Combating Marketing Analytics Measurement Paralysis

Technology

4 min read

Handling organizational questioning and debate by understanding the topics of deterministic vs probabilistic modeling and attribution

Combating Marketing Analytics Measurement Paralysis

The practice of marketing analytics contains a lot of unique characteristics that makes handling this type of business data a lot more challenging than others. It is decentralized, adapts to marketing strategy and the data contains varying data schemas and definitions. These traits contrast the structure and control commonly associated with other types of business analytics like product or financial, and can lead to a lot of organizational questioning, debate and measurement paralysis. Questions like why doesn’t the data align between platforms? What is the source of truth and how should we use this data? are common and a good indication that your organization is facing marketing analytics issues. This is not a place you want to be in with money on the line and while it may feel like an overwhelming challenge to tackle, many of the answers to your problems can be found in understanding the concepts of deterministic vs probabilistic data matching and attribution. These topics are important aspects of marketing analytics that shape the story of a company’s growth.

Deterministic vs Probabilistic Data Matching

Data matching is a difficult topic to communicate and educate to others due to tracking limitations for certain types of marketing like offline advertising or word of mouth (organic), as well as the proprietary measurement systems used by third party analytics and advertising platforms. Each type of marketing will have its own unique method for tracking and matching data with all commonly grouped under two measurement categories, deterministic and probabilistic. Deterministic matching is defined as the ability to track and stitch together an individual’s interactions, while probabilistic matching means there is randomness present in the data requiring probabilistic models to be used to predict the likelihood of an event occurring. Both need to be understood to a certain degree when dealing with marketing data and in the paragraphs below I highlight the areas to focus on for each category.

Deterministic matching is commonly used by third party analytics and advertising platforms (Google Analytics, Facebook Ads, etc.) and each system as it pertains to your marketing efforts needs to be thoroughly understood. The key areas to focus on when understanding these platforms and their unique measurement systems are data tracking, unique measurement protocols and any data caveats or platform limitations. While it might seem like a herculean effort to find and master this material, it is actually quite easy to locate educational material ranging from the platform documentation to brand or user generated online video academies. This content should become second nature to you as you will need to act as a knowledge base for your team. While you would assume marketers know these platforms in and out, many are not aware of what goes on beyond the numbers presented in the platform dashboards and are very likely to misinterpret certain data points without deeper knowledge. You should strive to become an expert in the deterministic platforms you utilize, as this match type is the most widely used in marketing analytics and one that is regularly relied on for source of truth measurement.

Probabilistic matching is used in the absence of deterministic tracking or in cases where an in-house holistic approach is preferred over stitching together deterministic data points. This match type involves statistics and is an evolutionary balance of art and science. Unlike deterministic matching, there are no tracking validation methods outside of commonly used statistical testing tools. This makes it extremely important that those using this technique have a solid understanding of the signals being used and the methodology they created in order to minimize errors and establish trust within the organization. Anything involving statistics can lead to questions and confusion, so it is extremely important to document as you go and keep an open line of communication within your organization so results are not misinterpreted. Trust should be established prior to using this type of measurement within any reporting and it is important for the key stakeholders who developed the methodology to break down its complexities into Feynman-like answers. Simplicity in your explanation is a key quality to focus on for this type of measurement to be successfully used.

Attribution

Customer purchasing journeys can happen over various devices and points in time. Attribution is the key element of marketing analytics that stitches together this journey and credits the marketing efforts that influenced the purchase. Like probabilistic matching, marketing attribution is a careful balance of art and science with an importance on understanding your customers purchasing journey and the attribution reporting possibilities between the various attribution systems that exist within the analytics and advertising platforms you use. These requirements complicate matters because there is no right answer when selecting an attribution model. It is highly dependent on knowledge and gut decisions.

While the decision is ultimately up to you, the most important consideration you can make in the decision making process is to find the most unbiased attribution system (mobile analytics is in a better state than web here) to serve as your source of truth and then determine what the utility will be of using the other platforms for other reporting needs. Your best bet for finding an unbiased system is to research independent or well-established analytics platforms. The utility found in other platforms like advertising is typically in the form of additional data points supporting marketing performance optimizations and incremental lift.

Once you have made your decisions on how the different systems will be used, then it is important to know how to communicate the differences between the systems, how they should be used and the pitfalls present in each. In particular, understand tracking capabilities and the discrepancies that will be present between the platforms despite the similarities in the model and weighting being used. Additional concepts that may be discussed at some point are the limitations of tracking website sessions across devices, advertising platform biases and view-through versus click attribution so it is important to document and educate your organization during this process to prevent questioning of the model selection further down the line.

Taking a Step Back

Dealing with Marketing data is not easy. You need to constantly be on your toes and adapt to the impact that changes in your marketing strategy have on your analytics framework. Change in measurement and tracking are unavoidable, but the impact that change brings doesn’t have to halt growth and progress. By mastering the topics mentioned in this article, having conviction in your decision making and focusing on educating other members of your organization there should be no disruption. This probably doesn’t sound like anything new, but seldom do we take a step back and focus on these actions. Rather than waiting in anticipation of questions in the future, put in the time while you are working on a project to document and think about answers to common questions that may come down the line. By installing this practice into how you work, there will be less disruptions in the future for both you and your organization's growth.

Sign up for my monthly newsletter to receive more content like this: