Marketing Intelligence

Addressable Media

Addressable media refers to any media that can be tied to an individual user, either through a probabilistic or deterministic identifier. For instance, a display ad served in-app can be tied to the user’s device id, making it addressable. However, an ad broadcasted on standard, traditional TV does not pick up any identifier and therefore cannot be traced back to a specific user or household and is therefore not addressable.

It’s easy to assume that all digital channels are addressable, but this is actually not accurate. Most marketers cannot retrieve specific identifiers from the Walled Gardens, leading to large sections of the digital marketing universe that remain non-addressable.

Likewise, it’s easy to assume that all offline channels are non-addressable. But once more, this is actually not accurate. Direct mail is very addressable, and the cable companies have been rolling out addressable TV to better compete against IP-enabled digital TV (Connected TV and OTT)

Algorithmic Attribution

Algorithmic attribution, also known as machine learning, is the process of assigning a portion of credit for a conversion to each touchpoint based on effectiveness. The key differentiator of algorithmic attribution is its use of advanced statistical modeling and inferences to determine an optimal, custom model that continually refines itself based on your data – put more simply, human assisted machine learning.

App Install

Many marketers do have very targeted goals of driving installs of apps they may have recently released, These are often run through Cost-per-install (CPI) programs where the marketer is able to pay their media partners for driving new users to install the app.

App install is often defined as the success metric for the CPI program, though many CPI programs wait until there is an actual post-app-install transaction before paying out. Because of their walled gardens, it’s actually quite difficult to measure whether the app install reported properly within certain advertisers. In cases where app installs need to be rewarded, the “install event” is often only recognized the first time the user starts up the app.

Attributed Conversions

Refers to the fractional conversions allocated by an attribution model to the channels, campaigns, keywords, placements or whatever attributable element. For instance, for a conversion path consisting of a paid search touchpoint followed by a display touchpoint leading to a single conversion, and assuming an even distribution of credit (i.e. a linear attribution model), the paid search channel gets an attributed conversion of 0.5 and the display channel gets an attributed conversion of 0.5

Attributed Revenue

Refers to the fractional revenue allocated by an attribution model to the channels, campaigns, keywords, placements or whatever attributable element. For instance, for a conversion path consisting of a paid search touchpoint followed by a display touchpoint leading to a single conversion worth $50 in revenue, and assuming an even distribution of credit (i.e. a linear attribution model), the paid search channel gets an attributed revenue of $25 and the display channel gets an attributed conversion of $25.

Note that advertisers can choose to provide pure revenue for their attribution analysis, but it’s usually better to use net revenue (which takes revenue and subtracts product costs) for your attribution calculation.

Attributed ROAS

ROAS means Return on Ad Spend, and is a derived metric that captures how effective your media investment was in delivering positive value versus how much you spent on that media. Since Return on Ad Spend (ROAS) is a derived metric, attributed credit is not directly distributed to these metrics. Rather, Attributed ROAS is calculated using Attributed Revenue.

The formula used is:

Attributed ROAS = Attributed Revenue / Media Cost

Attributed ROI

ROI stands for Return on Investment, and is a derived metric that captures how effective your media investment was.

It looks at:

(a) the positive value associated with the user performing the desired action (for instance, making a purchase, where the positive value is the money they spent)

(b) the cost of the media

(c) the cost of goods sold (the cost associated with the product.

As you can tell, Return on Investment (ROI) is related to Return on Ad Spend (ROAS), but also adds the cost of the product in the calculation of the derived metric. Because Attributed ROI is a derived metric, attributed credit is not directly distributed to it. Rather, Attributed ROI is calculated using Attributed Revenue.

The formula used is:

Attributed ROI = Attributed Revenue / (Media Cost + Cost of Goods Sold)

Attribution

The process of identifying a set of user actions (“events”) ?that contribute in some manner to a desired outcome, and then assigning a value to each of these events. Marketing attribution provides a level of understanding of what combination of events influence individuals to engage in a desired behavior, typically referred to as a conversion.

In general, marketers and agencies will use attribution to determine how to distribute credit for a conversion event based on the kinds of exposures and engagement a specific user has gone through in their customer journey on the way to conversion

Attribution Model

An attribution model is a methodology that is applied to all of a campaign’s or advertisers conversion paths in order to determine how to distribute the credit for a conversion. If you are a retailer, for instance, and you find that $200,000 of your revenue in the past month were to users who were exposed to your paid media, attribution models help you figure out how to allocate the credit for the $200,000 across the elements of your paid media. Attribution models help you understand the value of channels, campaigns, placements, keywords, etc… based on the revenue they may have helped generate

Various attribution models can be compared against each other to determine which is the best fit for your goals and gain a more holistic view of each media’s contribution. There is no one perfect model, an organization should continuously update their models and examine their ability to predict future performance.

Baseline Conversions

Baseline conversions, in attribution, refers to the estimated number of conversions that would have happened even without any of the marketing activity being measured by the attribution model. For example, baseline conversions may have been caused by external factors such as hard-to-measure word-of-mouth marketing, or by offline advertising — which cannot be measured by attribution models.

By establishing the baseline conversions before running attribution, the marketer is able to more precisely calculate the lift provided by their marketing initiatives, and only allocate the incremental conversions to the addressable initiatives being analyzed by the attribution system

Bathtub Model

A bathtub model is a rules-based model that allocates a set amount on the first and last touchpoints of a conversion path, and taking the remainder and allocating an equal amount to the middle touchpoints. For instance, let’s say your conversion path consists of Email >> Video >> Display >> Paid Search >> Retargeting. If we configure the bathtub model to allocate 70% to the end points (meaning that Email gets 35% and Retargeting gets 35%) then the remainder gets distributed evenly to the intervening touchpoints (meaning that Video gets 10%, Display gets 10% and Paid Search gets 10%)

Note that events may happen beyond a particular lookback window — and these will not receive any credit. For example, if the marketer has a 30-day lookback window in the above example, and a Paid Social event actually happened 31 days before the conversion event — then that Paid Social event does not receive any of the credit because it occurred before the 30-day lookback window.

Bias

The models you use for attribution can introduce bias. For example, last click is biased in favor of channels that appear later in the buy cycle, such as coupon sites that often attract customers right before they buy, though they likely would have bought anyway.

Category-level Attribution

Attribution is typically run across all conversions that occur within a time period. However, E-Commerce marketers have an opportunity to get more granular, and analyze a subset of their conversions specific to a particular category.

They can run attribution at the category level in order to answer questions like:

* Which channels are best at driving revenue for high-margin categories, like lady’s handbags and shoes?

* Which ones are best at driving conversions for my top-selling men’s shoes category?

and so forth.

Channel

Attribution is about examining all the various channels that are part of the customer journey – both online and offline. Online channels include search, social, display, affiliate, email and so much more. Offline channels like print, television, radio and outdoor are equally as important in an omnichannel customer journey. The nuance here is that the offline channels must be addressable, i.e. they can be traced back to an online visitor in order for the offline channel to appear in the journey. At the aggregate level, both offline addressable and non-addressable explain overall customer response to marketing stimuli.

Channel Predictions

Channel Predictions predict how a marketer’s KPIs are likely to trend over the next 30 days, allowing them to see how they are pacing to their goals based on a number of inputs.

Using Channel Level Predictions or Forecasting, a marketers can know in advance when to sit back and relax (because pacing indicates they are likely to crush their goals) and focus on other areas of growth, or, if Forecasting extrapolate that they will miss their KPI goals, they get early-enough warning to go on overdrive and take additional actions to spur growth.

Click

Clicks refer to the action of engaging with the advertiser’s media. On a mobile device, it’s more appropriate to refer to clicks as taps.

