Thank-you. We will contact you shortly.

Something went wrong! Please try again later.

Can MTA (multi touch attribution) Be Used for Forecasting?
FAQs

Unlike MMM, best in class MTA models estimate propensity to convert rather than demand and are therefore not directly applicable to forecasting.   While demand curves can be inferred from MTA models, they typically do not have much validity at the campaign level and only inform tactical decision making at the sub-channel level without MMM’s ability to forecast or support strategic decision making.

Incrementality of incremental contribution of Display advertising using audience holdouts, serving the held out audience a PSA ad, and comparing the measured response rate of the held-out audience versus campaign audience.

This process is called Design of Experiments (DoE). When expertly designed has the ability to deliver on the promise of incrementality at the vendor, campaign and audience level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom.  For heavily biased tactics like retargeting, DoE incrementality results can be actively incorporated into MMM as Bayesian Priors to improve MMM models across the board.  For retargeting tactics, DoE offers the most unbiased measurement approach, as it randomly selects a subset of website visitors for exclusion from retargeting impressions, both in total and at the vendor level, in order to measure true incrementality of these tactics on a customer group that has already established interest and intent.  For prospecting tactics, DoE carefully selects a subset of prospects to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of impressions.  Because this is designed at the group level, DoEs are not subject to all of the user level data challenges encountered by MTA requiring only that campaigns exhibit enough reach to establish statistical significance at the group level.  For most advertisers this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

Incrementality Testing creates an experiment which systematically withholds media channel exposure to a representative subset of customers (the control group) while maintaining normal media channel exposure to the broader customer set (the test group).  If the control group is both sizable enough to be statistically significant and selected at random such that they are broadly representative of the customer base, then the media channel lift can be determined by the difference in business outcome (conversion, revenue, profitability, etc.) between the test and control groups.

Design of experiments (DOE) is a systematic method to design experiments to measure the impact of marketing campaigns. DoE is a method to ensure that variables are properly controlled, the lift measurement at the end of the experiment is properly assessed and the sample size requirements are properly estimated.

DoEs can be designed for most major types of media, like Facebook, Search, TV etc.

Incrementality, like MMM, benefits from controlling for the broad set of media and non-media factors that impact the consumer decision making process. Like MTA it can measure the effects of media at a very granular (campaign/audience/adset) level but without relying on the challenges associated with user level data.

Furthermore a properly developed design of experiments that encompasses a broad of marketing channels can inform the tactical marketing optimization where MMM forecasting stops without the limitations and drawbacks of MTA.

Multi-Touch Attribution (MTA) collects cookie-level data for trackable addressable media and conversion events in order to determine the impact of each trackable event on the conversion path at the customer level.  By summing the impact of each trackable addressable media on each customers’ likelihood to convert, MTA quantifies the total media channel lift provided by all said trackable addressable media.  MTA does not account for the impact of non-addressable media, and furthermore much addressable media is either non-trackable or lost due to the innumerable challenges tracking data at the user level.

Marketing mix modeling (MMM) (sometimes also called Media Mix Modeling) collects aggregated data across marketing and non-marketing factors over a multi-year historical period.  That data is used to develop a demand model which quantifies the historical contribution of each marketing and non-marketing input to business outcome.  MMM typically estimates marketing impact on historical business outcomes at the channel level probabilistically, and can be subject to the correlation vs. causation dilemma. For forward looking projections MMM relies on a number of assumptions for non-marketing factors as well the assumption that channel level media mix, cost, and response does not diverge with the historic data that is the basis for the demand model.  While well built models based on high-quality data can overcome the correlation vs. causation dilemma to provide channel lift and forecasts, the limitation on degrees of freedom and challenges with overspecified models means that they cannot be used to inform tactical decision making at the sub-channel level.  Because models rely on multiple years of historical data to determine an average read for marketing inputs, they are challenged in sussing out dynamic changes to marketing channels and/or business changes in recent periods.

Incrementality in marketing refers to the incremental benefit produced per unit of input stimulation. Incrementality is the lift in desired outcome (awareness, web visits, conversion, revenue, profitability) provided by marketing activity.

