With major legal and technical changes going on in the area of customer privacy, here’s an overview of all the important things a marketer should know about tracking and attribution in 2021.
Exactly two years ago we released a blog post covering capabilities and limitations on tracking customer journeys and conversions within Google, Facebook and various other advertising platforms. A lot has changed in the meantime, the privacy landscape has changed significantly, and there’s much more to come.
It is not a secret anymore that tracking users across channels, browsers and devices holistically has become a utopia. Unstable, cookie-based tracking techniques and an increasing complexity in user journeys have made it difficult to combine various user touchpoints into a single journey, with clear attribution and ROI.
There has been a lot of development in the industry since our post back in 2018. It seems that today even the large platforms are struggling to provide complete pictures on conversion paths. That’s why we’re taking a shot to shed an updated light on the topic and see what has changed since then. Let’s start with the main challenges marketers are facing today when it comes to tracking and attribution.
Most measurement methods on the web are based on cookies, and we know that cookies were caught in the crosshairs of data regulators in recent years. The majority of third-party cookies, which are critical to measuring advertising effectiveness, are now either blocked by ad blockers, rejected by users who visit a website, or even blocked by browsers.
The Global Web Index has found out that 48% of internet users are using an ad blocker. Furthermore, research has shown that in some cases only 10% of users answer with “Yes” to all cookie consent boxes, accepting all types of cookies used by websites.
To successfully track buying journeys, each touchpoint in a conversion chain has to pass information to the next part of the chain, which has become almost impossible today. With all the recent and announced changes, cookies have become an unstable and inaccurate technique for mapping customer journeys, and the emerging alternatives are still not really established, failing to provide us with the much needed confidence.
Not only the measurement techniques, but also the users themselves have become more complex. Journeys tend to become longer, more diversified and therefore more fragmented than before, as Blake Morgan wrote for Forbes in a recent article. Here are three examples of recent developments that emphasize this increase in fragmentations of customer journeys:
In the last two years we have also seen a cultural shift when it comes to data protection and customer privacy. Data protection issues have increasingly moved onto political agendas, and users are now much more aware of how their data is exchanged between advertisers and platforms while surfing the Internet. And consequently more cautious.
The invalidation of the Privacy Shield, the introduction of GDPR in the EU, and the upcoming ePrivacy regulation have created a great uncertainty in the industry. Furthermore, Apple, the only company among the four tech giants that does not generate most of its revenue through advertising, has been very active in restricting access to data generated by its devices.
In 2018 Apple launched Intelligent Tracking Prevention (ITP) for Safari, which is blocking all 3rd party cookies, and have announced that iOS 14 will no longer be sharing the IDFA identifier with advertising platforms. Facebook and Google regularly adapt to these new conditions, with some unexpected turns - Google announced in January 2020 that they will also block all 3rd party cookies in Google Chrome in the next two years.
So, you as a marketer now might ask yourself: is anyone still capable of tracking customer journeys and conversions? In our blogpost from two years ago we came to the conclusion that it can be done if you utilize the advertising platforms of the two digital marketing giants: Google or Facebook. These two companies own large parts of today’s internet ecosystem and have access to vast amounts of data. Now, let’s have a look where these two platforms stand today in regards to attribution modelling.
Over the past two years, Google has released several features in favor of more successful tracking customer journeys, but also several ones impeding it. With the introduction of server-side tagging for Google Tag Manager, web analysts can now trigger tags from a server container rather than in the browser frontends of individual users. A new possibility associated with this feature (not officially announced by Google) would be the technical ability to trigger tracking pixels that would have been previously blocked by ad blockers or users not accepting all cookies.
On the other hand, Google announced to phase out support for third party cookies in Google Chrome by 2022, similar to what Safari is doing already. As these two browsers accumulate over 80% of today’s web traffic this would mean the definitive end of third party cookies.
When Safari announced to start blocking 3rd party cookies three years ago, Google answered with several technologies to circumvent cookies from being blocked, such as Conversion Linker, GTAGs or Parallel Tracking. Furthermore, Google has been extrapolating missing conversion paths by using self-developed AI models. With a constantly increasing awareness of users on the potential issues of privacy and data governance, Google seems to have shifted its focus more towards protecting user privacy.
In our previous blog post we mentioned the potential of Google Attribution 360. In October 2019, Google, however, deprecated this enterprise-focused beta feature, and two months later launched Attribution Beta for Google Analytics. The majority of features around this tool are free of charge. Google Attribution gathers data from different advertising systems and analyzes it cross-device and cross-channel. In the end, the tool’s users will be provided with paths that led to a conversion, model comparison features, and specific Google Ads Analytics.
