Pexitics Blog

Attrition Management – Explained though Case studies(Case Study 1)

Case Study 1 – Trend forecasting (Month on month) for an ITES major. The forecasts were broken up by department and locations. Also, the conditions under which the forecast would not hold true were also articulated (through a causality model)

Particulars about the data: – The data was from the employee portal where the exit management system was also configured. Every month a summary of the data was created and MIS generated. The system would run a forecast for the overall numbers. However, the accuracy of the forecast was not high and there was no provision to break down the forecast to make it at an operational level (departments, locations)

Definition of the problem- to create forecasts at an operational unit level so that certain decision related to retention and recruitment could be affected. Also, to create causality model linking Attrition trends to salary trends, engagement scores etc. to enable an understanding of when the forecast was likely to fail

The good part was that all the employee data as well as the growth trends for each unit / location was available with the organization

What we delivered: – We did Time series forecasting using ARIMA models, Error Trend Seasonality models and Moving Average models for each department level and location level and added it backwards to get the overall forecast for the company.

A linear regression model was created to define the exact causality relationship between engagement scores, salary against benchmarks, average time in system and type of location. This gave a clear understanding about when the variables start changing and hence, the forecast trends will start changing.

Insight implementation: – the forecast vs the actual line would get traced every month at every operational level. A greater than 10% difference would get highlighted for discussion

Overall (company level) trend forecasting did not give enough details to decide operational tactics and policies.

Insight implementation: – the forecast vs the actual line would get traced every month at every operational level. A greater than 10% difference would get highlighted for discussion

Predict when Forecast prediction will fail  – Causality model

Author: Subhashini Tripathi

Spread the love

So, what do you think ?