Introduction
In today’s digital-first world, a customer rarely makes a purchase decision after just one interaction. Instead, they typically pass through several touchpoints before completing a purchase—something especially common in e-commerce. Fortunately, these touchpoints are easier to track than ever.
As marketing becomes increasingly consumer-centric, understanding which channels influence conversions has become crucial. By identifying the right channels, companies can allocate budgets effectively and engage customers in the right place at the right time.
However, businesses often focus disproportionately on the final channel before conversion, overlooking earlier interactions that play an equally important role. To capture this broader perspective, marketers rely on multi-channel attribution modeling.
In this guide, we’ll explore the concept of channel attribution, see how it connects to Markov chains, and walk through a real-world e-commerce case study. We’ll also implement the approach in R to show how it works in practice.
Table of Contents
What is Channel Attribution?
Understanding Markov Chains
The Removal Effect
Case Study: An E-Commerce Company
Implementation in R
What is Channel Attribution?
According to Google Analytics, an attribution model defines how credit for sales or conversions is distributed across touchpoints in the customer journey. For example:
Last Interaction model gives 100% credit to the final touchpoint before conversion.
First Interaction model assigns full credit to the very first touchpoint that initiated the customer journey.
Let’s consider a simple example. Suppose a customer begins through channel C1 or C2, each with a 50% chance. The probability of conversion can be calculated as:
𝑃
(
conversion
)
𝑃
(
𝐶
1
→
𝐶
2
→
𝐶
3
→
Conversion
)
+
𝑃
(
𝐶
2
→
𝐶
3
→
Conversion
)
P(conversion)=P(C1→C2→C3→Conversion)+P(C2→C3→Conversion)
0.5
×
0.5
×
1
×
0.6
+
0.5
×
1
×
0.6
0.45
=0.5×0.5×1×0.6+0.5×1×0.6=0.45
This means there’s a 45% chance of conversion across the given paths.
Markov Chains
A Markov chain describes transitions from one state to another with associated probabilities. It has three key elements:
State space – all possible states (in marketing, these are channels).
Transition probability – the chance of moving from one state to the next.
Current state distribution – the likelihood of being in any given state at the start.
Customer journeys resemble a Markov process because the likelihood of reaching a new channel depends only on the current channel, not on the past sequence.
The Removal Effect
One practical application of Markov chains in marketing attribution is the removal effect. This technique measures the contribution of a channel by removing it from the model and observing the drop in conversions.
For example, if removing C1 reduces conversions from 45% to 30%, the removal effect is:
0.3
/
0.45
0.666
0.3/0.45=0.666
This shows that C1 has a significant impact, though conversions can still occur without it. In contrast, removing C2 or C3 results in no conversions, indicating their critical role in the journey.
Here, channels act as transition states, while the probabilities of moving between them are transition probabilities. Together, they form a directed Markov graph representing the customer journey.
Case Study: An E-Commerce Company
An e-commerce retailer conducted a survey to understand where customers interacted with its products before purchasing. The data covered 19 channels, followed by three final outcomes:
20 – customer has chosen a device
21 – customer completed the purchase
22 – customer undecided
Channels spanned categories such as:
Websites (1–3): company or competitor sites
Research Reports (4–8)
Online/Reviews (9–10): search engines, forums
Price Comparison (11): aggregators
Friends (12–13): social media
Experts (14): online/offline consultations
Retail Stores (15–17)
Misc. (18–19): promotions, campaigns
The challenge: Which channels should the company invest in for maximum impact? To answer this, we turn to R for implementation.
Implementation in R
We’ll use the dataset provided, clean and process it, then apply both heuristic models (first-touch, last-touch, linear attribution) and Markov chain modeling for deeper insights.
Step 1: Install and Load Libraries
install.packages(c("ChannelAttribution","ggplot2","reshape","dplyr","plyr","reshape2","markovchain","plotly"))
library(ChannelAttribution); library(ggplot2); library(reshape); library(dplyr)
library(plyr); library(reshape2); library(markovchain); library(plotly)
Step 2: Prepare Data
Read the dataset
Create a path variable representing each customer journey
Count the number of conversions for each path
Step 3: Apply Models
Heuristic Model: assigns credit based on first, last, or equal touchpoints.
Markov Model: evaluates channel contribution using transition probabilities and removal effects.
Step 4: Visualize Results
Using ggplot2, we compare results across models.
Results
From the analysis:
First-touch importance: Channels 10, 13, 2, 4, and 9 stand out.
Last-touch importance: Channel 20 dominates (as expected since it marks the decision stage).
Linear attribution: Channels 20, 4, and 9 emerge as key.
Overall conversions: Channels 10, 13, 20, 4, and 9 consistently play major roles.
Conclusion
Through this case study, we’ve identified the most influential channels across different attribution models. Markov chain modeling provides a data-driven framework for understanding how each channel contributes to conversions.
For e-commerce companies, these insights enable smarter budget allocation, targeted investments, and more effective marketing strategies. Combined with advanced analytics and AI consulting, businesses can transform attribution data into actionable strategies that maximize ROI.
This article was originally published on Perceptive Analytics.
In Atlanta, our mission is simple — to enable businesses to unlock value in data. For over 20 years, we’ve partnered with more than 100 clients — from Fortune 500 companies to mid-sized firms — helping them solve complex data analytics challenges. As a leading AI Consulting Services in Atlanta, we turn raw data into strategic insights that drive better decisions.
Top comments (0)