MTA and DoE are complementary because incrementality testing addresses many of the data and data tracking gaps that currently serve as severe limitations to MTA’s ability to measure marketing contribution across all addressable marketing channels.  Currently MTA has a major data gap in the so-called walled gardens (Facebook, AdWords,Instagram, Pinterest, YouTube etc.) in which no customer level data gathering is permitted.  MTA has no answers for these channels with no clear avenues for improvement short of a 180 degree reverse of course on data sharing by the likes of Facebook (don’t hold your breath).  Even in trackable addressable media channels, pixel related data loss can be severe, ranging from 5% in paid search to as much as 80% in channels like online video.  While cookie level data tracking has lower rates of data loss, it’s ongoing viability is in question after Google recently announced the discontinued sharing of Google User IDs that this approach relies upon beginning in Q1 2020.  DoE can both fill the gaps created by the so-called “walled garden” media channels as well as validate and inform media channels suffering from pixel related data loss.  As the market continues to evolve, and legislation to address privacy concerns like GDPR proliferates, MTA measurement unsupported by DoE will likely become obsolete.

Incrementality or incremental impact of media channels towards sales or another business outcome is calculated using one of the following techniques:

  1. Design of Experiments (DoE)
  2. Marketing Mix Modeling (MMM)
  3. Multi-Touch Attribution (MTA)

Yes, incrementality can be measured on Direct Mail. The typical approach to incrementality measurement on Direct Mail is to use systematic holdouts and compare the response rate and revenue per piece from users in the holdout group versus the users in the mailed group.

Yes, Incrementality of incremental contribution of Facebook advertising towards sales or other business outcomes can be measured. The approaches used for this purpose are called advanced marketing measurement techniques.

The most commonly used advanced marketing measurement technique for measuring incrementality on Facebook is called Design of Experiments (DoE).

DoE, when expertly designed, has the ability to deliver on the promise of incrementality measurement at the campaign, audience and ad set level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom and MTA cannot due to the aforementioned limitations on user level impression data.   DoEs carefully select a subset of users to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of Facebook impressions. A carefully designed Facebook DoE can go into much more detail than simply channel level incrementality and instead can execute always-on simultaneous experimentation at the campaign, audience and ad set level with the only limitations being the number of days required to achieve statistical significance which is driven by reach at campaign/audience/adset level.  For most mid-market advertisers as scale this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

There are many challenges to measuring marketing effectiveness.

  • Types of data available: The type, richness and quality of data vary widely from channel to channel. For example, on programmatic display, marketers can get user-level, impression-level data, and track the user all way to a conversion if that user converts. However, on TV, marketers would only get reporting of when their ads were flighted on air. On Facebook, marketers can rich aggregate data reporting about their targeted audience, but almost no user-level data.
  • Frequency of data: Some media channels offer data on a real-time basis, whereas others only offer data post-campaign.
The types and frequency of data available limits the universe of advanced measurement techniques that can be used to measure a specific media tactic. Types of decisions: The types of decisions that the measurement informs also plays a big role in how useful marketing measurement itself is. Marketers looking to make limited tactical decisions, like optimizing between creative A vs creative B running on a specific audience within a specific media channel, can use data made available by that channel. However, more advanced decisions like balancing budgets across media tactics based on incremental sales and incremental ROI would have to use advanced marketing measurement techniques to get to an answer. Other more strategic questions, like the impact of weather or interest rates or government policy on sales would need to use specific marketing measurement technique that can answer those questions.

Catalog DoE, using holdouts, is the classical approach for measuring Catalog and Direct mail. When expertly designed, has the ability to deliver on the promise of incrementality measurement at the drop and keycode level.

Typically, holdouts are selected by picking the nth name within each keycode (or granular RFM-type segment) within each campaign. By carefully selecting test and control groups from both house CRM and prospecting files DoEs can control for addressable and non-addressable media as well as non-media factors at the group level, focusing on the true incrementality provided by Catalog tactics at the campaign and audience level once statistical significance has been achieved, usually in a matter of months for most catalog focused retailers.  These reads can then be meaningfully updated afterwards for each new catalog campaign, typically in as little as a week after any new mailing when Measured is providing the match back analysis.

MMM measures Catalog impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.

