Chapter 07: Measurement and Attribution

Digital advertising’s greatest promise has always been its measurability. Unlike a billboard on a highway or a commercial on broadcast television, every digital ad carries the potential to be tracked, analysed, and optimised. This accountability is what transforms advertising from a creative art into a data-driven science.

Table of contents

What is measurement and attribution?

The different measurement and attribution models

Viewability and measurement standards

The future of measurement

In this chapter, we will explore the fundamental concepts of how digital media is measured.

We’ll break down the key metrics advertisers use to evaluate success, the technical methods used to capture data across different environments, and the models used to make sense of it all.

We will also examine the profound challenges facing measurement today, as privacy regulations and technological shifts force the industry to reinvent how it defines and proves value.

Key takeaways

• Measurement is the process of quantifying campaign performance through key metrics, while attribution is the science of assigning credit to the marketing touchpoints that led to a conversion.
• These processes are critical for justifying ad spend, optimising campaigns in real time, and understanding the complete customer journey.
• Core metrics include impression-based (CPM, viewability), click-based (CPC, CTR), and conversion-based (CPA, ROAS) models, each answering different questions about campaign performance.
• Measurement techniques vary significantly across environments, from cookie-based tracking on the web to SDKs and mobile advertising IDs in apps, and household-level identification in connected TV (CTV).
• The primary challenge facing measurement and attribution today is signal loss, driven by the deprecation of third-party cookies, privacy regulations like GDPR, and the opaque nature of walled gardens.

What is measurement and attribution?

While often used together, measurement and attribution are two distinct but deeply connected disciplines. Understanding their individual roles is the first step toward appreciating how they work together to provide a complete picture of advertising effectiveness.

Measurement is the process of collecting, aggregating, and reporting on advertising data. It answers the question, “What happened?” The role of measurement is to quantify the performance of a campaign using a defined set of key performance indicators (KPIs).

This includes tracking fundamental metrics such as:

• How many times an ad was shown (impressions)
• How many people saw the ad (reach)
• Whether the ad was actually visible on the screen (viewability)
• How many users clicked on the ad (clicks)
• What actions users took after seeing or clicking an ad (conversions)

Measurement is the act of gathering the facts. It is the raw material from which all strategic insights are derived.

Platforms like ad servers, DSPs, and analytics tools are all fundamentally measurement systems, designed to count and categorise user interactions at scale.

Attribution, on the other hand, is the process of assigning credit for a desired outcome, typically a conversion, to the various advertising touchpoints a user was exposed to along their journey. 

It answers the question, “Why did it happen, and which touchpoint was responsible?”

For example, if a user sees a display ad on a news site, clicks a social media ad a day later, and finally makes a purchase after clicking a search ad, attribution seeks to determine how much value each of these interactions contributed to the final sale.

As we explored in Chapter 02, there are many models for this, from simple last-click approaches to complex algorithmic systems.

The role of attribution is to connect advertising spend to business results, providing the intelligence needed to allocate budgets effectively.

Why is it important?

Measurement and attribution are not merely technical exercises; they are fundamental to the business of advertising.

Their importance stems from several key functions:

1. Justifying investment and proving ROI: In an era of tight budgets and high expectations, marketers must be able to prove that their advertising spend is generating a positive return on investment (ROI) or return on ad spend (ROAS). Measurement provides the data to calculate these figures, while attribution connects that return to specific campaigns and channels, justifying their existence and informing future investment.
2. Enabling campaign optimisation: Programmatic advertising is not a “set it and forget it” activity. It is a process of continuous optimisation. Real-time measurement data allows advertisers to see what’s working and what isn’t, enabling them to shift budgets, adjust bids, and swap out creatives mid-campaign to improve performance.
3. Understanding the customer journey: Modern consumers interact with brands across dozens of touchpoints before making a decision. Attribution helps map this complex journey, revealing how different channels work together. It might show, for example, that video ads are excellent for building initial awareness, while email marketing is effective at closing the sale. These insights are invaluable for developing a holistic marketing strategy.
4. Informing strategic planning: Over the long term, measurement and attribution data provide the strategic insights needed for high-level planning. By understanding which audiences, channels, and messages have historically performed best, advertisers can make more informed decisions about where to allocate future resources for maximum impact.

Without robust measurement and attribution, advertisers would be flying blind, unable to distinguish effective strategies from wasted spend.

Key measurement metrics and how they are measured

As discussed in Chapter 03, digital advertising performance is evaluated using a hierarchy of metrics, each corresponding to a different stage of the user journey.

