Ramgopal Prajapat:

Learnings and Views

SIP Investors - who will terminate in next 30 days

By: Ram on Jul 16, 2022

There around 40+ Mutual Fund Houses in India and they have over 2,500 Mutual Fund schemes.  There is an increased interest and investments toward mutual funds from the retail customers in the recent past.  

The investment to a mutual fund scheme is done either as Lump sum or SIPs (invest regularly - Systematic Investment Plan).  Number of SIP accounts is all time at 5.54 crore and the monthly investment toward SIPs  is  around 12276 crore [Ref].

Investment via SIPs is great way to accumulate wealth over a period and most investors are aware of the power of SIP. But not all customers are disciplined and committed toward long term goal.

One of the key challenges in the mutual fund industry is to keep customers active on the SIP portfolios.

Customers can terminate a SIP for various reasons and some of the high-level dimensions could be

  • Fund Performance – Performance of Mutual Fund Scheme
  • Switch - Nudges from a Financial Advisor to switch the investment
  • Customer Service and experience with the Mutual Fund Provider
  • Financial Needs - Financial situation has changed and not able to invest further or there is need to withdraw fund from the SIP portfolio
  • Market Volatility - Share market has a high level of uncertainty and customer’s risk appetite has changed
  • Social Media Discussions – customers are on social media platforms such as Facebook or LinkedIn and discussions & information on these platforms may influence investors to relook at the investments.

And there could be combinations of these reasons. How long SIP investor is expected to remain invested or when s/he will terminate his SIP?  This is pertinent to question for Mutual Fund Portfolio Manager.

Artificial intelligence (AI) and machine learning (ML) frameworks can be used to leveraged to answer some of these questions and manage SIP Portfolio.

Case Study

One of the leading Mutual Fund companies had a high level of SIP termination. The leadership team at the company had a vision to manage the SIP Book proactively. They wanted to identify the customers who have higher chances of terminating SIPs next months. And develop appropriate communication plan to reach out these customers.


Problem Statement

Develop a machine learning model that identify the customers with higher chances of terminating at least one SIP next month.


  • Customer Profile
    • Gender
    • Age
    • Location - Top 30 City etc
    • First Investment Date with Mutual Fund Company
    • Channel of acquisition
  • Investment Mix
    • Investment by Mutual Fund Schemes (e.g., Debt, Equality etc)
    • Number of Mutual Fund Schemes
    • Number of SIP terminations in the latest 1/3/6 months
    • New SIPs in the latest 1/3/6 months
  • Fund Performance
    • Fund Performance for the SIPs in the latest 1/3/6 months
    • Relative Fund Performance – compare with Nifty/Sensex
  • News and Social Media Sentiment for the company – this will be level multiplier and not useful directly for the investor level model
  • CRM Interactions
    • Contacts via email or Call Centre

Data Model


  • Prepare features from base attributes e.g., first investment date to days/months since first investments



Description automatically generated


  • Aggregate transactions into different groups and levels
    • Time Basis e.g.  Number of investments/redemptions in the last 1/3/6 months
    • Category Basis e.g., Number of investments/redemptions in Debt/Equity etc. mutual funds

Fund Performance – MF Scheme Performance

  • Based on published NAV, prepare fund performance in the last 1 week/month etc before SIP Termination date


  • Customer interactions data aggregation - # of calls or logins in the last 1 week, 1 month prior to SIP termination


Once each of the fundamental dataset is read, we need to combine or merge these datasets to prepare the final dataset with 1 investor (or customer) has one row. And the dataset will have all the relevant investor characteristics and label variable (whether SIP terminated in the 30 days or not)


Explanatory Data Analysis (EDA)

One the most important stage of predictive model development. In this phase, we build understanding of investor behaviours – how customers who terminated SIPs are different or behaving differently. Additionally, we will validate of data seems logically correct. Too good is always suspect. If one or two features help very strongly segregating SIP Termination and continuing investments, we will review the data.


Model Development

  • Split the dataset to Training (e.g., 70%) and Testing (e.g., 30%) randomly
  • Use any or multiple of the Machine Learning techniques e.g., Decision Tree, Logistic Regression, Random Forest, Boosted Trees, Support Vector Machine (SVM) or Neural Net
  • Parameter Optimization to improve the final performance of each of the competing models
  • Selecting one model or ensemble set of models to maximising the performance

I will always recommend using Decision Tree first and prepare visual tree. The tree will help in visualizing behavioural groups.


Model Validation

  • In-time Validation using left out sample (Testing data) and compare performance of the model between training and testing samples.
  • Higher drop in performance between training and test sample indicate overfitting of the model
  • Out of Time Validation – we may have used SIP termination of certain months - for example between Mar to Jun for preparing development data. We can pull customers data for Jul month and predict their SIP termination and compare with actual SIP status. If performance is great, we have a confidence to deploy the model (put into regular use)

Action Strategy


We a well performing Machine Learning Model to predict customers who going to terminate SIPs in the next 30 days. Now, time to action on.  For preparing action strategy there are a few important dimensions to consider

  • Customer Value – Not all customers are of same value so our focus will be more toward higher value customers (may be size of investment corpus)
  • Contact Channel – how are we doing to connect with customers and who should be contacted with broker, relationship manager, calls or emails
  • Communication and Messaging – we need to preparing communication plan so that our focus on reducing SIP terminations deliver business impact.

Leave a comment