Clicks on many paid search or standard banner ads will typically take users to the advertiser’s website or app.

However, this may not always be the case — clicks or taps on certain types of display ads may trigger a video or other interactive element that keeps the user on the same page.

Click Tracking

Click tracking allow tracking solutions such as Impact to track when the user clicks on something. In truth, clicks can either be tracked directly on the website (the user ends up clicking through into a landing page anyway) or can be trafficked in the advertiser’s ad management system as a 3rd party click through callout. In a web environment, the click tracker is often an executable tag, though pixels are also viable. In-app requires a dedicated API for click tracking since Javascript tags cannot be used within an In-app environment. All necessary contextual information about the referrer (the original publisher or media partner source of the click) is passed along with the click tracker

Consumer Journey

A consumer journey refers to the set of the advertiser’s marketing touchpoints that a particular user is exposed to or engages with over a period of time. It’s easy to confuse the consumer journey with the conversion paths — but they are not the same because many consumer journeys don’t end up with conversions.

Users who end up converting (i.e. in retail, a conversion is often a successful order. In auto, a conversion is often when a user chooses to ‘schedule a test drive’. Consumers who convert are typically exposed to a number of touchpoints beforehand. When customer journeys lead to a conversion, the customer journey is called a conversion path.

Not all conversions are driven by advertising. Some people just go directly to the advertiser’s website and make a purchase – even without receiving any exposure to any paid media. Such a conversion would essentially be organic and have a zero-length conversion path

Content Marketing

Content marketing refers to a marketing technique where the marketer publishes their own short or long-form content (in any format — written, audio, video) and pushes it out to their audiences in the hope that the intellectual property provides value for the reader. Content marketing can take on a variety of angles, such as beginner’s guides, educational pieces, infographics, thought leadership, research papers, buyer’s guides and more.

Content marketing stands in contrast to advertising, which is mostly paid marketing used to build awareness or persuade viewers to take action — the concept of intellectual capital is less pronounced in the advertising world versus the advertising world. However — they are very complementary in nature, since advertising can be used to promote and increase awareness of new content marketing pieces provided by the marketer.

Content marketing is distributed using a variety of methods:

a) Published on-site. Content marketing is often posted on the marketer’s website, and the marketer uses a variety of technique (paid advertising, organic social posts, email, etc…) to reach audiences and make them aware of the new piece of content marketing, and drive them to the site.

b) Published on 3rd party sites. Content syndication can happen in a variety of ways. A marketer may work with a trade association to publish their content marketing on their site (and other promotional channels, for example — content marketing may be disbursed to the 3rd party’s newsletters, etc…). A publisher may incorporate the piece of content marketing into their site as a “Sponsored Article” — which is a form of native advertising

Conversion

Conversions refer to success events — they represent actions that marketers want to their audiences to do. There are online conversions — success events that happen on the digital channel, and offline conversions — success events that happen in the physical world. When a user successfully checks out of the advertiser’s e-commerce site, then that’s an example of an online conversion. Another user may go to the advertiser’s brick & mortar location and buy something — an excellent example of offline conversion.

Conversion De-duplication

A marketer will typically use multiple systems to manage different channels. For instance, they may use an SEM like Kenshoo or Marin to manage paid search, and they may use an Ad Server or DSP like Doubleclick, Sizmek or the Trade Desk to execute their display ads. Each may track conversions independently — and if channel managers are not coordinating, each channel manager is watching their own conversion tracking, and the total number of conversions end up far exceeding the true number of conversions because they are getting overcounted across systems.

In our example above, if a marketer using different systems for SEM and Display noticed 50 conversions the past day, and noticed that all 50 involved both one Paid Search and one Display event each — then if no one does conversion de-duplication, then the marketer may wrongly conclude that they received 100 conversions over the past day — 50 from paid search and 50 from display.

That’s why cross-channel leaders recognize the importance of conversion de-duplication. Conversion de-duplication consolidates and reconciles all conversion events, so that duplicated conversion events recognized by separate systems are unified. It is a necessary step for any reliable Customer Journey Analytics or Multi-touch Attribution analysis.

Conversion Paths

A conversion path refers to the specific subset of consumer journeys that end with a user converting

Conversion Tracking

Conversion tracking allow tracking solutions such as Impact to track when the user converts on something. In a web environment, the conversion tracker is often an executable Javascript tag, though pixels are also possible. In-app requires a dedicated API for conversion tracking since Javascript tags do not run within the In-app environment.

Cookie

Cookies are still the primary deterministic identifier in the desktop and mobile web world. Cookies can either be first-party cookies or third-party cookies.

Apple Safari has been the most restrictive browser and does not allow setting of 3rd party cookies by default on iPhones (though users have the option to alter this behavior from their Browser Settings), and have dramatically limited the lifespan of even first-party cookies with its ITP updates.

Cookie formats are typically non-standardized as most companies maintain their own cookie pools. They are also pseudonymous – that is, they can be tied to personally identifiable information (PII) but when viewed on their own, don’t tell the viewer anything beyond a string of letters and numbers.

Cookie (First Party)

First party cookies are cookies that are issued by the domain that they are currently browsing on

Cookie (Third Party)

Third party cookies are cookies issued by a domain that is different from the domain the user is currently browsing on

Cost Importers

Cost importers are Impact’s tools for pulling in media cost data from 3rd party systems within the advertiser’s tech stack. Cost importers are generally IT-less (they do not require a technical resource to implement the integration) and can be fully configured by non-technical resources from the Impact platform directly.

CPA

CPA, or Cost per Acquisition or Cost per Action, is a metric that is tracked in many direct response and performance campaigns, particularly in verticals that are tracking user conversions — whether that conversion represents a sale or a form submission, depending on what the advertiser decides. This is why it’s also often referred to as Cost per Conversion.

Some marketers will include “clicks” as a viable action — in those cases, the calculation is essentially equivalent to a CPC (Cost per Click).

A related concept is eCPA, or Effective Cost per Acquisition. This is often calculated by advertisers who pay on another cost basis such as CPM or CPC, but wish to convert it to a Cost per Acquisition in order to optimize their media buying to some Cost per Acquisition target.

It is calculated as follows:

( sum of the relevant media costs / total # of acquisitions )

So, if a display campaign spent $1,000, and garnered 20 conversions, then the the eCPA = $1000 / 20 = $50

CPC

CPC, or Cost per Click, is a metric that is tracked in many branding, direct response and performance campaigns across any vertical. A click often refers to clickthru on an ad that directs them to the advertiser’s website, though many rich media campaigns may count a click on the ad that triggers some engagement (for instance, the user clicks on the ad to start playing a video, or playing a mini-game on the ad unit); in the rich media situation, this can also be referred to as Cost per Engagement (CPE).

A related concept is eCPC, or Effective Cost per Click. This is often calculated by advertisers who pay on another cost basis such as CPM, but wish to convert it to a Cost per Click in order to optimize their media buying to some Cost per Click target.

It is calculated as follows:

( sum of the relevant media costs / total # of clicks )

So, if a display campaign spent $5,000, and garnered 250 clicks, then the the eCPC = $5000 / 250 = $20

CPCV

CPCV, or Cost per Completed View, is a metric that is tracked in many video-based campaign across any vertical. A completed view is often triggered once the viewer of the video reaches the end of the video, though due to the idiosyncracies of many video player platforms, it may be triggered when <100% of the video is viewed.

A related concept is eCPCV, or Effective Cost per Completed View. This is often calculated by advertisers who pay on another cost basis such as CPM, but wish to convert it to a Cost per Completed View in order to optimize their media buying to some Cost per Completed View target.

