The Journey From Deterministic To Probabilistic Marketing
For years, brands & marketers have been using 3rd party cookies & tracking pixels to track website visitors, improve the user experience, and collect data that helped them target ads to the right audiences. However, now this is going to change forever. Digital tracking capabilities are more restricted than ever. Third-party cookies are already redundant on browsers like Safari (iOS devices) as well as Firefox and now Chrome will discontinue supporting them from 2023. All apps on iOS require consent from the user to be able to track/monitor them. Law enforcement bodies across the globe are laying strict rules and guidelines for advertisers to restrict online tracking to avoid misuse of any PII (personally identifiable information) data. And above all the growing use of ad blockers by internet users restrict the data flow from one server to the other.
While this is progressing in the right direction as far as protecting users’ online privacy is concerned, on the flip side, these changes are taking digital marketing measurement back to its initial years. Back then, the digital infrastructure was just evolving and the most an advertiser could track from their digital marketing was how many people did they reach, how many impressions were delivered, and how many clicks/interactions came through. A major disruption in digital marketing came almost a decade back with the introduction of remarketing which was built on the back of third-party cookies and along with that followed a whole new eco-system of digital marketing attribution. Now marketers have complete visibility on who came to their website, from what source, using which device, which page of the website they dropped out, what products did they buy & finally retarget them with ads based on their interaction on the website. And the business world was only getting used to how sophisticated they can get when it comes to running as well as tracking their digital marketing campaigns when the lawmakers and the big tech giants decided to overhaul the system all over again.
So, let’s stop for a minute and imagine what the worst-case scenario would be – that all tracking goes away completely no more google analytics, no more data from Facebook you would have no idea who’s coming to your website, who’s buying your products or services, all tracking would be eliminated. How do we survive this situation & what should we do today to prepare ourselves if and when tracking starts to be depreciated? As per the latest update from Google, it will delay the deprecation of third-party cookies on Chrome by another 2 years or so. Therefore, it’s very important to use this time to have a strong plan to transition from what we call deterministic marketing to more probabilistic marketing.
In deterministic marketing, you can be 100% sure that John came from Facebook and brought a hat on your website. Probabilistic marketing on the other hand, only lets you know with a high degree of certainty that someone came from a particular channel or the probability that the user has certain attributes. Having said that, probabilistic marketing has its pros and cons. One of the best things about working in a probabilistic environment is that it’s going to work regardless of what they do to tracking. So, if we can start to train ourselves to rely on probabilistic data today and start to transition how we think about marketing into a probabilistic thought process we’re going to succeed when the big tech players continue to ratchet what tracking is available to us. Another advantage of probabilistic marketing is that it will still allow us to make optimization across all the channels and across various audiences so even if we don’t know specifically that John is John, we can still make a determination about the probability that John is John and that the audience that describes John will behave in this way. Another merit of a probabilistic marketing environment is that it really forces us to focus on the bigger picture without getting too far down in the weeds looking at individual landing pages, conversion points, customer journey, etc. We get to take a step back and analyze across all channels, landing pages, assets & customer journey what’s working well and what levers can we pull in order to increase the campaign performance. Working in a probabilistic environment will allow us to have access to campaign insights much faster as compared to that in a deterministic environment. In a deterministic setup, we need to track every single point of contact, know where the customer is in the journey, where they came from, have cross-device tracking, and all other aspects of tracking put in place perfectly. In a probabilistic setup, we just make assumptions and get close to similar results. It’s not going to be perfectly accurate, but the net results will resonate with what one would expect from having a perfect deterministic marketing setup.
However, not everything about probabilistic marketing is good. At its core, the data itself is less accurate. It is not as scientific as a deterministic marketing approach. A big drawback is that we lose the individual customer journey. All these journeys are grouped together into a probabilistic journey, but we can’t zero in on John to find out exactly what John did, how long he spent on each page, etc. If John comes in and he’s a top customer who did everything right and we want more Johns it’s pretty hard to determine exactly the steps that this one individual did so that we can get more of that customer. Instead, John is lost in the population of other people of who some were good, and some were bad but they weren’t all Johns and so with probabilistic attribution we lose that ability to get down to the individual level to know for certain what each person did. Another demerit is that as consumer preferences change and traffic on the website starts coming from a new source, the system needs to re-learn what this new traffic source is, how well is it converting, and then update the overall probabilities across channels. This disrupts the learning established so far and, in a way, resets the whole attribution pattern all over.
Therefore, to have an accurate attribution or to improve our marketing campaign with higher accuracy based on data, we need to rely on quality data. And typically, that is our own first-party data. We should start creating our own first-party universe and if we are successful in doing so, we should be able to thrive in today’s environment but also be future-proof in the event that cookie tracking goes away. As an immediate step, we should start capturing customer information by incentivizing the visitors on the website to register or login before they are able to access full content on the website. Once we build our own first-party data, we will not be reliant on the likes of Google, FB, and Amazon to target the right audience and get a better return on our advertising spends.
