Business Data Miners

BACKGROUND

Introduce your business and what you do there.

I’m the CEO and co-founder of ezCater, a marketplace for corporate food service providers.

OPPORTUNITY / CHALLENGE

What challenge were you trying to address with Business Data Miners?

We’re highly data-driven, and our internal data analysis team was struggling with capacity constraints. Our team also lacked experience, and we were falling short in terms of complex analysis.

SOLUTION

What was the scope of their involvement?

They created a machine learning engine that predicted outcomes of various business decisions. The engine refined its processes as it received more data. Their team also analyzed search data for us.

What is the team composition?

We worked directly with Larry (President, Business Data Miners).

How did you come to work with Business Data Miners?

I heard about Business Data Miners from a colleague whom I very much respect. I gave them a call, knowing they would give me a worthwhile reference, and they gave me Larry’s contact information. I spoke with Larry for about two hours, and it was clear from the beginning that he is very knowledgeable. I liked his response when I told him about the project, and I felt that I would get significant value for my money if I partnered with him.

How much have you invested with them?

We spent $25,000–$50,000.

What is the status of this engagement?

We worked with them between May–July 2017.

RESULTS & FEEDBACK

What evidence can you share that demonstrates the impact of the engagement?

Their work has completely changed the way we approach data analysis. Using Larry’s insights, we’ve shifted directions in terms of how we use data to inform our business decisions.

How did Business Data Miners perform from a project management standpoint?

Though they were able to meet deadlines, I would have preferred more communication from their end.

What did you find most impressive about them?

Larry and his team were dedicated to providing meaningful analysis for our business. Their work was never theoretical and always centered around decisions that would have immediate effects.

Are there any areas they could improve?

Although it wasn’t necessary for the success of the project, Larry and his team could have done a better job of keeping us updated regarding their work.

Do you have any advice for potential customers?

If you can, have a data analyst or data scientist communicate with their team. With a general business background, I didn’t feel able to make sure that I was getting the most out of the partnership. Larry and his team will work better with people who understand the nuances of their work.

BACKGROUND

Please provide a brief description of your firm.

We are one of the largest telecom carriers in the U.S., behind Verizon.

What is your role and responsibilities? 

I was the director of the Credit Risks department at the time. I was responsible for bad debts, and setting credit rules for our company.

OPPORTUNITY / CHALLENGE

What was your business goal for the project with Business Data Miners?

Our company was using off-the-shelf modeling software to evaluate consumers. As we grew, we realized we needed to build a custom model and we talked with a few vendors. We gave them a common set of data, and ask them to build a predictive model for whether consumers would be a good credit risks for wireless service. Their team built the best model from that competition, so we chose them to evaluate perspective customer financing risk.

SOLUTION

Can you talk more about the model?

It was designed to give the most accurate evaluation based on a credit bureau score and basic demographic information. The model gave the likelihood of a customer being able to pay, and our management decided which scores were acceptable for credit. Sometimes, we decided to maximize sales, and sometimes we decided to minimize bad debt. Business Data Miners did their job by coming up with the pay score accurately.

When was that project completed?

In 2001, I think.

Can you give us a sense of the size of the initiative in dollar terms?

I don’t recall the exact figures, but it had a pretty significant revenue impact. Every transaction they worked on involved a few cents add-on per transaction, and there were more than one million a month being conducted. That adds up to a pretty significant number.

RESULTS & FEEDBACK

Do you have stats or metrics to track improvement?

The model worked, and I believe it was used for three to four years. That longevity alone shows how accurate and predictive the model was. Most models will be updated before four years.

What is unique or special about Business Data Miners compared to other vendors?

They are classic data geeks, in a good way. They have very smart Ph.D.s working for them for years. They have a great ability to pick a very technical topic and learn the modeling techniques quickly. They also use innovative, new trends to help the work. The model must provide results that are useful to help a business make decisions. They get the job done without any frills or fuss.