It is calculated as follows:

( sum of the relevant media costs / total # of completed views )

So, if a display campaign spent $10,000, and garnered 200 completed views, then the the eCPC = $10000 / 200 = $50

CPL

CPL, or Cost per Lead, is essentially a subset of CPA, or Cost per Acquisition, specifically used by verticals that require the audience to complete a form with their contact info. For instance, in the insurance vertical, an interested user may have to enter their personal info in order to request an insurance quote or have a broker contact them.

CPM

CPM, or Cost per Mille, or Cost per thousand impressions (mille is the latin word for thousand) is one of the most common ways to purchase advertising today. Though it is used across many branding and direct response campaigns, it is particularly suited for verticals and campaigns intended to raise awareness.

For instance, if the CPM is priced at $2, and you wish to deliver 1,000,000 impressions, then the cost of the campaign is

(Total # of impressions / 1000) * $2

(1,000,000 impressions / 1000) * $2 = $2,000

Many advertisers running direct response or performance campaigns who pay for media based on CPM will often calculate an eCPC or eCPA as a KPI in order to track their success and to optimize toward lowering that KPI.

CPvM

CPvM, or Cost per Viewable Impression, is a metric that is tracked in many campaign across any vertical. A viewable impression is generally measured based on IAB standards — that is, for a display ad, 50% of the ad appears on the screen for at least 1 second, and for a video ad, 50% of the ad appears on the screen for at least 2 seconds.

Since most display or video campaigns today are paid in CPM rather than CPvM, CPvM is usually calculated.

It is calculated as follows:

( sum of the relevant media costs / total # of viewable impressions )

So, if a display campaign spent $2,000, and garnered 200 viewable impressions, then the the CPvM = $2000 / 200 = $10″

Cross-Device Journey

Cross-device Journey depicts the customers journey regardless of which of their owned device a marketer’s touchpoint reaches them on. This is in contrast with a journey that does not factor in cross-device. A user who was exposed to the marketer’s media touchpoints across their mobile device, tablet and desktop will appear as three separate users with three distinct single-device customer journeys instead of one unified user spanning their many devices.

This has always been important, but is growing more and more so. In the US, the average user owns over 3 devices — and that number continues to increase each year. In order for marketers to have any reliability in their Customer Journey Analytics or Multi-touch Attribution solution — it must understand the users’ cross-device journey

Custom Model

Custom Models are rules-based attribution models that are completely defined by the marketer’s business rules. They can start with some base rules-based model (i.e. start off with a a linear attribution model) and can be customized to meet just about any business rule the marketer has. For instance, they can implement a Custom Rule that says “Allocate 30% of the credit to the first touchpoint unless the first touchpoint is a website visit. Allocate 20% to the final touchpoint and distribute the remaining credit to the rest of the central touchpoints.”

Altitude (by Impact) attribution models are very malleable, and Altitude provides pretty flexible ways to shape and customize the attribution model to fit exactly whatever business rule customizations are needed

Customer Journey

The value of attribution is to examine the journey that led to the desired action – this includes cross channel (online, offline) and cross device (desktop, mobile, tablet). 79% of users own three or more devices. Recent studies show that users switch between devices up to 27 times per hour.

Customer Journey Analytics

Customer Journey Analytics refers to a category of marketing intelligence products that deal with analyzing metrics and structures associated with the customer journey.

Marketers can, for instance, ask questions such as:

* What is the average number of touchpoints along the customer journey for converting paths?

The marketer may choose to anti-target a user who dramatically exceeds that average by a wide margin.

* What is the most popular way that converting paths start?

The marketer may choose to dial-up some of their investments on these first touch channels or campaigns

* What is the average duration of a conversion path?

The marketer may choose to anti-target users who far exceed the typical duration that most users take to convert

Many of these Customer Journey Analyses can be performed directly from the available Impact reports, though a user who would really like to dig deep into them can analyze individual paths through PAQL, Impact’s proprietary querying language for customer journeys.

In the future, we anticipate marketers to leverage Customer Journey Analytics to start activating marketing investments to guide users down higher-conversion rate paths.

Dashboard

A dashboard is a set of visual widgets that are used by specific roles within a data-driven marketing department to run their business and make decisions. Visual widgets can include longitudinal charts, snapshots-in-time breakdown charts, tables, lists, trending or forecast graphs, real-time KPI scoreboards, goal meters and many other innovative mechanisms to visualize numerical data in order to simplify and bubble up insights.

Generally, different members of the marketing organization will want to have organize, assemble and tailor their own dashboards to support their unique role, root-cause analysis methodologies and visual preferences. For instance, the CMO Dashboard will in general be far broader and shallower than the Paid Search Manager or Display Dashboards, which would be channel-specific and far more granular

Dashboards visuals generally fall into a number of major purposes:

* Monitor Performance – High-level mission control views to monitor on general performance on a regular basis to ensure that day-to-day performance is going according to expectation, and there are no major anomalies in the data (e.g. If one of your channel systems goes down, for instance, marketing leadership may immediately notice a drop in delivered impressions)

* KPI and Goal Monitors – A data-driven organization always measures KPIs and tracks it to strategic marketing goals. It’s important that every member of the marketing team keeps close tabs of how they are tracking to their goals, and constantly making the required adjustments to make sure they hit them

* Compare Longitudinally – Time is one of the most important dimensions in marketing analytics, and most growth organizations will want to ensure that certain important metrics (like Attributed Revenue or Return on Ad Spend) are growing month-over-month or year-over-year (particularly for seasonal businesses)

* Root Cause Analysis – These are drill-down widgets that allow you to look at anomalies and dig deeper into what might be causing a particular trend. The ability to get granular is a crucial part in being able to answer “Why?” questions and derive smart insights that can be used to take action and optimize wisely

Data Integration

With so much data available these days, the challenge is to consolidate it all and extract clear, actionable insights. Finding a platform that can systematically integrate data from various sources will help to tame your big data madness.

Data Quality

When it comes to data, many marketers intuitively believe in garbage in, garbage out. The data used in attribution modeling needs to be harmonized and cleaned to a common level of granularity so that it is useful. Utilizing data from various sources guarantees disparate data and finding a way to correlate it is critical.

Data Silos

Data silos generally refer to a particularly insidious issue in marketing intelligence that has arisen from the precambrian explosion of channel-specific systems over the past 20 years. As the number of ways for a marketer to reach their audiences through digital media have grown (and continues to grow), point solution systems have emerged to supply planning, workflow and optimization tools for those channels. These tools have generated an ever-growing mass of data, and marketing organizations have typically kept these point solution data as separate siloes to keep their channel teams’ management and optimization processes streamlined.

Unfortunately, data silos gave rise to a number of problems that have gotten in the way of providing reliable marketing intelligence (and many marketing intelligence systems have simply ignored many of these problems)

* No Omni-channel View – When data remains fragmented in siloes, then marketing leaders are not able to truly understand, at a holistic level, everything that is going on across their media. Many marketing organizations have taken to exporting reports from different systems, and manually patching together reams of unreconciled Excel spreadsheets together, an error-prone and time consuming task that often arrives too late after campaigns are already over, all in order to simply understand what is happening at a high-level

* Duplicated, Unreconciled Data – Most systems have mechanisms to optimize for their own channel. These often require firing a conversion tag when a user reaches a success event within the advertiser’s website or mobile app. Unfortunately, each channel system is firing and measuring its own conversion events in an unreconciled way, leading to each channel system claiming credit and resulting in the over-counting of conversions.