Impression-based metrics

Impressions are the most basic unit of digital advertising, representing a single instance of an ad being displayed. 

An impression is typically counted when a user’s browser requests the ad creative from the ad server. This is known as a “served” impression. However, this count doesn’t guarantee that the ad was actually seen by the user.

To address this gap, the industry developed the concept of viewability. The Interactive Advertising Bureau (IAB) standard defines a display ad impression as “viewable” if at least 50% of its pixels are in view on the screen for at least one continuous second.

For video ads, the standard is 50% of pixels in view for at least two continuous seconds. Viewability is measured using JavaScript tags from verification vendors like Integral Ad Science (IAS) or DoubleVerify.

Click-based metrics

Clicks indicate active engagement, where a user has been interested enough by an ad to interact with it. 

When an ad is served, it is wrapped in a tracking link. When a user clicks the ad, they are first redirected through the ad server, which records the click, before being sent to the final destination landing page. This redirect happens almost instantaneously.

Click-through rate (CTR) defined as (clicks ÷ impressions) × 100, measures the percentage of impressions that result in a click and is a primary indicator of creative effectiveness and relevance.

Conversion-based metrics

Conversions are the ultimate goal of most advertising campaigns, representing a specific, valuable action taken by the user. Conversions are typically tracked by placing a pixel or tag on the confirmation page that appears after a user completes an action (e.g., a “thank you for your purchase” page). 

When a user who has previously seen or clicked an ad reaches this page, the pixel “fires” and sends a notification back to the ad server, which records the conversion. With the decline of cookies, server-to-server tracking (using APIs) is becoming more common as a more reliable alternative.

Key conversion metrics include cost per action/acquisition (CPA), which measures the average cost to generate one conversion, and return on ad spend (ROAS), which calculates the total revenue generated for every dollar of ad spend.

How are ads in different environments measured?

As we covered in Chapter 06, advertising now runs across a diverse set of channels, and measurement methods must adapt to the technical realities of each environment.

• Web advertising: For years, measurement on the open web has relied on cookies and JavaScript-based pixels for tracking impressions, clicks, and conversions. This has provided a relatively standardised, albeit imperfect, way to follow users across different websites.
• Mobile app advertising: Because mobile apps do not support cookies, measurement relies on different identifiers. This is primarily done through software development kits (SDKs) embedded in the app, which communicate with the device’s unique mobile advertising ID (MAID) – either Apple’s IDFA or Google’s GAID. However, Apple’s App Tracking Transparency (ATT) framework has significantly limited access to the IDFA, making app measurement more challenging.
• Connected TV (CTV) advertising: CTV is one of the most challenging environments to measure due to its fragmentation. There is no standardised identifier like a cookie or MAID. Measurement often relies on household-level IP addresses or proprietary device IDs from platforms like Roku or Samsung. Connecting CTV ad views to conversions on other devices requires complex cross-device attribution modelling.
• Digital audio advertising: Audio measurement faces unique hurdles, as consumption is often passive and occurs in screenless moments. The IAB has developed standards like the video ad serving template (VAST), which includes support for audio ads, to standardise delivery and measurement. However, linking audio ad exposure to a later online action remains a significant attribution challenge.

The challenges with measuring and attribution in AdTech & programmatic advertising

Despite its sophistication, the world of digital measurement is facing a period of unprecedented disruption. Several powerful forces are converging to make tracking and attribution more difficult than ever before.

• The decline of third-party cookies: As discussed in Chapter 04, the phase-out of third-party cookies by major browsers is the single most significant challenge to web measurement. Cookies have been the technical linchpin for cross-site tracking, retargeting, and attribution for decades. Their absence breaks the traditional model for connecting ad views on one site to conversions on another.
• Walled gardens: Platforms like Google, Meta, and Amazon operate as “walled gardens,” meaning they do not allow independent, third-party measurement vendors to track users freely within their ecosystems. They provide their own reporting tools but prevent advertisers from getting a unified, cross-channel view of the user journey outside of their walled gardens. This forces advertisers to analyse performance in silos, making true multi-touch attribution nearly impossible.
• Cross-device and cross-channel complexity: The modern consumer may see an ad on their CTV, research a product on their work laptop, and make a purchase on their personal smartphone. Linking these activities to a single user is a monumental technical challenge, especially without consistent identifiers across environments.
• Privacy regulations: Laws like GDPR in Europe and CCPA in California impose strict rules on how user data can be collected and used for advertising purposes. These regulations require user consent for tracking and have increased legal and compliance burdens, further complicating measurement efforts.
• Ad fraud: Invalid traffic (IVT), generated by bots and other non-human sources, remains a persistent problem. Fraudulent impressions and clicks can skew measurement data, leading advertisers to make decisions based on inaccurate information. While verification companies have become adept at detecting fraud, it remains a constant cat-and-mouse game.

