The Future of Unified Measurement
The privacy landscape has transformed radically over the past few years. Marketers have had to jump through hoops, with the loss of a variety of crucial user signals due to Apple’s IDFA rollout, 3rd party cookie issues, new regulations, browser tracking restrictions, and more. These losses have made it increasingly difficult to track 1:1 attribution in paid media, and this issue persists across all the key biddable platforms.
With regards to IDFA, Apple has taken the stance that their users have a right to complete privacy, and they are currently taking steps to grant users an increased ability to opt out of targeting or tracking of any kind. Users have the power to block any tracking technologies or request the deletion of their data, as far as cookie-based tracking goes. Google has also said (as of their most recent announcement) that they plan to phase out third party cookie tracking from their browsers by the second half of 2024,
While there’s a lot of confusion and anguish around the loss of user signals at the hands of Google and Apple, let’s stop to look at the landscape today. In the US, Safari constitutes 37% of the browser market share, and is a well-known tracking-killing browser. Additionally, Apple’s privacy-preserving iOS and macOS combined make up 46% of the OSes in use in the US. (Source). This means that well before IDFA and Google’s announcement, the ecosystem was already moving towards a increased restrictions on tracking.
Now that marketers are more hamstrung than ever when it comes to attribution on a deterministic basis, there has been a lot of talk around Unified Measurement via modeling. Given the current tracking landscape, the only way to have a holistic understanding of your marketing efforts is with a nuanced, strategic, unified, client-side measurement plan.
What is Unified Measurement?
Forrester defines unified measurement as “a blend of statistical techniques that assigns business value to each element of the marketing mix at both a strategic and tactical level” (Source). Essentially, unified measurement mixes a variety of data, models and insights to provide a 360-degree view of their campaigns. This allows marketers to get granular with their data and comprehensively track campaign success across a range of KPIs.
Tracking in the current privacy environment needs to be a two-pronged approach. If part one is Unified Measurement, part two needs to be Marketing Science.
What is Marketing Science?
Simply put, Marketing Science is a systematic, numbers-driven attribution modeling framework with which you can maximize the insights and actionable takeaways you can get from all of your user data. The following paragraphs will get into some detail.
Platform Models have their limitations
Platform models should definitely be viewed as a key resource (indeed, in many cases they are even considered the single source of truth), but they don’t always tell the full story. For example, platform-specific models may over credit themselves and they simply can’t see everything when it comes to attribution. Facebook is notorious for this – we have sometimes observed up to a 50% disparity between the purchases Facebook claims credit for, versus what shows up on the backend.
If we’re going to look at the big picture of which channels contributed what, we need to remember that platform-based models skew heavily towards themselves. They aren’t built to consider things like cross-channel touchpoints and the incremental impact of each channel on others. Since platforms can miss certain transactions or take credit for others that are influenced elsewhere, it becomes very important to explore other models for a holistic view. This is what Marketing Science helps with.
Some Model Options Marketers Should Consider
MMM (Media Mix Modeling) – This is one of the most popular alternatives to platform models. It’s a top-down approach that evaluates how historical campaigns, offers, pricing, seasonality, and external factors such as economic environment impact sales. It also provides a measured marketing ROI that accounts for other external factors such as unemployment and weather
Experimental Design – This is a method to ensure that all campaign variables such as bid, budget, copy, time of day, device, etc are appropriately controlled to determine the performance of a particular paid campaign tactic. Some of the common ways to do this include:
- 1PA (First-party audience) holdouts
- Causal inference (ML Powered)
Incrementality Testing – This is a statistical technique that allows marketers to measure lift from media performance through the use of control groups. According to AdExchanger, this “allows advertisers to understand the relative values of their media channels compared to one another, as well as whether paid media is driving new users at all, or just reaching users who would have converted organically anyway” (Source).
RFM (Recency, Frequency, and Monetary Value) – As per Investopedia, “Recency, frequency, monetary value is a marketing analysis tool used to identify a company’s or an organization’s best customers by measuring and analyzing spending habits” (Source). This enables marketers to reach specific audience groups with specific behavior-customized communications, resulting in improved engagement rates.
Moving parts to make this work
There are some important components both brands and marketers need to have in place to successfully implement a unified marketing strategy:
This is self-explanatory. The more data we are able to send to these models, the better the models learn, and therefore the better they get at attribution. Look through all your data – business, sales, marketing, etc. and decide which ones to feed to your models. With data, more is always better.
Understanding of Overall Business
It’s key to understand the big picture of your business, and how this marketing effort fits within that larger goal. You’ll need to understand how your online + offline effort works to ensure you’re taking the correct approach. Every business will have to customize their marketing science effort based on nuances specific to them – geographical location, strengths and weaknesses, existing customers, etc
The Right Source of Truth
This is perhaps the most important. There are so many tools to measure performance today – the platforms themselves, Google Analytics or some other clickstream tool, MMP (Mobile Measurement Partners), backend data, etc. Even the most sophisticated models will end up being ineffective if you don’t have the right source of truth.