* Potential Bias – Several channel systems have stepped forward to offer themselves as a solution for consolidated channel tracking, but many of these systems are owned by enormous media owners. If the systems that are evaluating performance are also owned by the media owners who are being evaluated, then the potential for introducing bias is great

Deterministic Identifier

A deterministic identifier is an identifier that can be definitively tied to a specific user’s device.

The most common deterministic identifiers include:

* Cookies — which, despite many actions taken by Safari and Chrome, remain an indispensible identifier in the desktop and mobile web world

* IFAs — identifiers for advertisers, which are primarily used in the in-app world. Android and iOS platforms maintain their own proprietary scheme for device identification

* PIIs — short for Personally Identifiable Information, this refers to data that can be tied to an individual, such as login info, email, phone numbers, names, social handles and others

Device Fingerprinting

Device fingerprinting often leverage either proprietary or open-source methods for collecting data from digital transactions in order to uniquely identify a user.

This can sometimes lead to a surprising level of accuracy, depending on the technique used. A common fingerprinting mechanism, for example, leverages the specific collection and order of fonts on a user’s device to uniquely identify them.

Many companies may use some of these methods combined with their own. Because these are simply an approximation of the unique user versus s clear delineation of one, fingerprinting is a probabilistic method – and there is a chance that two users may collide and be confused for each other because they have the same fingerprint.

Because each vendor has their own secret fingerprinting recipe, the lifespan and scope of a fingerprint varies from vendor to vendor.

Digital Media

Digital media often refers to all media techniques delivered over the internet or wireless environment, including email, SMS marketing, paid search, paid social, digital video, display, native, digital audio and more. This is in contrast to offline media, which refers to all media techniques related to traditional pre-internet channels

It is often mistakenly referred to all digital media as addressable media which is erroneous because many digital marketing activities, such as advertising on YouTube or Twitter, actually non-addressable outside of the walled gardens’ tools.

Earned Media

The term Earned Media is often used in conjunction with the other two types of media: Paid Media and Owned Media. Earned Media, as opposed to Paid Media or Owned Media, represents word-of-mouth marketing (content that is generally not paid for) that helps build awareness for the brand, or drives visitors into the advertiser’s owned media.

Examples of earned media would include social mentions, likes, reviews, SEO, retweets, recommendations. Producing great content (eBooks, webinars, blog posts, etc…) is also an effective vehicle for driving earned media, because that content can be syndicated and generate inbound links, etc…

External Factors

A strong attribution model will take into account non-marketing elements such as seasonality, major holiday events, macroeconomic factors and competitive activities which can also greatly influence sales.

First Touch Model

A First Touch model is a rules-based model that allocates 100% of revenue to the very first touchpoint of a conversion path within a given lookback window.. For instance, let’s say your conversion path consists of Email >> Video >> Display >> Paid Search >> Retargeting. In a First Touch Model, 100% of the revenue is credited to the Email event since it is the first touchpoint in the conversion path.

Note that events may happen beyond a particular lookback window — and these will not receive any credit. For example, if the marketer has a 30-day lookback window in the above example, and a Paid Social event actually happened 31 days before the conversion event — then that Paid Social event does not receive any of the credit because it occurred before the 30-day lookback window.

Forecast

Attribution is no longer about just looking back to see what led to the desired action, it’s about being able to forecast how shifts in spending will ultimately affect your revenue. Forecasting, or marketing mix modeling, is a great tool to help marketers determine optimal media investments.

Goal Tracking

Goal tracking refers to a practice used by data-driven marketing organizations to measure and keep track of the pacing of their Key Performance Indicators. Well-designed marketing goals and KPIs are designed such that they support even higher-level cross-departmental business goals and KPIs

Granular Data

User-level customer journey data provides a level of granularity that isn’t part of marketing mix models (MMM). The ability to construct the exact sequence of touchpoints leading to a conversion provides a level of insight that can identify correlations between channels and make it possible to optimize your integrated marketing strategy.

Gross Rating Point

GRP stands for Gross Ratings Points, and is used to measure a combination of reach and frequency of a particular ad campaign across the population corresponding to the marketer’s desired audience. It is often used as a measurement of legacy TV reach.

GRP is calculated using the following formula:

GRP = 100 * Reach (% of Target Audience) * Average Frequency

For example, if a marketer wishes to reach females 18-30, and executes a TV campaign that airs on 5 TV episodes for a TV show that reaches 30% of the target audience of females 18-30, then the GRP is 150 (i.e. 100 * 30% * 5).

Homogeneous Data

Disparate data is the root of all evil when it comes to attribution. Mapping data from various sources into a single source of the truth is necessary to establish a homogeneous data set for modeling. Don’t start modeling until your data is homogeneous.

Identifiers

Identifiers are attributes or mechanisms that are primarily used to establish the identity of a user. They are an important building block in much of performance marketing because they help tie different marketing touchpoints (such as ad exposures and paid search clicks) to actual success events (conversion events).

Identifiers come in two flavors: deterministic identifiers (which can be used to definitively identify a user or device) and probabilistic identifiers (which can be used to approximate the identity of a user or device). The most common types of identifiers are cookies, IFAs, PII and device fingerprints/snapshots

Identity Graph

We’ll use the term identity graph and device graph interchangeably. A Device Graph (as per Digiday) is a map that links an individual to all the devices they use. This could include a person’s computer at work, laptop at home, tablet and smartphone. As the internet of things starts increasing the number of connected, digital, IP-enabled devices owned by a user, the identity graph will grow to also include their OTT/Connected TV, smart speaker, and other smart devices. Instead of counting each device as the behavior of a different person, a device graph counts them as one person, so there’s no duplication. Advertisers can then see things like what time of day a person was exposed to an ad and on which device, which helps show what role any particular ad had in a purchase.

Identity graphs consist of identifiers matched up with data assets that help link together different identifiers into something that may represent an individual.

A simple identity graph may consist two identifiers, like cookies, matched together by some shared unique data asset:

a) A more common identity graph might consist of a set of identifiers that have been mapped to a user through an abstract concept such as a User Id. In this case, we’re not tying the identity of the user to some pseudonymized piece of PII information such as a hashed email, but to a unique user identifier:

b) As you can see above, the identity graph attempts to “identify” a user by linking together a series of deterministic identifiers such as cookies, IFAs along with pseudonymized deterministic through hashed emails and cookie synching along with probabilistic links through device fingerprinting.

Identity Resolution Services

Identity Resolution Services refer to solutions providers such as TapAd, Drawbridge, Screen6 and others, whose primary activity is building out, enriching and maintaining an identity/device graph of users. These solutions are often integrated with other advertising systems to offer perceivable customer value to the marketer, such as the ability to provide accurate reach metrics, maintain frequency caps, perform smarter targeting, offer more reliable metrics and more.

Impact solutions such as Radius and Altitude leverage a combination of 3rd party Identity Resolution Services and its own proprietary identity graph, to recognize users across their devices to stitch together omni- channel customer journeys, provide deeper customer journey analytics and calculate more reliable attribution for smarter media optimization.

IFA

IFA stands for Identifiers for Advertisers, and are particularly relevant for the in-app world. These identifiers are maintained by the platforms they are on (usually Apple iOS or Google Android) and are useful for identifying a unique device across all apps on that device. It is typically inaccessible on the mobile browser though.

Like cookies, they are deterministic and consist of a string of 32 alphanumeric characters and are pseudonymous. Unlike cookies, they are controlled completely by the platforms they are on, and typically (with the exception of fraudulent device reset farms) have a long lifespan.

Impact Consortium

The Impact Consortium is Impact’s own proprietary identity graph, used to power Impact’s expansive attribution capabilities.