MTA attributes the impact of catalog impressions on conversion events at the user level.  For CRM related catalog activity based on an internal file of existing customers,  catalogs are typically mailed to known users then connected to online or in-store purchase events based on an internal CRM file and known user ID.  However, in order to accurately attribute credit to catalog at the impression level, even for known customer IDs, it is required to connect CRM based user IDs typically based on place ID (name/address) or email to digital impressions and user IDs served across multiple devices (desktop, mobile, browsers, etc.) typically based on cookie IDs.  In order to do this effectively a third party identity and cross-device provider is typically required at potentially significant additional cost to the marketer.  Without this additional data service MTA typically over attributes credit to catalog as it inaccurately assumes that little or no digital impressions have been served to or impacted the decision making process.  Even when this service is included in MTA at an additional cost, match rates rarely exceed 50%.  Prospecting catalog marketing activities executed via third party vendors are subject to all of the same aforementioned limitations with the added complexity that mailings sent at the place ID level must also be connected to conversions by unknown users either online at the cookie ID level or in-store via credit card or other PII collected at the time of purchase.  All of these challenges for both CRM and prospecting at the addressable level are in addition to MTAs previously mentioned and well known limitations in accounting for non-addressable media and non-media factors.

Media channel lift or incrementality can be measured or approximated in a number of ways.

Design of Experiments (DoE) creates an experiment which systematically withholds media channel exposure to a representative subset of customers (the control group) while maintaining normal media channel exposure to the broader customer set (the test group).  If the control group is both sizable enough to be statistically significant and selected at random such that they are broadly representative of the customer base, then the media channel lift can be determined by the difference in business outcome (conversion, revenue, profitability, etc.) between the test and control groups.

Media mix modeling (MMM) collects aggregated data across marketing and non-marketing factors over a multi-year historical period.  That data is used to develop a demand model which quantifies the historical contribution of each marketing and non-marketing input to business outcome.  MMM typically estimates marketing impact on historical business outcomes at the channel level.

Multi-Touch Attribution (MTA) collects cookie-level data for trackable addressable media and conversion events in order to determine the impact of each trackable event on the conversion path at the customer level.  By summing the impact of each trackable addressable media on each customers’ likelihood to convert, MTA quantifies the total media channel lift provided by all said trackable addressable media.

In short, yes.  Experimentation relies on steady state operation of your media channels as the vast majority of your customer base, receiving business as usual media exposure, will comprise the all important test group in the experimental design.  Good experimentation carefully selects out a small, but representative subset of your customers and withholds media exposure at the channel level in order to serve as the test group which will bear out the true incremental sales driven by channel level media exposure.

The impact of a marketing tactic (eg: Facebook Prospecting, Retargeting, Catalog for Housefile, Direct Mail, National TV) is typically measured in terms of incremental sales driven by that tactic. Incremental sales driven by a media tactic is calculated using advanced marketing measurement techniques. There are 3 major types of advanced marketing measurement techniques.

  • Design of Experiments (DoE)
  • Marketing Mix Modeling (MMM)
  • Multi-Touch Attribution (MTA)

Once incremental sales is calculated for a tactic using one of the above techniques, the ROI for that tactic is calculated using the following formula.

Direct Mail DoE, using holdouts, is the classical approach for measuring Catalog and Direct mail. When expertly designed, has the ability to deliver on the promise of incrementality measurement at the drop and keycode level.

Typically, holdouts are selected by picking the nth name within each keycode (or granular RFM-type segment) within each campaign. By carefully selecting test and control groups from both house CRM and prospecting files DoEs can control for addressable and non-addressable media as well as non-media factors at the group level, focusing on the true incrementality provided by Catalog tactics at the campaign and audience level once statistical significance has been achieved, usually in a matter of months for most catalog focused retailers.  These reads can then be meaningfully updated afterwards for each new catalog campaign, typically in as little as a week after any new mailing when Measured is providing the match back analysis.

MMM measures Catalog impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.

MTA attributes the impact of catalog impressions on conversion events at the user level.  For CRM related catalog activity based on an internal file of existing customers,  catalogs are typically mailed to known users then connected to online or in-store purchase events based on an internal CRM file and known user ID.  However, in order to accurately attribute credit to catalog at the impression level, even for known customer IDs, it is required to connect CRM based user IDs typically based on place ID (name/address) or email to digital impressions and user IDs served across multiple devices (desktop, mobile, browsers, etc.) typically based on cookie IDs.  In order to do this effectively a third party identity and cross-device provider is typically required at potentially significant additional cost to the marketer.  Without this additional data service MTA typically over attributes credit to catalog as it inaccurately assumes that little or no digital impressions have been served to or impacted the decision making process.  Even when this service is included in MTA at an additional cost, match rates rarely exceed 50%.  Prospecting catalog marketing activities executed via third party vendors are subject to all of the same aforementioned limitations with the added complexity that mailings sent at the place ID level must also be connected to conversions by unknown users either online at the cookie ID level or in-store via credit card or other PII collected at the time of purchase.  All of these challenges for both CRM and prospecting at the addressable level are in addition to MTAs previously mentioned and well known limitations in accounting for non-addressable media and non-media factors.

