There is not one measurement method that will suit everyone, and a company should explore what approach best fits the business. This chapter focuses on two popular methods – attribution modelling and attention-based measurement – and explores the pros and cons of each.

  • Attribution modelling
    • Benefits of attribution modelling
    • Challenges of attribution modelling
    • Alternative options around attribution modelling
  • Attention-based measurement
    • Using attention-based measurement

 

Attribution modelling

One of the most popular approaches used by marketers to understand the impact of different elements of a campaign on driving a customer action has been attribution modelling.

Attribution models assign credit to marketing channels to help marketers understand which touchpoints along the customer journey have contributed to a customer action – for example, a purchase, download or website visit, depending on the objective of the campaign. It is about understanding the value of each type of marketing and the role it has played towards driving a sale or action.

Some of these approaches work by allocating credit to just one touchpoint, otherwise known as single-touch attribution models. These include:

  • First touch: all credit is given to the first interaction (e.g. ad view or website click).
  • Last touch: all credit is given to the last interaction (e.g. ad view or website click).
  • Last ad interaction: all credit is given to the last ad that was displayed or interacted with, i.e. it does not include website clicks or other non-advertising related interactions.

Until browsers started blocking third-party cookies and enhancing privacy features for their consumers, making this approach more difficult to implement, it had generally been regarded as better practice to use multi-touch attribution models, which instead gives credit to more than one touchpoint along the customer journey. These options include:

  • Time decay: credit is distributed across the interactions, assigning the least to the first touch, building to the most for the last touch.
  • Linear: credit is distributed evenly across every interaction.
  • Position based: credit is divided in a U-shaped model, where 40% is given to the first and last touch, and 20% is spread across other interactions in between.
  • Custom: credit is divided up between the interactions exactly as the marketer wishes.

Benefits of attribution modelling

Multi-touch attribution modelling enables marketers to take into account all the touchpoints a customer has engaged with and allocates a certain amount of importance to each. Using this information, marketers are then able to see the value that each type of marketing activity and touchpoint has had on driving a conversion or action. This also enables marketers to understand where the opportunities exist to increase marketing efficiency and drive sales.

Challenges of attribution modelling

Many advertisers employ multi-touch attribution modelling to better understand which parts of their digital marketing are contributing to the final conversion, relying on third-party cookies to bring in data about the user’s journey to feed into the model, such as which websites they visited or searches they made.

However, multi-touch attribution models that rely on third-party cookies will become less viable as access to them continues to decline, as without these cookies there is no information being passed across different websites. If there is no visibility of where a customer has been, and which websites they have visited in the run up to a conversion, it is impossible to credit any of those interactions in a multi-touch attribution model.

With third-party cookies no longer being phased out by Google, some companies may continue to consider using them as a measurement solution. However, it is already the case that a big proportion of the open web cannot be addressed by third-party cookies, which makes them increasingly unreliable as a method of measurement. 

Speaking to Econsultancy, Andrew Hood, Founder and CEO at independent analytics consultancy Lynchpin, suggests companies should “view these ongoing industry developments through the lens of browser and device market share – which can be very different for different brands depending on the demographic of their audience. Marketers will need to look at the context of what their addressable audience is or is not via third-party cookies and weight their plans accordingly.”[1]

For those companies who continue to rely on last-touch, a key challenge is that it does not take into consideration how other elements of digital advertising have contributed to a conversion. Allocating budget based on this information means guessing the role of each channel and is likely to lead to wasting budget on some channels.

“If companies rely on last-touch they are undervaluing digital ‘upper funnel’ marketing activity such as generic search, social, video and email, and instead overvaluing lower funnel channels like brand search, voucher code websites and retargeting. In particular, the influence of brand marketing is not being factored in, and it doesn’t account for companies running a range of marketing activities at the same time.”

Gabriel Hughes, CEO and Founder, Metageni

“One of the biggest challenges we see with attribution within Sky is that last-touch always under indexes display and social advertising because people invariably don’t click on a display ad they see. Instead, at some point they will do a search and click through or come to your website directly. It will therefore under index display over search ads.”

Kumar Amrendra, Head of Digital Marketing, Sky UK Ltd

Alternative options around attribution modelling

With multi-touch models that rely on third-party cookies becoming less reliable as an option, there are alternatives companies can consider when seeking to understand the value different elements of their campaign are having on driving a conversion. They can create their own custom attribution models using machine learning and predictive modelling techniques, or work with an agency on this.