Google also published a small paper that describes the attribution methodology used for assigning credits to touchpoints in this tool. It mentions that only clicks are taken into account, meaning that advertisers investing heavily into top-of-the-funnel activities might not get an accurate result. What’s also missing is a possibility to programmatically retrieve the data via an API or to export it programmatically.
The entire tool is still in an early phase, but data-driven marketers can be excited about what features it has to offer in the upcoming months. In October, Google announced several post-cookie tracking capabilities with the launch of Google Analytics 4. Google, however, officially announced already to use several features to extrapolate missing user journeys. Russel Ketchum, a product manager at Google Analytics, puts it like this:
"The norm is that we’re going to have a mixed set of data: we’ll have event data, but not necessarily a user identifier associated with it. We’ll have gaps in data altogether and this is going to be true of all of all measurement providers."
Some reports by Google Analytics users have already revealed that the platform still has some bugs in tracking data, but Google reminds us that this new property type is still in the beta phase of development.
Facebook still has one major advantage in comparison to Google: all of its data collection and matching techniques rely on a logged-in user-cookie logic. This enables Facebook to track actual users instead of cookies, and to track users more easily and precisely across devices and browsers. Furthermore, many top-of-the-funnel campaigns happen on Facebook, and they have access to a vast amount of impression data that can be also taken into consideration for conversion paths.
In recent times, however, Facebook also faced multiple events that limited them in tracking users across devices and browsers. One of these was the update of iOS 14, the operating system powering Apple’s mobile devices, released in September. With the new iOS version, users can choose if they want to be tracked by any of the apps installed on their Apple mobile device. That’s a major change that will likely have a ripple effect: by allowing users to reject tracking, it’ll reduce the amount of data that’s collected, preserving user privacy.
In our previous post we reviewed a beta version of Facebook Attribution. In the meantime, this feature was released to all advertisers and is now officially out of beta status. The platform is leveraging one of its biggest assets here and uses the enormous amount of logged-in user data to track cross-device user journeys.
Facebook Attribution now also offers access to conversion data via an API (good news: Adverity offers a connector to Facebook Attribution to its users). Furthermore, several new connectors have been added to support integration of other ad platforms, such as Adform, Criteo or Pinterest. A big advantage is that also unpaid, organic engagements on the Facebook ecosystem are also taken into account for the attribution model.
One downside to mention is the integration with Google Ads, that seems to be a bit “hacky” - Facebook requires a script to be set up in Google Ads to push Campaign and Ad Group IDs to Facebook. In some cases I experienced the matching automations to not work properly, which resulted in poor data quality.
However, it seems Facebook invested a significant amount of resources into the development of this feature over the past two years. Especially with the reduced efforts from Google, Facebook’s tool might have the potential to become a new industry standard for attribution modelling.
Over the last two years, another major player has entered the attribution modelling market. In 2018 Amazon Ads launched Amazon Attribution; the tool has been initially available only for the US and UK markets, but has now finally launched in additional European countries. This tool allows marketers to see the amount of sales on Amazon that have been generated from channels outside the Amazon ecosystem, such as Google or Facebook.
One downside might be that again clicks from Facebook or Google can only be taken into consideration when attributing conversions to ads. However, for Amazon marketers this is already a big step forward, as before it wasn’t possible to track any conversions coming from Facebook or Google. This tool is again at an early stage, and marketers relying heavily on Amazon can be excited about what will come in the next few months.
Marketers are now asking themselves what is the best alternative to attribution modeling? An alternative that provides a realistic picture of the impact that each marketing dollar has had.
It might be worth evaluating techniques that rely on more robust, non-micro-level measurement techniques. Marketing Mix Modeling is a technology that has long been used, and with MMM it is possible to both understand and measure the impact of marketing spend on sales. It does not correlate marketing spend with revenue on a user level, but it uses regression models to measure the impact of marketing activities against a target metric at a macro level.
In this way, not only can offline sources such as TV or print be additionally integrated into the calculation, but basically any possible impact on sales can be taken into consideration. These influences can be, for example, the number of streamed podcasts, the number of webinars offered, or the prices of competitor products or services. Even if this technique does not work as ad-hoc as attribution modeling, it could experience a stronger digital renaissance due to the increasing uncertainty in user-based tracking.
Of course, this is just one of the potential solutions, and we’ll see what the future brings, as the current trends seem to be relying less on technology, and more on old-school marketing methods and tools that might not be so precise, but are far from useless. If marketers combine them with available 1st party data and a bit of AI-supported data analysis, they will get a pretty good picture of how their customers move from awareness to purchase.