Course contentsShow
← Data Science
Lesson 1733 of 2,14541. Marketing and Growth AnalyticsFree lesson

Markov Chain Attribution Models

Modeling customer journeys as state transitions and calculating removal effects to determine touchpoint value.

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

  1. Build the transition graph: Map all observed customer journeys as state transitions (e.g., Social → Email → Conversion)
  2. Calculate transition probabilities: For each state, determine the likelihood of moving to the next state
  3. Compute baseline conversion probability: The chance a random user converts given the current channel mix
  4. Apply removal effect: Remove one channel completely, recalculate conversion probability
  5. 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.