Online Display DoE, when expertly designed has the ability to deliver on the promised of incrementality at the vendor, campaign and audience level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom.  For heavily biased tactics like retargeting, DoE incrementality results can be actively incorporated into MMM as Bayesian Priors to improve MMM models across the board.  For retargeting tactics, DoE offers the most unbiased measurement approach, as it randomly selects a subset of website visitors for exclusion from retargeting impressions, both in total and at the vendor level, in order to measure true incrementality of these tactics on a customer group that has already established interest and intent.  For prospecting tactics, DoE carefully selects a subset of prospects to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of impressions.  Because this is designed at the group level, DoEs are not subject to all of the user level data challenges encountered by MTA requiring only that campaigns exhibit enough reach to establish statistical significance at the group level.  For most advertisers this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

MMM measures Online Display impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.  MMM typically encounters challenges in Online Display with retargeting tactics as they are highly correlated to business outcomes and therefore must be handled with extreme care to prevent reads based on spurious correlation.  Due to these known challenge, retargeting contribution is oftentimes practically limited via Bayesian approaches rather than truly estimated via statistical techniques.  Where Bayesian priors are used they have typically been heavily informed by incrementality measurements.

MTA attributes Online Display impressions on conversion events at the user level.  These are typically connected via cookie ID and website level conversion activity.  For retargeting impressions served after a site visit data tracking at the user level is fairly accurate, however the inclusion of other user level impression data that happens prior to a site visit is subject to either data loss or the inability to track data the the user level (e.g. Facebook).  For these reasons, and due to the fact that multiple retargeting impressions occur in the vast majority of converting and non-converting sequences MTA is biased to highly overattribute the impacts and credit given to retargeting in converting sequences.  For online display impressions served prior to a site visit, e.g. prospecting tactics, user level data collection is subject to data loss at the user level which can be significant depending upon the channel and vendor partner.  So while well executed MTA can deliver meaningful insights at the user level where good impression level data tracking exists, in aggregate these tactics tend to be under attributed due to the data loss problem.

Testing to measure the impact of media channels are designed using a process called Design of Experiments (DoE).

Design of experiments (DOE) is a systematic method to design experiments to measure the impact of marketing campaigns. DoE is a method to ensure that variables are properly controlled, the lift measurement at the end of the experiment is properly assessed and the sample size requirements are properly estimated.

Over the years, various attribution techniques have been developed and deployed as Software-as-a-Service (Saas) applications that marketers have come to rely on. Together, this class of software is called attribution software.

Attribution modeling is a method for assigning credit to advertising intended to drive sales. The most common and simplistic approach for attribution is called last-click attribution. This method offers 100% credit to the last click in the user’s path. In general, last click attribution is considered overly simplistic, and is used in a limited tactical way by marketers for making decisions.

For making more impactful decisions, more advanced methods have been developed over the years. A variety of techniques have been developed to determine how much credit to attribute to specific marketing campaigns. This process if called attribution modeling.

Media channel incrementality refers to the incremental business benefit produced per unit of media channel stimulation. Media channels composed of paid, earned and owned media whether online or offline having messages put into market with a goal of driving some desired awareness or response goal. Common media channels include Paid Search, Online Display, Social Media, Online Video, TV, Radio, Out of Home (OOH) and Direct Mail.

MMM, MTA and DoEs would take different approaches (see above) and provide different insights to Facebook measurement.

DoE, when expertly designed, has the ability to to deliver on the promise of incrementality measurement at the campaign, audience and ad set level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom and MTA cannot due to the aforementioned limitations on user level impression data.   DoEs carefully select a subset of users for to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of Facebook impressions. A carefully designed Facebook DoE can go into much more detail than simply channel level incrementality and instead can execute always-on simultaneous experimentation at the campaign, audience and ad set level with the only limitations being the number of days required to achieve statistical significance which is driven by reach at campaign/audience/adset level.  For most mid-market advertisers as scale this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

MMM measures Facebook impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.