Advertisers who onboard into the Impact Platform have the option to join into the Impact Consortium. If the advertiser passes in customer identity data (say, their email address when the user logs into the secure area of the advertiser’s site) into our Universal Tracking Tag, Impact captures a deterministic identifier that ties a specific user to a device. When the user logs in across multiple devices, and when the advertiser fires the UTT tag across those devices, then the Impact platform is able to tie the user and their multiple devices.

The Impact Consortium is fully compliant to privacy legislation such as GDPR.

Impression Trackers

Impression Trackers allow platforms like Impact, with its powerful tracking capabilities, to track when the user receives an impression — usually of a display or video ad.

Impression trackers are often trafficked in the advertiser’s ad management system as a 3rd party impression callout. In a web environment, the impression tracker is often an executable Javascript tag, with an pixel trackers as backup for environments that do not allow Javascript to be executed.. In-app requires a dedicated API for Impression tracking since Javascript tags cannot be used within an In-app environment. All necessary contextual information about the publisher is passed along with the image tracker

Incrementality

Incrementality refers to a measurement of advertising effectiveness that can be measured by attribution at multiple dimensions of granularity: channel, campaign, keyword, placement, etc… It indicates the amount of lift to a particular metric (i.e. incremental sales, incremental conversions, etc…) that is brought about by the marketing investment — comparing, for example those who were exposed to or clicked on a particular channel, campaign, keyword, placement, etc… versus one who had not had that touchpoint.

Incrementality can often be measured effectively by more advanced attribution algorithms, such as ones that leverage advanced statistical or machine learning techniques that calculates the likelihood of an increase on the target metric based on the presence or absence of a particular touchpoint in both customer journeys that end in conversion and ones that do not.

Install Tracking

Install Tracking is specific to the mobile/tablet world and allows an advertiser to track when their ad campaigns have resulted in a new install. Many marketers run their own Cost-per-install (CPI) programs to encourage users to download their app and use it.

Since there’s really no way way for you to fire 3rd party tracking code directly in the app store, meaning that there is no way to detect the install event directly from the app store event, most advertisers usually end up firing the Install Tracking Event when it detects that the app has only been opened for the very first time by the new mobile owner.

Intra-Device

Intra-device is particularly applicable to the mobile world, and refers to the ability to recognize a user within the same device, but across mobile web and in-app. Recall that 3rd party cookies are often deactivated in many devices in the mobile web, and unless the user does not clickthru on an ad or affiliate link, there are few alternatives to recognizing the user outside of probabilistic identifiers. When the user goes to a mobile app, on the other hand, there is often a way to recognize the user through deterministic identifiers (IFAs).

Identity Resolution Services that can bridge the gap and recognize users as they move from mobile web to in-app can map out and include the intra-device journey, which can be woven into an overall understanding of the user’s cross-device journey

Javascript Tracking

Javascript Tracking is used to signal events to a Tracking Service in the web environments where Javascript is enabled (which will happen in most cases — most users browsing the web on desktop or mobile will usually have Javascript enabled). In web environments where Javascript is disabled, tracking can usually still be accomplished by image trackers. In the in-app world, tracking is usually achieved through API tracking integrations.

Javascript tracking can be used to track any important web events: impression events (when ads are shown to a user), click events (when ads are clicked on, or when the user lands in one of the advertiser’s properties through an outbound link from an ad or a link from another site, a clickthru from an influencer mention or affiliate link, etc..) and conversion events (when the user makes a purchase, or completes a lead gen form, etc…). It can also measure other related metrics typically associated with web analytics, such as session-level duration, # of pageviews, etc…

KPI

A KPI, or Key Performance Indicator, is a measurement that will directly affect your marketing objectives. They can be identified by examining your strategic business goals, and deciding how to measure your progress towards those goals. Every business has unique KPIs so be sure you are measuring the most meaningful metrics to make more educated marketing decisions.

KPIs sometimes correspond to individual metrics, but more often, they are calculated from a series of metrics you are tracking. One common example of a KPI for advertising is ROAS (return on ad spend). ROAS is a measurement that evaluates gross revenue generated for every dollar spent. The math is simple if you have the tracking data you need. ROAS = revenue from ad campaign, minus the cost of the ad campaign, divided by the cost of the ad campaign.

Last Click Model

Last Click attribution assigns 100% credit to the final touchpoint (i.e. clicks) that immediately precedes a sale or conversion. While last click is important in identifying the closer, marketers should be sure to also examine the introducer (first click) and influencers (middle touches) as well.

Lifetime Value

The Lifetime Value (or LTV for short) captures the total value generated by a particular customer for a given advertiser, usually because of repeat purchases or conversions made by a given customer. Consumers with high LTV are a brand’s most valuable consumers, and many marketers rightfully attempt to locate audiences that increase their average LTV.

Linear Model

A linear model is a rules-based model and one of the simplest ones for those who are starting out when moving from single-touch attribution models to multi-touch attribution models. A linear model allocates an equal amount of credit to all involved touchpoints of a conversion path. For instance, let’s say your conversion path consists of Email >> Video >> Display >> Paid Search >> Retargeting. The linear model allocates an equal amount to each touchpoint, so Email gets 20%, Video gets 20%, Display gets 20%, Paid Search gets 20% and Retargeting gets 20%.

Optimizing marketing channels based on an even model means that the advertiser is rewarding frequency alone but not any external factors such as seasonal or macro-economic factors. An issue with this model is that diminishing returns and relative channel effectiveness are not accounted for as all channels and path positions are credited equally so more spend leads to linearly more conversion.

Note that events may happen beyond a particular lookback window — and these will not receive any credit. For example, if the marketer has a 30-day lookback window in the above example, and a Paid Social event actually happened 31 days before the conversion event — then that Paid Social event does not receive any of the credit because it occurred before the 30-day lookback window.

Lookback Window

A lookback window represents the amount of time (usually specified as a number of days) prior to a conversion that a marketer decides would be a reasonable period of time for a marketing touchpoint to have credibly influenced a customer’s decision to convert. The lookback window is applicable for both single-touch and multi-touch models — both rules-based and machine-learning attribution models.

If a marketing event took place prior to the lookback window, then it is not considered when the attribution model is applied. For example, if the marketer decides to use a 30-day lookback window (meaning, consider only marketing events 30 days prior to a conversion, but no more), then if a paid social event happened 31 days before a conversion, then it would receive no credit whatsoever for that conversion, regardless of attribution model.

For most products, a 30-day lookback window is reasonable and standard. Certain types of products, such as autos and durable goods, may opt for a 90-day lookback window as more appropriate to reflect the longer purchase and decision-making cycle for those types of products

Machine Learning Attribution

A machine learning algorithm leverages advanced statistical techniques such as linear and nonlinear regression, cooperative game theory and other data mining methods, to allocate credit in the fairest possible way possible, based on a touchpoint’s propensity to increase an audience’s likelihood to convert. It looks at all the touchpoints — both the presence and absence of touchpoints — and their role in driving incremental value — looking at both the baseline, converting paths and non-converting paths.

It is often perceived to be the most bias-free of distributing credit, but receives pushback from marketing organizations due to the perceived black box nature of its algorithm, particularly for those unfamiliar with its specific methodology or data science techniques in general. Most attribution vendors will have their own proprietary implementations of data science methodologies and will mix in some of their “secret sauce” in order to provide what they believe, would yield the most optimal set of incrementality calculations for their customers.

Marketing Event

A marketing event represents a trackable event such as an exposure to a display ad, watching a video ad, clicking on a paid search or paid social ad, tapping on an affiliate or influencer link, clicking through from an email or newsletter into the website, etc… These marketing events or touchpoints become the basic building blocks of a customer journey, and can be stitched together to illustrate all of the ways the brand has engaged with their audiences in hopefully persuading them to eventually convert.