Looking back on the project, is there anything that you think Business Data Miners can improve upon, or something you would do differently?

I don’t think Larry [Bookman] is a real slick presenter. Larry and his team are good people to have in the room during client meetings, but probably not the right guy to lead the presentation when you’re pitching the material to the client. However, they really know data and understand the technology. This is the most important thing. They help us make correct business decisions.

BACKGROUND

Can you describe your company in brief detail?

The company that Business Data Miners did the work for is an indirect auto finance company.

What was your role?

My role was the chief risk officer.

OPPORTUNITY / CHALLENGE

What business challenge were you trying to address with Blue Fountain Media?

We wanted to have a custom scorecard built to track subprime auto loans. In order to do a custom scorecard, their team had to take historical data on more than 20,000 files, and then analyze that data and come up with predictive analytics that would allow us to use that historical data to predict the performance of new applications.

That company was a finance company just financing cars. We offered our services to automobile dealers and those dealers who had non-prime customers would submit applications to us. We would use predictive analytics that [Dr.] Larry [Bookman, co-founder and president of Business Data Miners] built for us to score a credit bureau, and tell us whether that bureau was a risk that was worth taking or not. If we didn’t want to take the risk, then his model also helped us decide how we should price that risk. The model calculated the rate to charge that customer based on their risk profile in order for us to meet our economic targets.

SOLUTION

What technologies or platforms were used for that data work?

In our own environment, it’s an electronic bureau from Acrofax, TransUnion, or another provider. That bureau would be transmitted to us electronically. It would automatically get fed into this algorithm that they built based on the analytics. That algorithm would spit out a credit score, and that score told us what we needed to know about the risk of that particular applicant.

What was your process for selecting Business Data Miners to work with?

They were referred to us by a company that we were already doing some business with. In fact, Larry was providing a contract role. I don’t know what the exact arrangement was. This was a Canadian company based in Toronto, Ontario, Canada. When we said we wanted to have a custom scorecard built, they put us in touch with Business Data Miners.

I looked at about 20 or 30 different cuts to optimize the data. I think we gave him around 300 different data points. He had to sort through all that. Scorecards typically only have about 10 to 15 different data elements to them. You’re sorting through mounds of data in order to come up with the 10 or 15 elements within that data that actually have a bearing on risk and predicting risk and predicting performance. It’s not an easy process, but the scorecard that he built for us was highly effective. It gave us some real competitive advantage, and indirectly saved us a lot of money by not taking on as many bad loans.

Can you give us a sense of the initiative in a dollar figure?

It was in the low six figures.

RESULTS & FEEDBACK

How did Business Data Miners perform throughout the project?

In terms of the service level and useful dialogue we had with Business Data Miners, it was very professional. The end product did exactly what it was supposed to do and performed very well. That particular score card was only in use for about four years. Then the business was sold to a bank. The bank had its own analytics team. The bank used its own team to build the next evolution of the scorecard. Even with four more years of data, they didn’t really build something that was more predictive.

I think what Business Data Miners built was as predictive as it could be based on where the business was in its current state of development. By the time they built the next evolution, they tripled the size of the business. I think they got a 10 or 15 percent lift in terms of the predictability of the latest scorecard.

What is unique or special about Business Data Miners?

That wasn’t a tendered scenario. I’m missing some comparativeness. Typically, when you have a scorecard company do some work for you, they want clean data. The data they get from companies is typically not that clean. Their team not only worked with us to help, but clean up the data. Most companies would just reject anything that didn’t look like clean data. In this case, they identified a number of different things that seemed to be systemic. We were able to do a very material cleanup of our data rather than just throwing it out. He took the extra step to do that, which makes Business Data Miners stand out.

Documentation is very important in these projects. I know that four years later when the bank and their own people reviewed all that work, they commented that his documentation was outstanding

Is there anything that you think Business Data Miners could improve upon or something you might do differently?

No.