MTA, due to the inability to track user level impression data, can only attribute credit to Facebook ads via connecting click through to site data to  conversions at the user level.  It is widely known that click only attribution misses 80%+ of total Facebook media attribution, as the number of Facebook impressions in converting sequences vastly outweighs the numbers of clicks in converting sequences.   This renders MTA unable to measure holistic Facebook contribution or incrementality both in total and with respect to other media channels.

Media channels are tested using  a process called Incrementality Testing creates an experiment which systematically withholds media channel exposure to a representative subset of customers (the control group) while maintaining normal media channel exposure to the broader customer set (the test group).  If the control group is both sizable enough to be statistically significant and selected at random such that they are broadly representative of the customer base, then the media channel lift can be determined by the difference in business outcome (conversion, revenue, profitability, etc.) between the test and control groups.

Marketing ROI is calculated as the ratio of incremental contribution to sales from a marketing channel divided by the spend in that marketing channel.
Once incremental sales is calculated for a tactic using one of the above techniques, the ROI for that tactic is calculated using the following formula.

Media channels are measured by the incremental impact or contribution the make towards sales or another business outcome. The incremental impact is calculated using one of the following techniques:

  1. Design of Experiments (DoE)
  2. Marketing Mix Modeling (MMM)
  3. Multi-Touch Attribution (MTA)

One of the biggest questions in marketing is how much credit marketing campaigns deserve in driving sales. A variety of techniques have been developed to determine how much credit to attribute to specific marketing campaigns. This process if called attribution modeling

The most common and simplistic approach for attribution is called last-click attribution. This method offers 100% credit to the last click in the user’s path. In general, last click attribution is considered overly simplistic, and is used in a limited tactical way by marketers for making decisions. For making more impactful decisions, more advanced methods have been developed over the years. Together, these methods are called advanced marketing measurement techniques.”

Yes, incrementality can be measured on Catalog. The typical approach to incrementality measurement on Catalog is to use systematic holdouts and compare the response rate and revenue per book from users in the holdout group versus the users in the mailed group.

This can be done in a number of ways, but there is no doubt that best-in-class cross-channel reporting is created by collecting and aggregating data via direct vendor-level API based data feeds updated at the daily level.  For vendor without API based data feeds automated email to FTP uploads minimize latency and maximize data accuracy compared with manual data ETL operations.

For marketers who are getting started, Facebook can be measured using LastClick attribution reporting that Facebook natively offers. For more mature marketers, Facebook’s true contribution towards sales or other busieness outcomes have to be measured using advanced marketing measurement techniques.

MMM, MTA and DoEs are three different approaches that provide different insights to Facebook measurement.

DoE, when expertly designed, has the ability to to deliver on the promise of incrementality measurement at the campaign, audience and ad set level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom and MTA cannot due to the aforementioned limitations on user level impression data.   DoEs carefully select a subset of users for to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of Facebook impressions. A carefully designed Facebook DoE can go into much more detail than simply channel level incrementality and instead can execute always-on simultaneous experimentation at the campaign, audience and ad set level with the only limitations being the number of days required to achieve statistical significance which is driven by reach at campaign/audience/adset level.  For most mid-market advertisers as scale this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

MMM measures Facebook impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.

MTA, due to the inability to track user level impression data, can only attribute credit to Facebook ads via connecting click through to site data to  conversions at the user level.  It is widely known that click only attribution misses 80%+ of total Facebook media attribution, as the number of Facebook impressions in converting sequences vastly outweighs the numbers of clicks in converting sequences.   This renders MTA unable to measure holistic Facebook contribution or incrementality both in total and with respect to other media channels.

Marketing impact is typically measured in incremental sales or business outcomes relevant to a business. Incremental sales driven by a media tactic are calculated using advanced marketing measurement techniques. There are 3 major types of advanced marketing measurement techniques.

  • Design of Experiments (DoE)
  • Marketing Mix Modeling (MMM)
  • Multi-Touch Attribution (MTA)

Most marketers who are investing reasonable amounts of advertising dollars on Facebook graduate from using last click to advanced marketing measurementtechniques.