Google’s data-driven attribution reporting

Many marketers using Google Analytics had relied on multi-touch attribution reporting in the tool to help determine the impact and value of different touchpoints in driving conversions. On 1 July 2023 Google Analytics 4 (GA4) replaced Universal Analytics, with the move seeing significant changes to the data, metrics and reporting available to users of the tool, in addition to some added and expanded features. One notable update was the switch to ‘data-driven attribution’ as the default attribution model.[2]

Google’s data-driven attribution uses the historical data from a company’s account to determine how people interact with their ads and decide to convert. Machine learning algorithms assign fractional credit to customer touchpoints which may have previously been undervalued. Google’s Smart Bidding can then react to these opportunities, resulting in performance gains, according to Google.[3]

Google describes its data-driven attribution reporting as different from other attribution models since it uses a company’s conversion data to calculate the actual attribution of each ad interaction across the conversion path. Data-driven attribution looks at data from a company’s website, shop and Google Analytics conversions from Search (including Shopping), YouTube, Display and Discovery/Demand Gen ads. By comparing the paths of customers who convert to the paths of those who do not, the model is able to identify patterns and determine which steps have a higher probability of leading a customer to a conversion. The model then allocates more credit to those valuable ad interactions on the customer’s path.

With Google Analytics 4, Consent Mode generates anonymous data which is used to power Google’s behavioural modelling and conversion modelling. By using machine learning, the tool can use the data generated by users who consented to cookies and fill in the blanks to provide a fuller picture of all users. For those users who do not provide consent, Google records pings, which are events that are not associated with the user. The tool can then compare those pings to the actual behaviour of similar consenting users to ‘fill the gap’ of missing user data and provide marketers with a more accurate look at website metrics, while adhering to users’ privacy preferences. The Cookie Information blog article provides some questions and answers for companies wanting to further understand Google’s Consent Mode v2.[4]

Google’s Privacy Sandbox attribution solutions

The proposals for Google’s Privacy Sandbox will enable users to measure the performance of digital ads using the Attribution Reporting API[5] and the Private Aggregation API.[6] These solutions operate as follows:

  • Attribution Reporting API: Measures when an ad click or view leads to a conversion by matching data which is then aggregated and encrypted. The API allows advertisers to gain insights into conversions without tracking an individual’s activity across websites.
  • Private Aggregation API: Allows developers to generate aggregate data reports using data from the Protected Audience API and cross-site data from Shared Storage. It allows a histogram to be constructed from the aggregation of data across users in defined buckets. It differs from the Attribution Reporting API, which is designed to measure conversions, instead having been built for cross-site measurements.

However, a key challenge for marketers is having confidence in the numbers, as often how the attribution has been calculated is regarded as a ‘black box’. In addition to this, each platform’s reporting tools attribute sales in different ways, making it hard to draw direct comparisons if a company is using multiple tools. This is further complicated by the fact that sales can be potentially claimed by different parties as their own. For example, the default attribution setting for a campaign with Meta Ads will be different to Google Ads, and Google Analytics does not include view-through conversions by default.

Custom attribution modelling

By nature, attribution modelling provides a short-term view of results, which makes measuring the effect of a company’s branding a challenge and prevents them from having a full view of how they are performing in the long term. In response to this, some agencies have been developing customised approaches to attribution modelling in order to be better aligned around the entire customer journey.

“We strongly encourage our clients to shift their focus from marketing-centric to customer-centric approaches. The challenge in a marketing measurement context is that the way in which each marketing channel tends to influence the customer journey is very different, so you have to think about both the marketing channel and the customer at the same time. The concept that links the two things together is the customer journey itself. This is why attribution is so important, as it resolves the question of what exactly your marketing contributes to the customer journey and therefore contributes to customer engagement and conversion.

“With this in mind, a central requirement of our custom attribution approach is to apply a different attribution model for different categories of customers. This means that our machine learning predictive attribution models are given information about the customer in order to predict the contribution of marketing to conversion likelihood. The most basic customer segment that must be applied is the distinction between a new and repeat customer.

“It is astonishing to consider that most attribution models attribute new and existing customers in exactly the same way, when we know their purchase behaviour is very different. Machine learning is able to spot patterns in the data which are not obvious to the human analyst, and also incorporate these into a predictive attribution model that blends marketing signals with analytics and human behaviour. In our approach we incorporate available customer data signals into the first-party attribution models we train for clients, thereby leveraging their customer-centric data to help evaluate the ROI from marketing.”

Gabriel Hughes, CEO and Founder, Metageni

Attention-based measurement

Attention-based measurement is seeing renewed interest in recognition of the fact that treating all impressions as equal is not really reflective of the truth of customer behaviour. This is supported by a study from Teads, which found that attention is three times better at predicting outcomes than simple viewability.7] Over a third of marketers (37%) in Econsultancy’s Future of Marketing report indicate they are either using or considering this an alternative form of measurement to using third-party cookies.

