As well as addressing the measurement challenges I discussed in part one of this column, advertisers need to get their martech setup in order too. Measurement is hard enough in the modern world when you have a great setup, it’s next to impossible when you don’t.
Core systems need to be well installed and reliable, from analytics to CRM to ad servers, key systems need to be reliable and integrated. Look at the use of advanced techniques that can mitigate some of the tracking issues such as server side tracking, conversion APIs (the Facebook one is particularly important) . All the data should be brought into one place including key publishers like Google, Facebook, DSPs etc. Ideally log (individual) level records of sales, visits and other marketing actions should be ingested that allow for advanced techniques later.
This marketing data setup is a vital component as it allows for data use to be seamless, common reports automated and different attribution types and systems can be compared. Crucially it allows for machine learning and AI to be used on the data as we shall discuss. The setup needs to be considered as modular not set in stone as in a changing world different data sets and technologies will need to be added. Flexibility here really is a superpower. Arguably this section can be the most intimidating but there is plenty of specialist help available.
Fixing measurement and attribution
So how should we fix measurement and attribution? Well there is a lot of detail to cover here and it’s not easy. However, all experts in the space would agree that it’s both possible and desirable to improve from where we are and that this improvement will have a dramatic benefit both to the industry and to individual companies that action these changes. The first step to solving a problem is to admit there is one and advertisers and individual marketers must lean into that problem. Building a new attribution paradigm is built on four key pillars.
Testing and Incrementality.
Quality and Engagement.
Strategy and Framework
Linear tracking, as provided by Google ads, Facebook or a DSP will work on some occasions. Logged-in users is one clear path, universal IDs, cohorts and other techniques will also allow platform tracking to work. However, how it works will not be consistent across all journeys. E.g. it may work if a user is on Chrome for both the publisher and the sale (or other action) but not on Safari or when apps drive to browsers.
The task to understand then is to map out all the journeys that your customers make and identify how likely linear tracking is to work. This can be done either simplistically (by browser for example) or by mapping out more precise journeys that look at apps, different publishers, where users have given consent and more. This understanding helps educate and inform strategic decisions.
For example most advertisers today cannot track well in Safari or from app to browser (particularly on Apple). By understanding what percentage of users this affects and clearly documenting this, it will aid in conversations about how much to spend on Google, Facebook or any publisher that relies on these numbers. By journey you can then estimate how much you spend on this journey, what results you are able to track, estimate how much might be lost. Campaigns and strategies can then be planned based on that journey. As a simple example all display campaigns should be setup distinctly on Chrome vs Safari and other browsers where cookies are not effective.
Testing and Incrementality
Incrementality is the most important metric to understand and it cannot be done without a rigorous testing framework. It is important to know across all the journeys mentioned above and for each channel and campaign what the likely incrementality is. Facebook has excellent (though black box) testing capabilities and incrementality tests should be a regular component of display buying.
Incrementality can also be tested on search too. This can all be done with limited effort – though it must be mentioned that this tracking will be subject to the same issues as the linear tacking it uses. More difficult is testing that looks at the total impact that media has on total sales. Advertisers should be investing in resource to build test and learn programmes, and give these teams the autonomy to continually test current and future plans at scale.
A testing framework can start off simply yet it can also develop. The best advertisers are already building advanced capabilities that allow for AI to work on marketing and analytics data to uncover the true value of media. Media mix modelling was something historically only open to large advertisers with considerable TV spend, modern data techniques and technologies are democratising it and leading it as a solution for the many not the few. In time it may be possible to combine the data from media mix modelling with linear marketing data and have a strong understanding for value from it.
Quality and Engagement
Even with the best attribution techniques it is important to also look at the quality of media being bought and how users are engaging with it. Viewability, dwell time, video engagements as well as adjacent content quality and fraud are all important dimensions to understand. Not for everyone, but for many panels and surveys this can aid understanding. Some new breed data companies can also provide insights on real-world journeys without the bias of tracking techniques albeit at a high level. This data can be combined with linear tracking and testing in order to make better media decisions. Journeys too should be understood with quality in mind.
Strategy and Framework
The final step is to bring it all together into a consistent framework. By going through the above, marketers and advertisers will have a much better view of the effectiveness of their channels. Yet because it involves a myriad of techniques the final and perhaps hardest step is to bring it together. This will mean having a framework of what metrics to be used to make which decisions with clear governance and understanding as well as education for decision makers (including CEOs and CFOs talking to Google/Facebook).
This is an opportunity for marketers at a senior level to critique their current frameworks and look at how top-down frameworks can drive performance effectiveness at a short and long-term level-measuring outcomes rather than proxy metrics. This is tied hand in hand with marketing strategy and company growth and can allow CMOs to truly have the data they need to drive company growth.
Advanced Techniques and Technologies
Advanced techniques and technologies are emerging to help with the world described in this article including identity providers, customer data platforms, data collaboration systems and data clean rooms. AI will ultimately become a vital tool in this space. AI is often lauded as a technique for optimising media yet without a solid view of attribution its efforts are at best wasted, at worst harmful.
AI needs to be effective in attribution. These advanced techniques can be expensive and time consuming and will not suit everyone. For many, getting the basics described above is what’s needed first and perhaps all they need. More sophisticated firms with broader ambitions should lean into the emerging technologies as they can already provide a competitive edge to business intelligence, measurement and also to activation. Many have already made huge strides, though more work is needed as at the sharp end what best in breed looks like is still to be invented and the world is changing fast.
Whether marketers choose only to look at the basics or push further down the rabbit hole I truly hope that in 2022 more marketers meet the measurement challenge head on. If they do it will help in so many areas from advertiser competitiveness and the health of the independent sector.
It’s going to be my focus in 2022 to give CMOs the data they deserve to drive growth for their companies.