Insurance broker automates staffing model process with forecasting model
This company refreshed their outdated staffing model process with a new forecasting model that would predict the ideal number and type of support staff to assign to each account per year.
Client background
This company is a privately owned insurance broker who provides insurance, employee benefits and risk management support for customers across a variety of industries.
The business challenge
The company’s objective was to forecast staffing needs based on data from prior years of business. Their current staffing model process was created nearly a decade ago and needed to be refreshed with new time tracking data to better support the company’s current needs. The company also wanted to add additional factors to the staffing analysis, including number of customers and revenue per customer, to update guidance on the staffing levels required for a given account.
The company engaged Baker Tilly to build an AI-driven foundation around designing a systematic and automated method to assign admin and support staff to specific business leads and their respective client accounts.
Strategy and solution
Baker Tilly began by reviewing the company’s current state staffing model, a regression equation created with their personal company data, before identifying and analyzing key attributes that met business metrics and explained the company’s current business process. We then worked with the company to extract new time tracking data from their current ERP system and collected each department's files containing client account profile information.
This data was then tested in various classic and machine learning (ML) forecasting algorithms and a variety of modeling assumptions. The team then validated the robustness of the models using a test and training set. While the company was only looking for the number of people to staff each account per year, our modeling approach provided more granular information surrounding ideal staffing levels based on location, team, position level and even time of day.
As a result, we presented the company with a model that would predict the ideal number and type of support staff to have assigned to a particular project. Having the correct number of staff on any given project resulted in a right-sized workforce and mitigated the risk of over/understaffing.