes to an individual
- Ad Server – a server which hosts your digital ads so that they can be served onto the sites where you book media
- AdStock – the amount of time a consumer theoretically remembers an ad for after seeing it
- Artificial Intelligence (AI) – AI is a form of computer technology in which machines are trained to behave and learn like humans
- Attention – a metric that seeks to capture the amount of attention a consumer pays to an ad, rather than simply
*seeing*it. Dwell time is often used. - Attribution – the task of understanding and explaining which marketing and media channels drove which business results.
- Awareness – the percentage of your target audience that is aware of your product or service
- BARB – Broadcast Audience Research Board, the body that manages the measurement and reporting of TV audience data
- Base – sometimes known as the intercept, or ‘constant’ this is the level of sales your business or brand generates without media or advertising spend intervention. Sometimes called Beta0.
- Broad Audience – a wide demographic target audience definition such as ‘Adults’, ‘Men’, ‘Women’
- Broadcast – channels that reach mass audience quickly through mass communication such as TV or radio.
- Categorical variables – are descriptive rather than numeric, for example gender, country or product type
- Churn – losing customers
- Churn Rate – the rate at which you lose customers, usually measured as the percentage of customers lost per week, month or year
- Coefficient – the “weighting factor” that the model estimates for each variable. So, we might say that for every £1 invested, 0.01 (our coefficient) sales are generated i.e. a £100 cost per sale.
- Confidence Interval – The expected range in which you would expect a value to fall if that value was sampled multiple times
- Consideration – the proportion of your target audience that would consider buying your product or service. Identified through tracking studies like YouGov or bespoke research.
- Conversion – the rate at which visitors move to the next action in the path to purchase
- Conversion Funnel – a unified view of the journey consumers takes from becoming aware of your product to purchasing and sometimes including post-purchase behaviour.
- CPA – Cost per Action, or sometimes cost per acquisition
- CPM – Cost per Thousand (the M stands for “mille”)
- CPT – Cost per Thousand
- CRM – Customer Relationship Management
- Cross sectional data – data collected at a specific place or point in time, like a research survey
- D2C – Direct to Consumer – selling products or services directly to consumers
- Data Science – using data to establish cause and effect using the scientific method
- Decay – related to AdStock – the theoretical rate at which the memory of an ad decays
- Decomposition – identifying the component parts of a time series data set. For example, identifying underlying seasonality and trend components.
- Demo – Demographic – usually applied to a demographic group
- Dependent variable -the variable we are seeking to explain. The behaviour of the dependent variable is explained by the independent, or explanatory variables, sometimes also called the
*x*variables. - Dimensionality – in data science, this refers to the number of
*x*variables (columns) in relation to the number of observations (rows). - Direct Marketing – Marketing activity that seeks a measurable response at the individual consumer level
- Distribution – the element of the Marketing Mix that concerns the amount of retail coverage a business has. There can also be an online manifestation of this in the form of delivery times, returns policy etc.
- eCommerce – online websites that sell products and service and allow consumers to buy direct (see D2C)
- Econometrics – the practice of analysing economic data with statistics. Derived from the words “economy” and “metrics”
- Error term –
*e*– the part of the model equation that captures the data that can’t be explained in the model. Sometimes also called the remainder term. - so – excess Share of Voice – the excess value when your share of voice is divided by your share of market – higher so has been shown to assist brand sales or share growth
- Exploratory Data Analysis – the process of analysing a dataset to ensure that it is suitable for modelling
- Effective Frequency – usually used in brand campaigns or launch campaigns. This is the number of times consumers might need to see an ad for it to register.
- Feature engineering – adjusting variables such as creating interaction variables to test an effect in a model
- Frequency – Sometimes called OTS (see OTS)
- ggplot2 – one of the leading graphics packages in the R modelling suite.
- GRP – Gross Rating Point -the percentage of your audience that sees an ad
- Heteroscedasticity – a certain type of pattern (too much variance) sometimes seen in residuals – it can suggest that the model is not properly capturing all the information it has available to it
- Homoscedasticity – another residuals patter where the error is constant, indicating the model is good
- Impact – the actual number of consumers who see your ad
- Incrementality – business outcomes that are incremental to your static state
- Inference – drawing conclusions about a population based on the selection and statistical analysis of a sample from that population
- Interaction variable – a variable that is created when you combine two variables to text their combined effect e.g. an increase in ad spends and a change in price
- Intercept – also the base and technically the point at which the models predicted line crossed the y axis
- Lag – any delay between seeing an ad and acting on it
- Lifetime Value (LTV) – the amount of revenue a customer or subscriber generates over the estimated average customer lifetime. Usually between 1 and 5 years depending on the product and category.