Attention has always been at the heart of advertising, but technology means brands can now look beyond traditional metrics around reach and viewability to understand their audiences in new ways. In 2022, a number of big ad agencies developed and scaled their own attention models, as brands around the world turned to this as a measurement option.

Heineken was one brand who developed its first globally tested attention measurement campaign in 2022.[8] According to the brand’s director of consumer connections in Brazil, Fernanda Saboya, attention metrics offered a more “qualified vision of impact” compared to previous measurements.

Using attention-based measurement

There are two main approaches to attention-based measurement:

  • Attention metrics: Measured by JavaScript tags on live campaigns, such as scroll rate and interactivity (including volume up/down, clicks or length of time the ad is in view, to name a few).
  • Attention data: Gathered from panels of opted-in consumers where eye-tracking software is deployed to gather data on what people are looking at, and therefore paying attention to, when consuming media. This data is then used (in combination with other datasets such as survey data and data from tags, as above) to create models to give attention norms to different advertising channels and formats.

The first approach allows marketers to go beyond the usual basics of measuring a campaign, such as if the campaign is merely ‘viewable’ (in other words, if the activity has the opportunity to be seen by a consumer). Instead, it uses more data points to understand if the campaign caught the attention of the target audience.

The second approach allows marketers to get some cross-channel insights about the marketing that is capturing the most attention, and to weight their campaigns accordingly. The suppliers of this data offer planning and measurement tools with which to better understand the impact of marketing activity across digital. This approach offers the further benefit of allowing some cross-media measurement, for example measuring the number of seconds an eyeball on average looks at a digital marketing message versus a TV ad.

Chloe Nicholls, Head of Ad Tech at IAB UK, describes how IAB UK’s members have helped to shape a guide for advertisers on attention as a way of understanding consumer engagement.[9] The resources include the IAB’s Attention Definition Matrix, which covers the four main methods used to measure attention (Figure 1).

Figure 1: IAB UK’s Attention Matrix

A table by IAB UK describing the "attention matrix" in advertising.

Source: IAB UK

The value in using attention as a metric is one which James Sharman, Northern Europe Digital Acceleration Lead at Haleon, describes they are heavily focused on to measure effectiveness throughout the funnel. Citing a study from Teads that revealed that the further down the funnel a customer goes, the more attention is required,[10] Sharman said: “This is the opposite of what we thought before, because normally we would spend longer on the big brand campaigns as we assumed that once people had heard of us, the purchase would be a relatively easy click. Their report changed the way we looked at campaigns with attention becoming much more of an important KPI for us.”

This guide is based on primary research which involved exploring findings from two reports:

  • Econsultancy’s 2023 Future of Marketing report, which was based on a survey of 835 client, vendor and agency-side marketers. The survey was fielded to Econsultancy and Marketing Week’s audiences between 9 June and 3 July 2023.
  • The Language of Effectiveness 2023 report has been produced using responses to an online survey of 1,369 qualifying marketers conducted by Econsultancy’s sister brand Marketing Week between 27 March and 28 April 2023.

In-depth interviews were carried out with industry experts. Econsultancy would like to thank the following interviewees for their invaluable contribution of time and expertise to this guide:

  • Kumar Amrendra, Head of Digital Marketing, Sky UK Ltd
  • Amy Blasco, Partner, Enterprise Data, Experience and Marketing Lead, IBM
  • Laura Chaibi, Director, International Ad Marketing and Insights, Roku Inc
  • Sebastian Cruz, Regional Digital Marketing and Media Director, Shiseido, Asia Pacific
  • Gary Danks, General Manager, AIM, Kochava
  • Mauricio Ferreira, Marketing Effectiveness Lead, Confused.com
  • James Hurman, Founding Partner, Previously Unavailable
  • Gabriel Hughes, CEO and Founder, Metageni
  • Dr Grace Kite, Economist and Founder, Magic Numbers
  • Chloe Nicholls, Head of Ad Tech, IAB UK
  • Roxane Panopoulos, Group Manager, Regional Measurement & Insights – Netherlands and Nordics, Snap Inc
  • Marina Peluffo, Head of Business Intelligence, Prima (speaking as industry expert)
  • James Sharman, Northern Europe Digital Acceleration Lead, Haleon
  • Steven Silvers, EVP, Global Creative and Media Solutions, Kantar

Lynette Saunders is a Senior Analyst at Econsultancy, where
she works on delivering industry-leading research, briefings and
reports for the digital marketing industry and speaks at a number
of external conferences.

Lynette’s previous experience includes delivering web analytics, measurements and insights, as well as leading usability and
customer experience programmes focusing on improving the
overall online customer experience for Cancer Research UK
and the Royal Mail Group.