Together, these challenges are forcing the advertising industry to move away from its reliance on granular, user-level tracking and toward new, privacy-preserving measurement paradigms.

The different measurement and attribution models

Once a conversion has been measured, the next critical task is to understand what drove it. In a simple world, a user would see one ad, click it, and immediately make a purchase.

In reality, the path to conversion is rarely so linear. It often involves a series of interactions across different channels, devices, and formats over several days or weeks. This is where attribution models come into play.

An attribution model is a set of rules that determines how credit for conversions is assigned to touchpoints in a customer’s journey.

Choosing the right model is critical, as it directly influences how an advertiser perceives the value of their different marketing efforts and, consequently, how they allocate their budget.

Main attribution models

Attribution models range from simple, single-touch approaches to complex, multi-touch systems. While the industry is moving toward more sophisticated methods, understanding the basic models is essential as they often form the foundation of many reporting systems.

Single-touch attribution models

These models assign 100% of the credit for a conversion to a single touchpoint. They are simple to implement and understand but are widely criticised for oversimplifying the customer journey.

Last-click attribution is the most common and often the default model in many analytics platforms. It gives 100% of the credit to the final touchpoint the user interacted with before converting. 

For example, if a user clicks a search ad and makes a purchase, the search ad gets all the credit, even if they had previously seen three display ads and a video ad. Its main drawback is that it ignores all the upper-funnel activities that built awareness and consideration.

First-click attribution is the opposite of last-click as it gives 100% of the credit to the very first touchpoint in the conversion path. 

This model highlights channels that are effective at initiating the customer journey but ignores the interactions that may have been more influential in driving the final decision.

Multi-touch attribution models

Multi-touch attribution (MTA) models recognise that multiple touchpoints contribute to a conversion and distribute credit among them. As we will explore later in this chapter, MTA is a broader discipline, but these rule-based models are its simplest form.

Linear attribution is a model that spreads credit equally across every single interaction in the user’s journey. If there were four touchpoints, each would receive 25% of the credit. 

It’s a step up from single-touch models but still assumes every interaction is equally valuable, which is rarely the case.

Time decay attribution is where the touchpoints closest in time to the conversion get the most credit. An interaction that occurred one hour before the sale would receive more credit than one that happened a week earlier. 

This model is useful for shorter sales cycles where recent touchpoints are likely more influential.

Position-based attribution (U-shaped) is a model that assigns a specific percentage of credit to the first and last interactions (e.g., 40% each) and distributes the remaining 20% evenly among the touchpoints in the middle. 

It values the interactions that introduced the customer and closed the deal, while still acknowledging the assisting role of mid-funnel touchpoints.

Data-driven attribution is the most advanced model as it uses machine learning to analyse all available conversion paths and determine the actual contribution of each touchpoint. It moves beyond simple rules and assigns credit based on the incremental impact of each interaction. 

This is the most accurate but also the most complex model, requiring large amounts of data to function effectively.

Online and online-to-offline attribution models

The models described above primarily operate within the digital realm. However, for many businesses, a significant portion of their sales still happens in physical locations. This creates the need to connect digital advertising efforts with offline outcomes.

Online attribution tracks a purely digital path to conversion, such as a user clicking a display ad and making a purchase on an e-commerce website.

Online to offline attribution aims to measure the impact of digital campaigns on in-store sales. This is a significant challenge that requires bridging the gap between online user identity and offline purchase data. 

Offline to online attribution measures the performance of offline touchpoints to online conversions. An example of this would be a person receiving a coupon or discount code in the post, which is an example of direct mail marketing, and using it to make an online purchase.

Advertisers are able to convert the offline touchpoint with the online conversion by using the coupon or discount code as the linking piece of data.

However, there are other ways advertisers can attribute offline advertising touchpoints to online conversions, such as via vanity URLs, online surveys asking people how they heard about the company, zip/post codes, and time-limited attribution windows.

The primary method for achieving offline to online attribution is through data onboarding, a process we detailed in Chapter 05. 

An advertiser can upload a list of in-store purchasers (e.g., from a loyalty program) to a data onboarding service, which then matches those customers to their online profiles to determine if they were exposed to a digital ad campaign.

