About Client
The client is a global event management company based in Chicago with a strong presence in the Americas, Europe, and the Middle East.
Challenge
The client’s existing forecasting model had a few unfavorable issues that impacted its accuracy-
The system lacked automation- The process was more people driven rather than being data driven
- The error percentage between the numbers from the forecast model and actual numbers recorded from their Oracle Financial ERP went above 30% in some cases.
- The model only supported monthly forecasting with no option for weekly forecasting
Our Solution
BizAcuity collected the existing forecast and actual numbers from the existing system- Identified all the attributes and parameters which can impact revenue.
- Conducted meetings with all business units to account for every variable important to each department
- The variables that were identified were both, from the internal as well as external sources
- Data preparation was done from multiple disparate sources
- Cross-checked the correlation between the identified variables and the actual revenue
- Data points which had a good correlation with the company revenue were selected such as pipeline, and group rental nights
- Auto Arima model was tested with the chosen data points as predictor variables
- Holt winter model, which predicts the number based on historical data, trends and seasonality without the need for any predictor variables was used to make predictions for the entire year.
- The Holiday effect on revenue was analyzed using the Generalized Additive Model (GAM) model. BizAcuity had discussion with all business units to validate the findings on the impact
- The pre and post-holiday effect too was looked into using the GAM model
- The holiday impact was incorporated in the revenue forecasting model to further improve accuracy
Outcome
- The new revenue forecasting model was completely automated and based on location.
- A centralized solution was implemented for all business units so that there is a single source of truth across the organization.
- Efficiency of the system doubled. Previously, only 5 months from the 1 year forecast were below the 10% error rate.The number increased to 10 months.
- The model helped the client with a flexible system which also shared weekly forecast numbers
- One of the extra benefits to client from the holiday impact analysis was that the business team could now plan better for the holiday season based on their location
