How Machine Learning Helped a Non-Profit Use Less Money to Create More Jobs
Data is an extremely powerful tool for businesses, especially if harnessed in a smart, efficient and predictive way. Organizations like JobsOhio, a non-profit that helps companies identify and pursue growth opportunities in Ohio, must parse hundreds of thousands of data points daily to be successful – a process that can prove time-consuming, complicated and a hindrance to business processes and objectives if using outdated data collection methods.
But modern solutions like machine learning, which leverages pattern recognition tools to identify trends in immense amounts of data, could help JobsOhio make valuable, informed decisions in less time and with less money.
While JobsOhio was no stranger to the power of data, they were unsure how they could best use it to prioritize projects, maximize resources, improve close rates and have a larger impact on the Ohio economy. And as a non-profit, they also wanted to do all this with limited funds.
JobsOhio’s funding model is unlike the funding model for any economic development organization in the country in that absolutely no tax dollars or other public dollars are used to support it. Despite being wholly self-funded, JobsOhio has been very effective at generating new jobs in Ohio. For example, in 2019, JobsOhio executed 306 projects generating $1.2B in new payroll, creating 22,770 new jobs.
As effective as they are with their limited funds, they still wanted to do more for the state of Ohio. That meant prioritizing the employers with whom they would negotiate to create, retain or expand operations in Ohio – a very complex challenge given over 32.5 million employers in the United States (250,000 of those in Ohio) spread across over 1,000 industry specialties. JobsOhio realized that the complexity of this problem was a good match for machine learning, but they needed the help of a partner.
Aware of our deep data expertise and proven success with machine learning, JobsOhio engaged Centric Consulting to explore predictive models’ development to help with this prioritization.
Enter Centric: Showcasing the Potential of Data with Proofs-of-Concept
At Centric, we want our clients to understand fully why we choose the methods we do and why they work. As we partnered with JobsOhio on this project, we wanted to give them a taste of what machine learning and data prediction could do for them before making large changes to their processes. Proofs-of-concept come in handy here, as they use real data to truly illustrate the potential value data modeling can bring to business without considerable investment or disruption.
“When we develop these proofs-of-concept, we say ‘Let us show you what is possible’ before we try to influence your operations,” says Centric Machine Learning Specialist Tanya Kannon. “It helps the client to understand fully how it works to be on board before they make the investment.”
We determined a wide variety of data sources to feed our models, including everything from news reports on new commercial construction projects to legal and government databases highlighting bankruptcies. Based on JobOhio’s business needs and evaluating five-years’ worth of available data, we worked together to identify three prospective models to investigate as a proofs-of-concept (POC).
- Site Capture Model: This focuses on prioritizing companies already in the JobsOhio pipeline by identifying those that would most likely accept a development offer and successfully expand in Ohio.
- Site Retention Model: This discovered candidates for job retention projects by pinpointing individual sites of companies already in Ohio that were likely to have a significant job reduction.
- Site Expansion Model: This identified candidates for job growth projects by finding individual sites of companies already in Ohio that had significant job growth potential.
We used Azure Machine Learning Services to build these models, and we helped establish data dashboards that would allow the JobsOhio team to track and visualize their work. With the current economic climate, finding the right projects and keeping focus on the potentially successful projects allows JobsOhio to drive job creation and investments better.
The Results: Overcoming Challenges Through Close Partnerships
There was an important learning experience for JobsOhio in this effort. Machine learning projects are more like Research and Development (R&D) efforts than software development — all R&D efforts come with some risk that the desired product is infeasible. By testing three different models, JobsOhio determined that some would work for them, while others may not.
In this case, while we were able to successfully source the data needed to develop strong predictive models for Site Capture and Site Retention models, we were unable to collect enough data to feed the site expansion model effectively. This could certainly change as more data becomes available in the future.
Despite this challenge, the close partnership we formed with JobsOhio during this project led to its ultimate success. We were able to develop two robust models — one that addresses prioritizing efforts and one that provides insight into companies that are likely to reduce their footprint in Ohio. And perhaps more importantly, we worked together to give JobsOhio profound, valuable insights into their business.
“The thing that struck me about working with JobsOhio was how invested they were in this project. They truly involved themselves in the nitty-gritty of it, to better understand how the predictive model works so they can sell it to the rest of the business,” says Tanya Kannon. “Now, they are walking away with a huge insight into their business, of who they are, and why they do what they do.”