I want to discuss a topic that every marketer and data analyst loves: attribution. Attribution is the process of understanding what prompted someone to become a customer or a user of your product. It is particularly challenging in today's software environment due to the complexity introduced by multi-device usage and new privacy laws like GDPR, along with changes by major players like Apple and Google in response to these laws.
Understanding where your users came from and what prompted them to become customers involves gathering as much data as possible. Although it would be ideal to get inside every user's head to understand their motivations, we still have a number of tools at our disposal. These tools provide insights that help us piece together a comprehensive picture of our customer acquisition paths. Some common sources of data include browser headers, which tell us what website referred a user to our site. Deep link providers like Adjust, Appsflyer, and Branch help track how deep links were triggered. Survey data, although underutilized, can be valuable after a transaction or sign-up has occurred, by asking users how they heard about your product. Data from paid marketing channels like Meta and Google, as well as data you append to URLs like UTM parameters, can show what campaigns led to user conversions. Promo codes can also link users to specific acquisition sources.
Once data is collected, it is essential to store it in one place, like a database designed to log user sessions and other relevant data without overwriting previous entries. The objective is to analyze three types of attribution: aggregate attribution, individual attribution, and session attribution. Aggregate attribution involves looking at the overall data to determine which channels brought in the bulk of customers, without delving into individual user data. It is probabilistic. Individual attribution focuses on understanding what prompted specific users, like myself, Casey, to start using the product. Session attribution helps identify the sources of individual sessions, which may not relate to the initial acquisition but are vital for understanding user behavior on specific visits. The latter two we should try to be definitive on.
Each type of attribution provides different insights. For example, aggregate attribution indicate our best guess of % of users that came from an acquisition source like Meta, while individual attribution shows specific users cohorted by acquisition channel to understand lifetime value by channel. This allows different targets for cost per acquisition based on the source's effectiveness, reflected in user lifetime values. Building an attribution model is the next step. This model might be last touch, first touch, multi-touch, or rules-based to determine attributions. Smaller startups often start with simpler models and evolve to more sophisticated ones as they grow. A rules-based model ranks data sources by reliability, such as prioritizing UTM parameters, deep links, and browser headers in a particular order based on our confidence in those sources. I usually recommend this method for scaling startups, with incrementality testing for certain channels.
By combining these sources in a structured approach, you can estimate your aggregate, user-based, and session-based attributions more accurately. For aggregate attribution, even when you leverage all of these sources, there will be a remainder, which most companies call direct traffic. This is a mistake. Direct is not a channel. You should estimate what caused the direct traffic based on the other data sources. I like to statistically apply my “how you heard about us” survey responses to people that didn’t answer the survey as one way of handling this. In conclusion, focusing on attribution, particularly after achieving product-market fit, is crucial. As you refine your attribution strategies over time, you will gain better insights and confidence in growing your business sustainably.
Share this post