- Linear response – a non-decaying response function e.g. for every cup of coffee consumed the satisfaction score remained unchanged.
- Machine Learning (ML) – is a subset of AI in which machines (computers) use data to make predictions
- Marketing Science – applying the scientific method to marketing data
- MMM – Marketing Mix Modelling – the technique of using statistical models to examine which variables drive business outcomes
- Model – a reproducible mathematical or statistical representation of something that occurs in life
- MOSAIC – A customer classification system similar to ACORN above
- MRR – Monthly Recurring Revenue – the amount of money a subscriber pays per month whilst in contract or signed up to a specific service.
- Noise – random changes in variables from which no or very little value can be extracted
- Non-linear response – a decaying response function e.g. the more coffee a person consumes, the less satisfied they are with each additional cup, as opposed to linear response in which case each coffee would just be as satisfying as the last
- OLS – Ordinary Least Squares is a regression technique that seeks to minimise the difference between actual and predicted values – related to the term ‘best fit’
- OTS – Opportunities to See, a measure of frequency
- Optimise – the act of reorganizing a fixed resource to produce more yield from
- Optimiser – a tool that can run optimisations to resolve an allocation problem to produce maximum possible business outcomes.
- Opt-In – a privacy state in which a consumer can decide whether or not to receive communications from a business.
- Outlier – a value in data that is outside the upper and lower quartiles of a dataset
- Panel model – a modelling technique which allows us to build models using the same variables from different data sets e.g. countries or cities
- PPC – Pay Per Click – a form of search marketing
- Prediction – applying your model to data to predict the value of
*y*at*x* - Programmatic – a system that bids for online display space automatically subject to a set of bid rules
- Prospect – a consumer who could buy but currently does not
- Python – a programming language conceptually similar to R used to build models
- R – a programming language (code) used to run statistical models
- Reach – the percentage of your target audience that see one of your ads at least once
- Recency Frequency Value (RFV) – a method for scoring and classifying customers using transactional data
- Regression – a statistical technique used to find the ‘line of best fit’ relationship between two sets of variables,
*x*and*y*. - Residuals – The difference between the values predicted by a model and the observed values in the data
- Response Curve – The consumer response function to your media investments
- Ridge Regression (sometimes called Regularised Regression) is a form of regression that scales down the influence of less important variables in a regression model
- RMSE – Root Mean Squared Error – measures the average difference between values predicted by a model and the actual values in the data. A key metric in model assessment
- R-Squared – Residuals squared – the total value of the residuals in a model, squared.
- R Studio – (Posit) – an integrated development environment, or IDE, like a GUI to write and edit R code
- Sales Funnel – a start to finish representation of your prospects’ path to purchase
- Seasonality – this is not about the seasons, but about identifying repeating cycles in a time series sample – these could be other days, weeks, months or years
- Share of Market (SOM) – your market shares i.e. your sales as a percentage of total category sales
- Share of Voice (SOV) – your advertising spends as a share of your category’s ad spend
- Spot Match – a technique for matching TV spots to web traffic using a time window, usually of a few minutes
- Standard deviation – a measure of the distance of each value in a dataset from the mean if that dataset
- Standard error – the standard deviation of the mean of multiple samples, say we tale 10 samples, get a mean of each, we have 10 means. they have different values. we take the mean of these means and calculate the standard deviation of that mean
- Static State – your BAU business metrics – the natural level of business you generate without advertising spend input
- Sub Group – any target audience that is smaller than the full adult population e.g. Men aged 16-34
- TGI – Target Group Index – a major consumer behaviour survey
- Time series data – a sample of data that is spread over time, such as sales data by week over three years
- Vector – a vector is a string of data that has magnitude and direction e.g. 1,3,5,7,9,11,13 can be called a vector
*x*variables – our explanatory variables (*x*1,*x*2,*x*3 etc) – these are the variables that we are testing for an effect on our dependent or Y variable*y*variable – the variable we want to explain with our explanatory variables (*x*1,*x*2,*x*3 etc)- YouGov – a major consumer survey of behaviour and attitudes, similar to TGI