We asked our software developer and AI-expert Chamod Kalupahana to answer our most common questions about our SQUID statistical attribution model and Markov Attribution that make up our model’s internal processes.
Q: What is Markov Attribution?
Markov Attribution is a statistical AI algorithm that is built on Markov chain principles, tailored to marketing data.
This new attribution model in SQUID is a better way of analysing how important each marketing source is for your user traffic. Consider you have your user traffic, with how many times they visited your website and what source they came from.
Traditional algorithms, such as “last touch wins” models, will just attribute the source of the user to how they last visited your site. While this works, it’s biased and inaccurate. Markov Attribution looks at all the data of your users, training on it and tells you accurately how important each source is for users visiting your site and converting.
Once you use this attribution algorithm in practice, you can directly see how influential your marketing sources and channels are.
Q: What does the new algorithm do?
Markov Attribution takes all the marketing touchpoints of a user and transforms them into a Markov chain. Like an AI model, the algorithm trains on the data and then assigns credit to each marketing touchpoint.
Q: What are the benefits of using a machine learning algorithm for SQUID
It gives more accurate credit attribution across complex, multi-touch journeys, adapts over time, and can infer influence even when data is incomplete.
Q: Why did you choose Markov Attribution as a basis?
Because it balances rigor and scalability, less biased than last-touch yet more computationally manageable than other AI approaches.
Q: What are the next steps for SQUID’s Self-Learning Process?
We hope to keep improving the model, training on new data and upgrading SQUID with new AI models and techniques.