Cross-device and cross-channel attribution

Another layer of complexity is added by user behaviour across different devices and channels. 

A user journey no longer happens on a single browser. Effective attribution must be able to follow the user as they move from their smartphone to their laptop to their CTV.

This requires a robust identity resolution strategy, as explained in Chapter 04. 

By using deterministic data (like a login) or probabilistic data (inferred connections), advertisers can stitch together a user’s activity across different environments into a single, cohesive journey. 

Without this, a user on a mobile phone and the same user on a desktop computer would look like two different people, making accurate attribution impossible.

Multi-touch attribution (MTA)

Multi-touch attribution (MTA) is an approach that evaluates the impact of every touchpoint a consumer is exposed to on their path to conversion. 

Rather than relying on simple, rule-based models, sophisticated MTA platforms aim to provide a granular, bottom-up view of performance by analysing user-level event data.

The goal of MTA is to move beyond last-click analysis and quantify the value of upper-funnel activities, such as display and video, that play a crucial “assisting” role.

By understanding how each channel contributes, advertisers can optimise their media mix, reallocating budget from underperforming touchpoints to overperforming ones.

However, the effectiveness of MTA is highly dependent on its ability to track users across platforms and devices. This makes it particularly vulnerable to the signal loss caused by cookie deprecation, the ATT framework, and walled gardens, which limit the user-level data MTA needs to function.

Marketing mix modelling (MMM)

As user-level tracking becomes more difficult, many advertisers are turning to an older, more traditional technique that has been updated for the digital age: marketing mix modelling (MMM), also referred to as media mix modelling.

MMM is a top-down, statistical analysis that measures the impact of various marketing activities on sales, while also accounting for non-marketing factors like seasonality, economic trends, and competitor actions. Unlike MTA, which requires user-level data, MMM works with aggregated data, such as weekly spend per channel and total sales revenue.

MTA vs. MMM

While both MTA and MMM aim to measure marketing effectiveness, they do so in fundamentally different ways.

MMM is increasingly seen not as a replacement for MTA, but as a complementary approach. MMM can provide a high-level strategic view of how different channels contribute to the overall business, while more granular, privacy-safe measurement techniques can be used for day-to-day tactical optimisation within those channels.

Viewability and measurement standards

For many years, the most basic digital advertising metric was the impression – a count of how many times an ad was served from a server. 

However, a “served” impression carries no guarantee that the ad was actually seen by a human. An ad could be served at the bottom of a webpage that a user never scrolled to, or it could load on a background tab that was never opened. 

This gap between an ad being served and an ad being seen led to a critical evolution in measurement: the concept of viewability.

Viewability is a metric that tracks whether an ad had a realistic opportunity to be seen by a user. It seeks to answer a more meaningful question than “Was the ad delivered?” – it asks, “Was the ad viewable?” This shift was a watershed moment for the industry, as it moved the currency of advertising from “opportunity to serve” to “opportunity to see,” bringing a new level of accountability and transparency to media buying.

IAB measurement standards

To ensure that viewability was measured consistently across the ecosystem, the Interactive Advertising Bureau (IAB), along with the Media Rating Council (MRC), established a set of global standards. 

These standards define the minimum requirements for an ad impression to be counted as “viewable.”

The primary IAB viewability standards are:

• Standard web display ads: At least 50% of the ad’s pixels must be in view on the browser for a minimum of one continuous second.
• Video ads: At least 50% of the video player’s pixels must be in view for a minimum of two continuous seconds of playback.

It’s important to note that these are the minimum thresholds.

Advertisers can, and often do, transact on stricter criteria, such as 100% of pixels in view or longer time durations, especially for high-impact brand campaigns.

The technology behind viewability measurement

Viewability is not something that can be measured by an ad server alone. It requires sophisticated, independent verification from specialised third-party companies like Integral Ad Science (IAS), DoubleVerify, and Moat (part of Oracle).

These companies use JavaScript tags (for web) or SDKs (for mobile apps) that are deployed alongside the ad creative. This code continuously monitors the ad’s position relative to the browser’s viewable area and the time it remains on screen, firing a notification only when the IAB standard has been met.

In the fragmented worlds of in-app and video advertising, ensuring consistent measurement used to require publishers to integrate multiple SDKs from different vendors, creating performance issues and technical overhead.

To solve this, the IAB Tech Lab developed the Open Measurement Interface Definition (OMID). 

