For an exceptional progression in Online Marketing and Enhanced E-commerce solutions, we need to decipher the 3 simple yet major components, Consumer Requirements, their Next Move and the continual Shift in the Market Trends. The world is going online, and so are we. The most valuable asset for any online business is its DATA. But without the adequate tools and tech, we are just underestimating the true power of it. For an instance, while exploring our dashboards and reports, we often come across questions like,
“What section of the funnel had the majority of drop-offs?”
“What are the GA’s top and the low performing Conversion Paths?”
“How can we predict the conversion rates for the next quarter?”
“Is there a way to predict upcoming Campaign’s conversions?”
“Which feature need optimisation in next release?”
Well, a sequence of meaningful systemic changes can help us figure that out. Besides, leveraging state-of-the-art analytics to model the role of various touchpoints (there is a certain probability associated with it while moving to each of the other channels) we can explore more.
It may seem abstract to understand, so let me put it this way: Imagine a process that begins with a new customer, on your website or application, and ends with or without a conversion. There are marketing channels with which a consumer might interact (organic, paid, social, etc) and then you can notice how these touchpoints are connected. This can be implemented using Markov Chain Analysis, which has been also adopted by giants like Netflix in their indomitable recommendation systems.
In simple terms, it helps us understand the transition of a user from one touchpoint to other in the funnel, and eventually the subsequent conversions. It also helps in forecasting the probability of occurrence of a particular event for time under consideration.

The Reality: Product management can be challenging and daunting. Bridging the GAP between multiple teams, coordinating release schedules, defining product vision, product planning and execution throughout the product lifecycle and endless other activities, it’s easy to get caught up in the hustle, bustle, and stress that is product management. Time to pause, take a breath, and re-energize yourself if you want to know your customers' next move. Because we have Markov Chain to the rescue.
FAQ no. 1: So, what are the prerequisites; oh, wait isn’t GA already doing this?
Well, the prerequisite is just User-level data which is available in the GA console. The data in GA reports is aggregated and it restricts further drilldowns. Additionally, we cannot take the ‘account paths’ those without any conversion. Although, it does provide us with a structured data for Markov Analysis. In case, we are willing to find the User-Journey analysis for the mobile application consumers, the source of data changes to consoles like Appsflyer, Branch.io, Mixedpanel and etc.
FAQ no.2: Do we need to make a model whenever we get a new batch of data points?
Fortunately, no we do not have to. We only have to upload the processed file on our UI and the results will be updated. The team shall need the data for the specific timeline for the website or mobile application and Voila!
So, if you’ve been following so far, you must be wondering how exactly this Markov Analysis output looks like. We have four different output explained below:
1. Paths: The output will consolidate all the unique journey paths that users have followed so far.
The conv column will give me the number of successful conversion number for each path, and conv_null will reflect the unsuccessful counts respectively for each path. Please note that when there are conversions, the conv_null column will be automatically 0 and vice versa.

2. Conversions: The next output will help me understand each touchpoint’s or channel’s conversion proportion. My goal is to maximize the counts of my conversion-channel, which in this case is “Subscription Completed”. It shows me the count of people converted for each touchpoint in the journey. I can compare my individual channel’s performance. You can further arrange them in the order of preference too.

3. Transition Matrix: Now this is one interesting output of the Markov Analysis implementation.
Let me explain this with the help of an example.

Points to note from above transition graph:
- 33% exit from navigation suggests that we need to improve ‘navigation’.
- 50% exit from Filter channel indicates that we need to optimise the channel at competitive level.
A matrix can be made representing each channel’s probability in terms of percentage.
4. Removal Effect: First let’s understand what a removal effect represents. Removal Effect tells us the effect on the conversions (in percentage) if we remove the given channel or touchpoint from our journey. It shows us the percentage of users we can lose if we remove a channel from our journey. For example, as shown in the figure below, if we remove the ‘First App Open’ channel in the customer journey we would lose 51.6% of the total conversions. Hence, it tells us about the channel that isn’t adding much value to our site or application.

Plausible areas where Markov Chain Analysis can come handy:
- To compare the conversion rates contributed by the various touchpoints/channels before, during and post a campaign.
- Additionally, how the conversion rates relate with your investment.
- We can check the contribution from each channel for different time period and combine it with spends to perform a regression analysis to check if the increased spends led to increased channel contribution.
- Brand loyalty and customer behavior can be analyzed at the same time.
- It is recommended that these activities need to be carried out every month to understand your Digital Product vs User demand.
Data helps you listen more to your customers. We hope this was helpful.
Merkle Sokrati offers Analytics Solutions, so if you would like to learn more about this article, please contact us. If you are concerned about your Google Analytics account, we can perform a full Google Analytics Audit.
*Picture source: https://setosa.io/ev/markov-chains/