Putting Transparency Into The Model Building Process Helps Predict Construction Costs
The client is a multinational real estate investment company developing about 100 new buildings annually. They serve roughly 5,000 customers across a wide range of industries in 19 different countries.
Organizational leaders wanted to reinvent their development cost analysis with a modeling system to estimate building cost at an individual project level. There were over 150 characteristics influencing costs, with 17 classified as high impact (accounting for two thirds of the building costs).
We collected data from a variety of sources and formats, cleaning and aggregating the information to prepare it for advanced analysis. A total of 318 variables were analyzed over the course of the project, including categories such as location, building dimensions, and labor rates by market. The approach involved exploratory data analysis (EDA) to summarize the data set characteristics with visualizations. Using an iterative approach, we held regular updates with the client to communicate the selection of building cost predictors based on findings during the EDA phase. We also met with each of the regions to gather information on current-state reporting and important metrics to their business. The software selected was Microsoft Power BI due to its inexpensive licenses and compatibility with our client’s needs.
The end result included linear and non-linear predictive models to assist in the bidding process by showing where construction bids deviate from expectations. Our predictive analytics framework offered transparency into the model building process to help the client best utilize the predicted costs during the construction bid process.