Causal Analysis of Marketing Campaign

Balaji Muthukrishnan (~balaji49)




Business Problem:

A leading apparel chain in India needed to find the effectiveness of their new in-store campaign ran across different stores during the year 2018. The sales and campaign data for Tamilnadu circle was available to analyze the effectiveness of the campaign, causal impact analysis was used to quantify it.


The tool of choice was either r lang or python, python was chosen as we had already deployed our previous data science projects in the form of python API.

The total increase in sale due to the campaign was to be identified (i.e) cause and effect relationship to be quantified. The causal approach was choosed, various python packages like dowhy, pycausalimpact, causalinference and many more, were analyzed for the problem fitment. Endogenous and exogenous factors affecting sales like promotions and calendar data are identified and collected during the data preparation stage using python package like pandas, scikit learn etc.., last 3 years sales data was available. Based on the availability of the required data, pycausallimpact was finalized.

Basic outline of the talk:

  1. Apparel Business Understanding [4 - 5 minutes]
  2. Campaign Analysis - problem definition [4 - 5 minutes]
  3. How does Causal Impact work? [5 - 7 minutes]
  4. How did we identify and quantify the increase in sales? [10-12 minutes]
  5. Q/A - [2 minutes]

Who is this talk for?

  1. Data Scientist and Machine Learning engineers approaching retail business problems day in and day out.
  2. Anyone who is curious to know about causal problems and how to solve them.
  3. Newbies to Data Science.

Key takeaways:

  1. How data science is applied in business settings?
  2. Significance of business understanding to identify and solve data science problems.


  1. Basics of Apparel business(Will be briefly covered in the talk)
  2. Basics of python
  3. Basics of Causal Analysis

Content URLs:


Slides - First cut, will be improvised

Video Preview

Speaker Info:

Balaji is a 'Machine Learning Engineer' at Pathfinder Global FZCO, Chennai. With 8+ years of working experience in building statistical, econometrics, machine learning and time series forecasting models in academia and industry, he has provided Analytical solutions to Retail, Retail Real Estate, Finance and Telecom domains.

Analytical solutions provided:

  • Merchandising and Replenishment planning.
  • Causal analysis of marketing campaign.
  • Customer 360 - Customer propensity, Customer lifetime value, Recommendations (cross-sell and up-sell), Customer segmentation and Customer churn (contractual and non-contractual settings).
  • Multi-tenant forecasting.
  • Face recognition.
  • Daypart and Product affinity analytics.

He has also delivered corporate training to clients (citizen data scientists) in the areas of Data Science, Statistics and Machine learning. He is passionate about speaking at data science meetups, educational institutions and writing articles.

Section: Data Science, Machine Learning and AI
Type: Talks
Target Audience: Intermediate
Last Updated:

Hi Balaji, thanks for your interest to talk in pycon India 2019.

There are some suggestions after going through your proposal.

  • Your description seems a bit general and broad, can you please be bit more specific about the problem statement and how you solved it?
  • I realise that your talk involves in "Causal impact algorithm". Can you add bit more specifics on how python played a role in that, for example: did you use any ML libraries or you built a wrapper around existing libraries?
  • You can also add an outline of what all you are going to cover in your talk

Also please refer Pycon India speaker best practices

(CFP coordinator)

Naren Ravi (~naren)
The comment is marked as spam.


How to attach slides and videos to the proposal.

Balaji Muthukrishnan (~balaji49)

Login to add a new comment.