How time-to-event machine learning prediction reduced sales agent attrition for major telesales company
Case Study
Background
A technology-enabled global business services company partners with ambitious, progressive executives around the world to future-proof their businesses: keeping ahead of the competition and customer expectations.
The company, specialising in customer engagement and improving business performance, tap into sentiment to build the emotional connections that keep customers and staff fanatical about brands.
They are able to help clients reimagine their business by creating exceptional customer engagement, accelerating digital transformation, and delivering actionable CX insights.
The Challenge
Facing high attrition rates of more than 10 per cent among its sales force, the company identified that a solution was needed.
Multiple spreadsheets, insufficient visibility and analytical bandwidth meant managers were caught in ‘analysis paralysis’ – leaving them unable to understand, let alone predict, when individual sales agents were at risk of leaving the business.
Looking for the expertise to deliver the solution, the technology services team approached T-DAB.AI (T-DAB) for support.
The client’s brief was three-fold:
- Develop a machine learning algorithm to predict the number of weeks of remaining tenure, as well as a certainty score to assess that prediction’s accuracy compared to past events.
- Create an easy-to-understand dashboard to present predictions and accompanying information to a non-technical audience.
- Deploy the model to production systems.
The Solution
Drawing on 36 months’ historical data across 290 variables, T-DAB undertook a full machine learning development workflow.
A data audit, followed by cleaning and wrangling, at the start ensured that the data could be utilised effectively before the build commencing. Feature engineering followed, as well as machine learning model experimentation, testing, benchmarking and validation.
Prototyping and experimentation was carried out using standard Python libraries such as Skitlearn and NumPy.
Once the final model was tested, benchmarked and validated, it was pickled before being made ready for production.
The dashboard was developed in Power BI for presentation of the outputs and included interactive graphs, tables, filters and dynamic call out boxes.
The Result
Using the dashboard, managers of sales agents are now able to identify agents at risk of leaving – quickly and easily – before they leave.
The smart tool does this early on – alerting managers up to 10 weeks ahead of the staff member’s potential departure, and as well as providing the number of weeks remaining, the system produces a certainty score.
Managers now have easy access to the supporting information they need – without needing to refer to other sources and spend valuable time trying to map information together.
The model monitors and highlights agents conservatively, ensuring managers have enough warning to take remedial action, such as improving working conditions, while minimising losses resulting from attrition.
About T-DAB.AI
T-DAB.AI is data science and data engineering innovation company. We develop innovative, bespoke machine learning-driven solutions to allow anyone to infuse technology with the spark of predictive intelligence.