Andy Hasselwander, Chief Analytics Officer at MarketBridge
Measurement and optimization remain critical imperatives for marketers. When hundreds of millions of dollars are being spent to acquire customers, CFOs take notice and want to know how various aspects of the marketing mix are working—or not working. The questions of ROI, cost per acquisition, marginal cost per acquisition and overall mix optimization are posed daily to CMOs. Unfortunately, while marketing may be the most scrutinized cost center, it is also one of the messiest when it comes to data.
Even though it has never reached full maturity, multi-touch attribution (MTA) was made possible by digital marketing, specifically cookies. In the early days of the internet, browsing privacy was an afterthought—people were more worried about credit card numbers getting stolen than they were about brands tracking their behavior.
Then came third-party cookies and pixels, which completely changed the game. Third-party cookies can be read across networks and domains, making it possible to “follow a customer” around the internet. They are particularly appealing to marketers as they can track consumer activity across broader stretches of the internet, not just their own sites.
Post-Cookie Measurement Gloom
Using third-party cookies for measurement is getting harder and harder, and it’s a good bet that within a year, it’ll be impossible. Stories of data breaches at large-scale companies like Yahoo and Facebook, along with general news-spurred anxiety about foreign data interference have led to an increase in consumer concerns over data privacy.
Furthermore, in a bid for consumer trust, Google announced in 2020 that it would be phasing out third-party cookies on the Chrome browser. With Chrome’s cookies officially set to crumble by 2023, more and more tech companies like Apple (on both Safari and in-app across devices) are joining the trend towards greater consumer privacy and more data silo-ing.
This decline in consumer information availability has caused uncertainty among marketers. Coupled with other budgetary and data management problems, few marketers feel prepared for the post-cookie world when it comes to measurement. Luckily, there are a few measurement options that are a healthy alternative to cookies.
Beyond Discrete Multi-Touch Attribution
Discrete, user-level MTA is probably dead—but this doesn’t mean that precise marketing measurement is impossible. In fact, discrete MTA always had serious flaws, most importantly its inability to measurement brand-focused, upper-funnel investment, and its digital bias. Media mix modeling (MMM) has slowly chugged along as the measurement and optimization framework of choice for upper-funnel and video-focused marketers—and with good reason. However, media mix modeling is still too high-level to really get to reliable cost per acquisition metrics by investment channel. Something more is needed, even without third-party cookies.
Cohort analysis means replacing individually identifiable data with aggregated groupings for analysis or targeting. These aggregations of individual records protect individual privacy while still allowing targeting and tracking.
Cohorts are very useful to estimate attribution. Cohorts can be created for very specific levels of the business—getting down to very specific audience cells of, say, 10 leads each. This has the potential to get around privacy concerns. It remains to be seen how cooperative ad networks are prepared when it comes to providing cohort data, but marketing analysts should keep their pulse on this trend.
Shift to First-Party Data / Customer Data Platform
Another approach to solving the measurement problem is to bring as much prospect data as possible in-house. By shifting to a syndicated or partially owned system to track first-party data and develop consumer identity graphs, companies can track customers in more private ways.
This doesn’t solve the problem of tracking consumers everywhere, but it can make the bottom of the funnel less opaque. In many ways, CRM systems are first-party data systems. While they were originally designed only to work with existing customers, companies are realizing that a unified identity graph for customers and prospects that leverages first-party data is a competitive advantage.
Aggregated MTA combines econometric techniques with aggregated, channel-level data to get to a more holistic picture of attribution. While it doesn’t provide the magical answer of “what drove this specific lead to come to my website,” it can still be very specific. Aggregated MTA uses the same statistical technique used in MMM, econometric time series analysis, but models many different “last touch” channels as mutually exclusive, collectively exhaustive dependent variables.
What results is a “model of models”—a network path of all of the different ways that a customer can be influenced by, and find, a brand. While aggregated multi-touch attribution can’t tell a marketer how one specific customer got to the purchase point, it can provide serious insight into each channel true contribution, both to its “own” last touch channel, and to other channels.
The Perfect Recipe
Marketers should embrace privacy push-back from consumers. The movement towards digital attribution, while short-lived, showed marketers and executives what should be possible when it came to marketing measurement. Adopting cohort thinking, first-party customer data centricity, and aggregated econometric multi-touch attribution have the potential to give marketers a stable, privacy-compliant foundation to measure across multiple channels.
About the Author: As Chief Analytics Officer, Andy leads the marketing data and analytics functions at MarketBridge, joining the firm (for the second time) in 2018. Andy brings 20+ years of marketing strategy, data science and software development experience to the firm.