As we introduced in Chapter 06, OMID provides a single, standardised API that allows any verification partner to measure viewability without needing their own proprietary SDK. 

This has dramatically simplified viewability measurement in mobile and is becoming increasingly important for CTV environments.

The nuances of viewability

While viewability has brought a much-needed layer of accountability to digital advertising, it is not a perfect metric. Advertisers must understand its context and limitations.

Viewability is not a measure of engagement: An ad can be 100% viewable but still be ignored by the user. Viewability confirms the opportunity for impact, not the impact itself. 

It is a foundational metric, but it must be analysed alongside engagement and conversion metrics to get a full picture of performance.

100% viewability is not a realistic goal: Due to factors like rapid scrolling, ads loading below the fold, and varying user behaviour, achieving 100% viewability across an entire campaign is nearly impossible. 

Industry benchmarks for viewability typically range from 60% to 75%, depending on the format, channel, and inventory quality.

Connecting viewability to ad fraud: Viewability measurement is a key tool in the ongoing fight against ad fraud, which we will explore further in Chapter 09. Verification vendors not only measure if an ad is on screen but also use their technology to detect if it’s being “viewed” by a bot or other forms of invalid traffic (IVT). 

A high viewability rate can sometimes be a red flag for sophisticated fraud, where bots are programmed to keep ads on screen to appear legitimate.

Ultimately, viewability has established a critical baseline for media quality. It ensures that advertisers are paying for a genuine opportunity to connect with an audience, laying the groundwork for more meaningful measurement of advertising effectiveness.

The future of measurement

The challenges outlined previously – cookie deprecation, privacy regulations, and walled gardens – are not simply hurdles to overcome; they represent a fundamental shift for digital advertising.

The era of ubiquitous, user-level tracking on the open web is ending.

In its place, a new ecosystem is emerging, built on a foundation of user consent, data privacy, and technological innovation.

The future of measurement will be defined by a portfolio of solutions rather than a single replacement for the third-party cookie.

Advertisers will need to master a combination of new identity frameworks and collaborative data platforms to continue measuring performance effectively.

Let’s explore the three pillars that will support this new landscape.

Universal ID solutions

One of the most direct responses to the demise of third-party cookies is the rise of universal ID solutions.

These identity solutions generate identifiers that can be used to recognise users across different websites and platforms in the open programmatic ecosystem, enabling many of the measurement and attribution functions that cookies once supported.

Unlike anonymous third-party cookies, most universal IDs are built on deterministic, first-party data, such as a hashed and encrypted email address or phone number that a user provides when they log into a website.

When a user logs into different websites that are part of the same universal ID network, that common identifier allows them to be recognised across domains.

Several solutions are competing for adoption, including ID5, The Trade Desk’s Unified ID 2.0 (UID2), and LiveRamp’s RampID.

The primary role of these IDs in the future of measurement is to restore a degree of addressability for cross-site frequency capping and attribution.

However, as noted in Chapter 04, their primary challenge is scale. 

Their effectiveness mostly depends on how many publishers can successfully encourage users to log in, and no single solution has yet achieved the near-universal reach that cookies once had.

Data clean rooms for attribution

Perhaps the most significant and rapidly growing solution for the future of measurement is the data clean room.

As we detailed in Chapter 05, a data clean room is a secure, privacy-preserving environment where two or more parties can collaborate on their first-party datasets without either party having to share its raw data with the other.

Data clean rooms are becoming essential for attribution in a world without cookies, especially for measuring performance within walled gardens.

The typical attribution use case works as follows:

• An advertiser (e.g., a consumer brand) uploads its ad exposure data (a list of anonymised user IDs that saw a campaign) into a clean room.
• A publisher or retailer (e.g., a large media company or a national supermarket chain) uploads its conversion data (a list of anonymised user IDs who made a purchase).
• Inside the clean room, the two datasets are matched based on their common user identifiers.
• The clean room then outputs an aggregate report showing the advertiser how many users who were exposed to their ad went on to convert, without ever revealing the personally identifiable information of those users to either party.

This allows for accurate attribution while fully respecting user privacy and data governance policies.

Major walled gardens have launched their own clean rooms, such as Google’s Ads Data Hub and Amazon Marketing Cloud, to allow advertisers to perform deep analysis on their data. 

At the same time, independent providers like InfoSum, Decentriq, and Snowflake are enabling data collaboration across a wider range of companies.

Together, solutions will form the foundation of measurement and attribution for years to come, enabling a more responsible and sustainable advertising ecosystem.