Marketing Intelligence

Marketing intelligence refers to the systems, skills and processes that allow marketing organizations to make smart, data-informed decisions, usually through well-designed reports, KPIs, dashboards. For our purpose, we hone in on a particularly important marketing question: how to allocate their marketing spend most effectively based on the ROI and incremental value provided by their different marketing investments.

In order to make informed, holistic decisions specifically around allocating spend, marketers need to look at all aspects of the marketing problem. Marketers thus have to capture information across multiple marketing domains, including customers (which includes current customers and prospective customers), channels, media, customer behavior, sales and more. Marketing intelligence consolidates all this information into a centralized location so that the marketer has an overarching view that they can use to make smart and informed decisions regarding their marketing initiatives and spend.

Note: The use of the term “Marketing intelligence“ can be confusing because it is used quite broadly. For instance, you can read various trade journals and magazines to receive “marketing intelligence“ around the latest developments in the industry. This is not what we mean by “Marketing Intelligence“ though.

Salesforce.com, a salesforce automation tool, may provide some marketing intelligence around the prospect funnel. Marketo, a marketing automation tool, may provide some marketing intelligence around customer engagement on the marketer’s email or landing pages. Many of these martech tools might even have sophisticated KPI trackers, visualization or querying platforms to provide intelligence to specific questions in marketing.

But for our purposes, these are not true “Marketing Intelligence“ systems because they focus on very specific problem siloes rather than providing systems that allow marketers to receive marketing intelligence across the larger marketing universe – across channels, campaigns, devices, audience types and vendors – as a whole – which is necessary for answering the larger marketing question focused on smarter allocation of media spend.

Marketing System of Record

A marketing system of record or marketing source of truth (we use the two terms interchangeably) allows users to consolidate all their data into a single platform, and leverage it for to achieve marketing intelligence by applying various data applications such as KPI/goal tracking, scorecarding, dashboarding, reporting and attribution on top of the consolidated data.

Why do marketing organizations need a system of record?

Because marketing organizations are experiencing an explosion of marketing technologies that have come about in the past few years to deal with the growing complexity and proliferation of channels they have had to oversee. With over 5,000 marketing technologies available in the market, marketing organizations have a harder and harder time gaining visibility into their investments, what media efforts are truly making an impact on their customers, and which marketing initiatives are delivering positive net value.

A Marketing System of Record, does the following:

Collect. Automates the ingestion and consolidation of the marketing campaign data from different systems and different sources stretched out across different channels Reconcile/Normalize. Data consolidated into a single system need to be cleaned up, unified and normalized. Customers who may be recognized by an email address in one system, a cookie in another, and a device id in a third system needed to be reconciled into a single identity Apply. Knowledge-based applications could then be built over this robust source-of-truth for marketing data. This runs a gamut, from analytical applications such as reports, KPI measurement and visualization tools to advanced data applications such as customer journey pathing and attribution analysis

Media Mix Modeling

A media mix model is an econometric top-down model that bridges the online world with the offline one. Media Mix Models are great for assessing whether non-addressable media like TV, radio, print, out-of-home and others are pulled into the media mix model, along with external factors such as macroeconomics, weather and seasonality – all these elements can also be factored into the Media Mix Modeling’s longitudinal statistical analytics.

Media Mix Models, marketers receive guidelines, informed by advanced econometric data, that tell them which factors are most impactful in driving a lift in revenues or conversions, thus giving marketers directional recommendations on how to allocate their media budgets across offline AND online advertising to maximize impact.

Model Overfitting

Overfitting is a modeling error which occurs when a function is too closely fit to a set of data points, which is a no-no in data science and limits the practical usability of a model. One can, in theory, create a model that explains all the data points of a particular test data set extremely well to the point that too many parameters are used to explain away most residual variation (i.e. all the noise). The consequences of using an overfitted model is that the overfitted model becomes ill- suited to explain the behavior of another data set representing the general population, because it has been over tailored to the test data set.

Model Validation

The statistical model used to generate attribution findings should be validated with in-sample as well as holdout sample, or “control“ (a sample of data not used in fitting a model) – the holdout sample is used to assess the performance of the models.

Multi-funnel Conversion

Most conversion funnels are simple – eCommerce funnels often involve only one: Land on the site > shop for a product > add it to cart > checkout > order confirmed!

However, some businesses rely on far more complicated conversion paths, and may leverage multiple conversion funnels, thus we use the term Multi-funnel Conversions to describe this. Conversion funnels often lead to intermediate “success events”, and it’s not uncommon for marketers to optimize towards these immediate “success events” – especially when the conversion process is complex, lengthy and true value only gets realized after the user makes their way through subsequent conversion funnels. For example, marketers may optimize towards driving users to subscribe to a service/create an account, but not necessarily use the service or perform some revenue-generating task. This is when it’s important for platforms to support the concept of “multi-step conversion funnels”

Analyzing the behavior across multi-funnel conversions allow marketers to define multiple “success events” and effectively stitch together conversion funnels. By doing so, marketers maintain a view of their short-term conversion performance (which media is driving the most account signups) but are also able to determine which media investments introduce customers who provide true value in the final conversion of a multi-step conversion funnel process (which media is driving the account signups that eventually perform some revenue generating activity later on).

Multi-touch Attribution Models

Multi-Touch Attribution Models (or MTAs for short) are more complicated than Single-touch Attribution Models. MTAs seeks to distribute credit across more than one touchpoint in a conversion path. One of the biggest deficiencies of single-touch attribution models is that it does not recognize a fundamental fact around marketing and advertising: that is, that that marketing and advertising is usually a “team sport” and that multiple touchpoints cooperate together to convince a prospect to eventually convert.

It’s usually not a one-person effort. Certain type of video advertising may be good in building out awareness. Rich media advertising or email campaigns may be good at building out interest and purchase intent. Paid search may be the final step after the user decides that they already want to make a purchase. All these channels come together to successfully drive a conversion.

Non-Addressable Media

Non-addressable media refers to any media that cannot be tied to a unique user because no unique identifier can be extracted when the ad is delivered. For instance, when an ad is delivered through traditional TV, Radio, or when an ad is printed on a newspaper or on a billboard, that ad is generally classified as non-addressable. This is in contrast to addressable media, which CAN be tied to an individual user, either through a probabilistic or deterministic identifier. For instance, a display ad served in-app can be tied to the user’s device id, making it addressable.

It’s easy to assume that all offline channels are non-addressable. But once more, this is actually not accurate. Direct mail is very addressable, and the cable companies have been rolling out addressable TV to better compete against IP-enabled digital TV (Connected TV and OTT)

Likewise, It’s easy to assume that all digital channels are addressable, but this is actually not accurate. Most marketers cannot retrieve specific identifiers from the Walled Gardens, leading to large sections of the digital marketing universe that remain non-addressable.

Non-Converting Paths

Non-converting paths represent customer journeys that do not resolve into a conversion. This may be because the user has not converted yet, or may never actually convert at all. It is important for marketing intelligence solutions to understand both converting paths and non- converting paths in order to truly understand, from an attribution perspective, how influential different touch points are in truly driving lift and increasing users’ propensity to convert.

