TL;DR Collaborated with a global project development and construction group to improve their lead list by incorporating various open data sources and customer expertise. Resulting model increased B2B sales success rate and led to development of user-friendly web platform, later spun out into its own company.
Excerpt from analysis
This case demonstrates that effective solutions don’t always require cutting-edge machine learning. Sometimes, a simpler approach combined with deep domain knowledge can yield the best results. It’s a reminder that in problem-solving, the shiniest solution isn’t always the most appropriate or effective.
When I started this project to find better leads for a client in commercial real estate, I had to pick the right data. The goal was to predict which companies might want to relocate soon. Here’s what I did.
Understanding moving patterns
I thought about what makes tenants want to relocate:
- Company changes: Fast-growing companies often need to upsize, while downsizing companies might look for smaller spaces to cut costs.
- Industry trends: Some industries, like tech and finance, tend to relocate more often than others, like education or manufacturing.
- Building quality: Older buildings or those with poor space utilization might push tenants to seek newer or more efficient spaces.
- Location quality: Good access to transportation can be a big factor in tenant retention.
- Lease vs. own: Owner-occupied properties are less likely to see tenant turnover.
- Recent moves: Companies that just moved are less likely to relocate again soon.
Picking the right data
I looked for information that tells us a company’s office needs:
- Company details: I used data like industry type, number of employees, and financial health. These help see if a company is growing or shrinking.
- Property information: I collected data on the office space size, building age, and whether it’s owner-occupied or a leased property. This shows if the current space fits the company’s needs.
- Location factors: I looked at how long the company has been at their address and the property’s proximity to commute possibilities like public transportation and highways.
By combining all this information, I made an indicator model, which I turned into “relocation probabilities”. Tenants with high scores became the best leads for the commercial real estate group. This method improved the cold calling success rate and helped the client find more potential tenants.
Presenting the data
The results were presented to the client as a data grid that ranked each company by its relocation probability and showed the supporting data in four groups:
- Building: area in square metres, construction year, number of other companies in the building, whether the company owns the building, and time at the address.
- People: current employees plus the change over the past 6 months, 1 year, and 3 years.
- Financial state: equity, liquidity ratio, and solidity.
- Commute possibilities: distance to the nearest train station, subway, bus stop, and highway.
Project outcome
The success of this project led to the development of a user-friendly web platform. This platform was later spun out into its own company named Tembi.