Markov Chain Attribution Models
What you'll learn: How to model customer journeys as state transitions and use removal effects to quantify the true contribution of each marketing touchpoint.
Core Idea
A Markov Chain treats the customer journey as a series of states (touchpoints like "Email," "Social Ad," "Organic Search") where users transition from one state to another with certain probabilities until they reach a final state: "Conversion" or "No Conversion."
Unlike rule-based models (linear, time decay, position-based), Markov Chains learn transition probabilities from actual customer paths. The key innovation is the removal effect: you simulate removing each channel from all journeys and measure how much the overall conversion probability drops. The bigger the drop, the more valuable that touchpoint.
How It Works
- Build the transition graph: Map all observed customer journeys as state transitions (e.g.,
Social → Email → Conversion) - Calculate transition probabilities: For each state, determine the likelihood of moving to the next state
- Compute baseline conversion probability: The chance a random user converts given the current channel mix
- Apply removal effect: Remove one channel completely, recalculate conversion probability
- Attribute credit: The difference in conversion probability represents that channel's contribution
Analogy
Think of a city's subway system. Each station is a touchpoint. By analyzing rider flows, you learn which routes people take. If you close one station and ridership to the destination drops 15%, that station was carrying 15% of the value—even if it wasn't the final stop.
Key Takeaway:
Markov Chain attribution models customer journeys as probabilistic state transitions and measure each touchpoint's value by calculating how much conversion probability decreases when that touchpoint is removed from the system.