Normalized Data

Normalizing data for the purposes of marketing intelligence is the process of organizing data from disparate data sources — often representing different channels and data models — into a centralized repository with data structures that can support all the necessary data regardless of source. Normalization also makes the assumption that the data is de-duplicated and redundancy is reduced, and all important dependencies between the data set are captured in the most efficient way possible

Offline Conversion

Offline conversions refer to success events that happen outside of addressable digital channels, such as sales in brick & mortar locations, or closing a sales through the advertiser’s call center, or closing a lead through a third party agent or franchisee. There is usually enough PII information collected from the offline conversion (information such as names, credit card numbers, etc…) that allow marketers to identify the individual performing the offline conversion.

Through integrations with marketers CRM systems, it should be possible to tie a user’s digital activities (including their marketing journey and online conversions) with offline conversion events.

Why would a marketer want to do that?

Because conversions, whether offline or online, do not happen in silos — and being exposed to marketing messages online has been shown to drive sales in the brick & mortar world. Because of the online/offline divide, too many marketers have taken the easier route of associating offline conversions with offline marketing, and online conversions with online marketing — but customers don’t think and behave in such a simplistic manner. By looking at customer journeys that drive both online and offline conversions, marketers are able to obtain a far more accurate picture about the incremental effects of their digital marketing on ALL types of conversion events

Offline Media

Offline media is often used to refer to legacy media techniques that predated the rise of the internet, such as TV, Radio, Print, Direct Mail, Call Centers, Cinema Advertising, Billboards and more. This is in contrast to digital media, which refers to all media techniques related to the internet

It is often mistakenly referred to as non-addressable media which is erroneous (many direct mail and call center techniques are highly addressable marketing activities).

Omni-channel

Omnichannel is defined as a multi-channel sales approach that focuses on an integrated shopping experience across all channels. Customers may encounter many touchpoints and move between online and offline channels, such as ordering online for in-store pickup. Each channel’s role is considered in relation to others and the customer experience is designed to be seamless and consistent.

Owned Media

The term Owned Media is often used in conjunction with the other two types of media: Paid Media and Earned Media.

Owned Media refers to all media efforts that are in full control of the advertiser, and generally does not incur any variable payout to an external publisher (i.e. An ad creative may be fully designed and built by the advertiser, but in order to disseminate it, you need to pay publishers to place it on their site). Examples of Owned Media include the advertiser’s website, any media properties or microsites they may own, any mobile apps they build, any blogs they maintain, posts and tweets they may do on any social channels they maintain, customer base email marketing they may do, etc… Furthermore, when marketers invest in enriching one’s owned media, it also pays dividends on the Earned Media front (and, to an extent, on the Paid Media front — for example — better quality landing pages on the advertiser’s site can help improve quality scores on their Paid Search efforts).

Paid Marketing

Paid Marketing refers to all initiatives undertaken by a marketing organization that requires some form of payment for delivery of exposure, engagement or conversion from the advertiser’s prospective audience. Paid Media, which is often used to describe advertising-like activities, is a subset of Paid Marketing. Apart from advertising, other initiatives that go under paid marketing could include affiliates, influencers, business-to-business strategic partnerships, local and client brand ambassadors and many more.

Paid Media

The term Paid Media is often used in conjunction with the other two types of media: Owned Media and Earned Media.

Paid Media is often referred to as advertising, and often refers to media exposure that is paid for at either a CPM or fixed-fee typed basis (though much advertising DOES get paid for through alternative payments models like Cost per Click (CPC), Cost per Lead (CPL), Cost per Install (CPI) or Cost per Acquisition (CPA).

Paid Media typically consist of these formats/channels: Standard Banners, Rich Media, In-Stream Video, Digital Audio, Native, Paid Search, Paid Social, Digital Out of Home, and of course, traditional offline formats/channels such as TV, Radio, Print (Magazines or Newspapers), Outdoor, Cinema, etc…

Pathing

Pathing refers to the process of stitching together customer journeys and analyzing their structure, metrics and characteristics for additional insight. Stitching customer journeys together can often be a tedious, difficult and error-prone process, and one that cannot be done manually without technological solutions to support it.

Here are a sampling of difficulties that come from doing this:

1) Collection and Assembly of Touchpoints. Most marketers on average leverage 12+ different systems to manage and launch their campaigns. Cost and transactional data need to be either ingested into a central system, or the data can be captured through Javascript, Image or API Trackers deployed on every system that needs it.

2) Reconciliation of Touchpoints across Devices. Most audiences today own 3-4 internet-enabled devices, and that number continues to do up. Stitching together customer journeys require a deep understanding of assembling a cohesive multi-device identity graph to ensure you have a reconciled customer path across all their devices.

Pathing and Attribution Querying Language (PAQL)

The Pathing and Attribution Query Language, or PAQL for short, is a proprietary patent-pending language for marketers to collect customer journeys, across multiple devices and multiple channels, in order to generate custom pathing and attribution marketing insights.

PAQL can also be used to tailor the behavior of rules-based attribution models in order to fulfill just about any custom business rules required by an organization. PAQL is a highly flexible way of creating bespoke attribution models necessary to meet the unique attribution modeling needs of any marketing organization.

People-based View

People-based view refers to looking at marketing data, not as a series of unrelated marketing touchpoints and events, but as a stream of inter-related events tied to a particular person. This means, the marketing intelligence solution needs to normalize and stitch all marketing events and related data points into a unified customer journey, rather than just flat and unreconciled exposure, engagement and financial data. It needs to be able to recognize that disparate events that appear to be happening in different devices owned by the same user (on the user’s desktop, tablet and mobile), even events that appear to be happening in different parts of the same device (mobile web versus in-app) and recognize that these are all the same person.

When marketing intelligence solutions provide a people-based view, it represents a far more accurate picture about how their marketing initiatives are truly driving prospects into customers, and leads to better optimization decisions versus non people-based views.

People-based views require that the customer journey:

1) Represents de-duplicated events. Without a people-based view, different channel systems may lay claim to their own conversion events, leading to significantly over-estimating how many conversions your marketing is actually doing

2) Represents a cross-device view. Media has long worked in conjunction with each other across a users’ different screens. An old adage says that users don’t think in screens or devices, but marketers, due to the difficulty of managing the proprietary complexities of each device, sometimes have to look at their data device-by-device, instead of recognizing the user behind multiple devices.

3) Represents an intra-device view. Just because it’s not easy to stitch together the probabilistic world of cookies for mobile web with the deterministic world of IFAs in mobile app, doesn’t mean it shouldn’t be done. Though mobile web only represents 15% of the mobile minute, it is still a significant presence in the mobile channel, the fastest growing channel. Being able to recognize the same user across mobile web and in-app within the same device is paramount for a people- based view

PII

PII is short for Personally Identifiable Information and generally refers to data that contains personally identifiable information, such as login info, emails, phone numbers, names, social handles, etc… They can be used to link together other identifiers and are a highly reliable, deterministic identifier.

Because it is information that has been entered by the user, it is often very reliable. It can also be used to bridge the online/offline identification challenge. PII need to conform to the growing demands of privacy legislation such as GDPR.

Pixel Tracking

Pixel Tracking is used to signal events to a Tracking Service in the web environments where Javascript is disabled. Fortunately, in most web environments, Javascript IS enabled, so the preferred methodology is to use Javascript Trackers instead of Pixel Trackers. In the in-app world, tracking is usually achieved through API tracking integrations.

Pixel tracking can be used to track any important web events: impression events (when ads are shown to a user), click events (when ads are clicked on, or when the user lands in one of the advertiser’s properties through an outbound link from an ad or a link from another site, a clickthru from an influencer mention or affiliate link, etc..) and conversion events (when the user makes a purchase, or completes a lead gen form, etc…)”

Post-click Conversion Rate

Post-click conversion rates represent a measure of users who have both clicked on an ad AND converted.