MMM, MTA and DoEs are three different approaches that provide different insights to Facebook measurement.

DoE, when expertly designed, has the ability to deliver on the promise of incrementality measurement at the campaign, audience and ad set level in a way that MMM cannot due to practical limits on data granularity and degrees of freedom and MTA cannot due to the aforementioned limitations on user level impression data.   DoEs carefully select a subset of users to serve as the control group, showing them a PSA advertisement usually for a charity of the marketers choice in order to determine the true incrementality of Facebook impressions. A carefully designed Facebook DoE can go into much more detail than simply channel level incrementality and instead can execute always-on simultaneous experimentation at the campaign, audience and ad set level with the only limitations being the number of days required to achieve statistical significance which is driven by reach at campaign/audience/adset level.  For most mid-market advertisers as scale this statistical significance is achieved in a matter of weeks, and can be meaningfully updated afterwards on a weekly basis and be used to inform tactical campaign optimization at the weekly level.

MMM measures Facebook impact on aggregated business outcomes at the channel or sub-channel level, providing insights into channel contribution relative to other addressable and non-addressable media and informs strategic planning and forecasting at the annual or quarterly level.

MTA, due to the inability to track user level impression data, can only attribute credit to Facebook ads via connecting click through to site data to  conversions at the user level.  It is widely known that click only attribution misses 80%+ of total Facebook media attribution, as the number of Facebook impressions in converting sequences vastly outweighs the number of clicks in converting sequences.   This renders MTA unable to measure holistic Facebook contribution or incrementality both in total and with respect to other media channels.

Measurement the impact of marketing on sales or other business outcomes is performed using a broad array of techniques called Advanced Marketing measurement techniques. They are:

Design of Experiments (DoE) creates an experiment which systematically withholds media channel exposure to a representative subset of customers (the control group) while maintaining normal media channel exposure to the broader customer set (the test group).  If the control group is both sizable enough to be statistically significant and selected at random such that they are broadly representative of the customer base, then the media channel lift can be determined by the difference in business outcome (conversion, revenue, profitability, etc.) between the test and control groups.

Media mix modeling (MMM) collects aggregated data across marketing and non-marketing factors over a multi-year historical period.  That data is used to develop a demand model which quantifies the historical contribution of each marketing and non-marketing input to business outcome.  MMM typically estimates marketing impact on historical business outcomes at the channel level.

Multi-Touch Attribution (MTA) collects cookie-level data for trackable addressable media and conversion events in order to determine the impact of each trackable event on the conversion path at the customer level.  By summing the impact of each trackable addressable media on each customers’ likelihood to convert, MTA quantifies the total media channel lift provided by all said trackable addressable media.

The type of attribution model chosen determines what factors are included in assigning credit to advertising towards recognized sales.

The most common and simplistic approach for attribution is called last-click attribution. This method offers 100% credit to the last click in the user’s path. In general, last click attribution is considered overly simplistic, and is used in a limited tactical way by marketers for making decisions. Other attribution models like First-click, U-shaped, even-attribution take the sequence of the advertising events as the only factor to determine credit.

More advanced methods have been developed over the years, which take into account other factors like frequency of ad exposures, recency of ad exposures, type of exposure and other such factors towards determining the right level of credit.

A/B testing is a specialized type of incrementality measurement.  In A/B testing randomized groups are shown a variant of a single variable (web design, landing page, marketing creative etc.) in order to determine which variant is more effective.  Incrementality measurements use A/B testing in certain media channels, such as prospecting, where tracking a media exposure for both the test and control groups is required.  In the case of media incrementality the A group  (test group) is shown business as usual media exposure while the B group (control group) has exposure withheld or is shown a null media exposure, typically a public service announcement (PSA) for a charity of the marketers choice.  The more generic form of A/B testing is called Design of Experiments (DoE).

Attribution modeling is a method for assigning credit to advertising intended to drive sales. The most common and simplistic approach for attribution is called last-click attribution. This method offers 100% credit to the last click in the user’s path. In general, last click attribution is considered overly simplistic, and is used in a limited tactical way by marketers for making decisions.

For making more impactful decisions, more advanced methods have been developed over the years. A variety of techniques have been developed to determine how much credit to attribute to specific marketing campaigns. This process if called attribution modeling.