agsandrew - Fotolia

From HVAC repair to prescriptive analytics software development

Ontegrity turns knowledge gained through years of onsite HVAC maintenance into a prescriptive analytics system built in-house that prescribes repairs before failures occur.

This is the third in a continuing series of stories previewing sessions of importance to cloud application developers at the Cloud Expo conference that takes place June 7 - 9 at the Jacob K. Javits Center in New York City.

Jonathan Quint is CEO at Ontegrity, a Framingham, Mass., company that provides onsite maintenance services for commercial and industrial HVAC systems; about as far away from being a software company as is possible. Yet Ontegrity has become a leader in applying predictive and prescriptive analytics to the world of equipment field service. His session, Harness IoT and Predictive Analytics for Efficient and Reliable Operations, is scheduled for Thursday, June 9 at 3:40 p.m. In it, he discusses how Ontegrity built a software-development organization from scratch, creating prescriptive analytics software that enables the company to provide cost-effective, before-it-fails field maintenance and repair.

Ontegrity repairs HVAC systems. That is not even close to being a software company. How did Ontegrity get on this path?

Jonathan QuintJonathan Quint

Jonathan Quint: We are in the business of providing preventative onsite maintenance services to businesses that have mission-critical HVAC systems. Whether they are in telecommunications, food service, or something else, they can't allow those systems to go down. We noticed that sometimes we'd go out for a scheduled preventative maintenance call only to find that a unit didn't need service. Other times a unit would fail before a scheduled service. The question was how to avoid unnecessary calls yet address issues before a failure could occur.

This is essentially a modern version of an expert system. Where did the expertise come from?

Quint: Our field service personnel provided years of subject matter expertise. By using their expertise combined with a group of young number-crunching algorithm geniuses, we started to install sensors in different places to measure temperature, pressure, and other factors, looking for patterns and correlation. It worked, but we are continually refining the algorithms. There are sensors in each physical location connected to an edge device that talks to our servers. Everything is accessed through our online system.

You can tell when an HVAC is headed for failure, even though it appears to be running fine?

Developing machine learning algorithms is a challenge. You don't just write them and have them work the first time out.
Jonathan QuintCEO, Ontegrity

Quint: Many companies have sophisticated real-time monitoring systems with dashboards that send an alert when a fault occurs, not before. With our algorithms, we detect system degradation and measure that against predetermined thresholds. It can tell us what needs to be done to restore optimal performance. Often, the customer has no idea anything is wrong.

What is the jump that took Ontegrity from predictive to prescriptive analytics?

Quint: Our system is being trained. We started with it telling us go to a location and do maintenance, but now we are in the final steps of validating where it tells us to go to a location, perform the following work, and be sure to bring these parts with you.

What is the IT infrastructure for this prescriptive analytics software?

Quint: This is a private-cloud infrastructure that runs on our own servers. We use ColoSpace, a local infrastructure provider [based in Rockland, Mass.] that also offers outsourced data center and managed services throughout New England. We've used them for several years.

Is the SME the key for other companies considering similar development?

Quint: The key lesson is having a SME working with our software developers. Having [a] mix of the right people helped.

Do you consider Ontegrity to now be a software company?

Quint: We are in transition. We will still have the field service business. We did not have software or application-development expertise, so perseverance is important. This process of developing machine learning algorithms is a challenge. You don't just write algorithms and have them work the first time out. It is a process of refinement.

What advice do you have for other companies considering a similar transition?

Quint: We already had the industry expertise, but we were not experts at developing software. We found that recruiting for talent was a significant challenge. There is a lot of competition for young data analysts and demand exceeds supply. We put together a great team to solve a specific need and then created a solution for it. This is a smart way to grow a business.

Next Steps

Does your dashboard know how much you are spending on IT ops?

Public vs. private infrastructure: Which one costs the least?

Machine learning is growing in popularity. Do you know why?

Dig Deeper on Cloud application development