The formula is:

( # of converting users who have clicked on an ad / # of impressions )

Post-impression Conversion Rate

Post-impression conversion rates represent a measure of users who have both clicked on an ad AND converted.

The formula is:

( # of converting users who have been exposed to an ad / # of impressions )

Probabilistic Identifier

Probabilistic identifiers do not rely on deterministic identifiers that uniquely identify an individual. Rather, probabilistic identifiers are simply an approximation of a unique user versus a clear delineation of one. Meaning, there is a chance that two probabilistically identified users may collide and be confused for each other because their probabilistic identifier mistakenly conflates them as the same user. Naturally, the chance for collision is highly dependent on the strength of the probabilistic identification mechanism.

Product Attribution

Product Attribution allows users to run their attribution models at the product-level versus the order or conversion-level. Generally, most attribution companies treat conversions as indivisible units when drawing their path-to-conversion. But, in truth, a conversion could actually represent an order that contains multiple products purchased.

Product Attribution allows the marketer to recalculate pathing and crediting at the more granular product-level instead of simply limiting pathing and attribution at the conversion level. This allows the marketer to see the attributed quantity and revenue for any given product, and is tremendously useful in helping advertisers drive product-specific marketing and media strategies

Query

Most people think of search when it comes to a query – the search query is the word or the string of words entered in the search box of a search engine to access some information on the web. The study of search query trends is key to search engine marketing (SEM) optimization.

Retargeting

Retargeting is a lower-funnel technique to target users who have visited the advertiser’s website. In the E-Commerce world, retargeting may get to the granularity of knowing when visitors visited one or more of their product pages, added items to their cart and/or started but did not finish their checkout. Understanding user behavior at this granular level allows retargeters to employ a series of optimization techniques, such as valuing and bidding more for users who are especially deep in the advertiser’s web conversion funnel, or personalizing the ad through dynamic creative.

ROI/ROAS

Return on Investment (ROI) or Return on Ad Spend (ROAS) are often used interchangeably in the media and paid marketing world to represent the value generated by specific marketing initiatives. It can be analyzed at a channel or campaign perspective, although channel managers who are managing a specific channel can use ROI/ROAS to optimize at a more granular level for their channel, such as ad groups and keywords for Paid Search managers, or placements and ads for Display and Video managers.

Rules-based Model

There are various rules-based attribution models such as last touch, first touch, first and last, position-based, time decay and linear. The difference between rules-based and algorithmic models is that rules-based applies the same rules to each and every conversion while algorithmic learns from your data to continually refine a custom data-driven model. Algorithmic takes into account correlations between your media and even external factors like economic conditions and seasonality.

Single-touch Attribution Models

Single Touch Attribution Models are the simplest kind of attribution model, because they allocate 100% of the credit for a particular conversion to a single touchpoint. The most common type of single-touch attribution model is called the Last Click model, which awards 100% of the credit for a conversion to the touchpoint that generated the last click before conversion happened (assuming that the click happened within a particular lookback window).

Statistical Significance

In the simplest of terms, when a statistic is significant, it simply means that you are very sure that the statistic is reliable.There is a lot of math behind this calculation but what you need to understand is the “p-value“ which represents the probability that random chance could explain the result. In general, a 5% or lower p-value is considered to be statistically significant.

Subway Graph

A subway graph visually shows the marketing touch points along a conversion path according to the “days before conversion” that the marketing touchpoint occurred.

Success Event

A success event is an event within a website or an app which an advertiser decides they want their website/app visitors to do. These can be revenue or lead-generating events such as successful checkouts, or completion of a lead-gen event, or noteworthy intermediary events, such as the creation of a new account, adding items to the shopping cart, or first starting with a lead gen form.

Time Decay Model

Time decay models reward the media touchpoint closest to conversion most and the others receive less credit the earlier they are in the path. Time decay models reward path position regardless of the relative effectiveness of each of the channels in the customer journey. Just like most rules-based models, they ignore external factors.

Note that events may happen beyond a particular lookback window — and these will not receive any credit. For example, if the marketer has a 30-day lookback window in the above example, and a Paid Social event actually happened 31 days before the conversion event — then that Paid Social event does not receive any of the credit because it occurred before the 30-day lookback window.

Trafficking

Trafficking refers to the operational process used to set up and launch an ad campaign. Trafficking processes exist on both the demand-side and the supply-side. On the demand-side, it usually involves ad operations personnel on the media agency side, executing campaign workflows on their 3rd party ad server or demand side platforms.

On the supply-side, it usually involves ad operations personnel on the publisher side, executing complementary campaign workflows on their publisher-side ad server or supply-side platforms. Because so much of attribution relies on firing impression, click and conversion trackers through Javascript or Pixel trackers — these are usually set up and implemented as part of the trafficking process by ad operations people on the demand-side.

TV Attribution

TV Attribution Models allow marketers to understand the impact of various dimensions of their TV advertising, such as the TV Ad Network (ABC, Fox, CBS, NBC, …), TV Program and TV Ad Airing Spot, on their website traffic, online sales and other digital activities. It has long been accepted by marketers that airing commercials on legacy TV can often lead to conversions online — but due to the non-addressable nature of TV Advertising, the impact is hard to accurately quantify.

With TV Attribution Models, marketers can collect TV data in aggregate (GRP), and analyze their incremental impact on driving conversions in the online world.

Unique Contribution/Revenue

When assembling a path-to-conversion, it is quite possible to find a number of paths that consist of one marketing touchpoint followed by a conversion. It is easy to conclude that, if that marketing event had NOT happened, then the conversion event may not have happened at all. We therefore measure unique contribution or unique revenue based on the amount of conversions or revenue delivered by a particular touchpoint where it is present in a single-touchpoint conversion path.

Unique contribution represents the number of conversions driven by a channel or media investment that you would not have received if it was not for that media investment’s sole contribution.

Unique revenue represents sales revenue driven by a channel or media investment that you would not have received if it was not for that media investment’s sole contribution.

With the complex channel overlap of today’s marketing environment, you want to measure each media’s unique contribution and unique revenue so you know where you’re getting the most bang for your buck.

For example, if a marketer, by analyzing their conversion paths, find that 50 conversion paths looked like this: Paid Search Click –> Conversion, delivering about $1,000 in revenue. Then the Unique Contribution of Paid Search would be those 50 conversions, and the Unique Revenue would be $1,000.

View-through Conversion

View-through conversions represents instances where an audience was exposed to a particular ad (in whatever format — display, video or other), but did not click on the unit to go to the advertiser’s website. However, the user may have registered the exposure to the advertiser’s message in their head. They may go to the advertiser’s site at some point in the near future and take the desired conversion action. When the exposure event occurs within the lookback window duration of the conversion event, then we say that the user has had a “view-through conversion”.

Walled Gardens

Walled gardens represent areas in the digital media space where paid media can be purchased, but limited or no tracking data can be extracted at the event level. Most walled gardens will certainly ingest advertiser data in order to optimize campaign performance within the walls of their walled garden, but often do not provide granular (or any) data back to advertiser to allow them to optimize throughout their initiative (across both walled garden and other publishers). The social channels are probably the most famous of the Walled Gardens, including Google YouTube, Facebook, Instagram, Twitter and others.

Walled gardens present a particularly large and existential challenge for marketers and other players within the digital media space since most media dollars are actually captured by the walled gardens

Yield

It’s all about the results. It can take a lot of time and effort to build a statistically-significant, accurate attribution model but don’t lose sight of the ROI. Keep your investment balanced with the potential savings or increase in revenue. You wouldn’t spend $100,000 to save $10,000 now would you? But if you have a large revenue stream, a 1% improvement in close rate can easily pay for your investment in attribution.