Know When to Pivot


Our CTO, Paolo Samontanez, was interviewed by the Department of Science and Technology (DOST) for their first book series “Science for the People.” The book showcases the innovations of science and technology, including success stories, and best practices of small and medium enterprises, as well as startup companies.


Read the full transcript of the interview below.



Unplanned downtime can spell millions in dollars of losses for a big corporation. OneWatt, a Filipino startup headquartered and backed by an investor in the Netherlands, makes sensors to listen for industrial machines that are about to conk out. Making sure business goes on without a hitch, their engineers get in, install, punch some numbers and get out. But it wasn't always this way. Its CTO, Paolo Samontanez, an electrical communications engineer, gives us practical advice about knowledge when to change direction- and what it would take to come back.


Paint us a picture of a factory you monitor.

Typically, at factories, there are hundreds of motors in a production line - from small motors to ones that are the size of people. If one motor fails, then the whole process fails, and that means they start losing their productivity. It’s essentially lost revenue for the facility. That was kind of our inspiration - that’s why we decided to literally listen to the problems of motors with our Embedded Acoustic Recognition Sensors, or E.A.R.S for short.


Would it be something like a virtual staff?

Yes, so essentially it’s like Alexa. It’s a cylindrical device- technically it’s a microphone array- and then the user puts it near the motor they want to monitor and the sensor collects the sounds from that monitor. Being a microphone array allows us to specialize our recordings in a noisy, industrial environment. To do additional signal processing, like beamforming and noise reduction.


What’s beamforming?

If you can imagine a factory: motors are side by side, right? If we put, for example, the E.A.R.S near one motor, the question is: how do we know we are listening to the right motor that we are aiming at? Beamforming allows us to direct - essentially selective hearing- to a specific direction. We reduce the effects of the other motors and that allows us to collect data from the motor we are interested in.


How do you set that up?

When I go to the facility, I stick the E.A.R.S to the wall and, based on the position, I do some calibration. I have to analyze what are the peak frequencies coming out from the asset. From there, I measure how I will steer the hearing of the sensor in that specific direction and that specific frequency. Like a radar, essentially. It’s completely non-intrusive. The installation is completed in two hours and we do not ask the factory to shut down their motors.


Are other service providers on the market usually intrusive?

Yes! Either they need to open up the motor or cut into the wires of the motor to insert the sensor.


Once you’re done with the installation, how do you gather the information?

Most of the AI processing is done [online] on the cloud. But for some cases, we attach a hard disk, and then we can collect data and do post-processing.


But is it ideal to actually do it online?

Yes! To have real-time monitoring of the motors. The vanilla version of the product - we give them an app they can see the notifications of events that they should be alerted to.


What do you miss about the scientific community back home?

I think we are much better when it comes to the science and technology part. But Europeans are better at the business part.


Didn’t you and your co-founders start OneWatt because you heard a lecture in college about the inefficient power grid in the country?

During that time - 2016 - there was a discussion about how expensive electricity is in the Philippines. When our power grid produces extra electricity, it just goes into thin air. It’s wasted electricity, essentially. We tried to use AI to predict the prices of electricity in the spot market and balance out supply and demand. We try to store the electricity while the prices are cheap. And then when the price rises, we use the stored electricity.


So, at first, you were prototyping… a battery?

Yes! Actually, the very first prototype, we entered it in AngelHack, it’s a competition. We were part of the top two so we were battling against a juicer. In the end, the judges chose the juicer over us and, actually until now, we’re still bitter about that. (Laughs)


Must have been some juicer

The problem with the battery, it needed a lot of capital. So it’s a very risky project, if the people in the industry don’t trust us, we wouldn’t have clients.


So how did you pivot to the sensors?

What happened was, we needed to have a way to monitor if the client is ramping up their production or not. The revenue that we get is based on how much savings they get [from the use of the battery in their factories]. If they are getting the savings but they ramp up their production, it would appear as if they didn’t have savings, essentially, they could take advantage of the system. So, we put in a sensor to listen [When we moved away from the battery business] we decided to give the sensor more functionality and focus on that specific part.


Would you ever return to the original idea?

Actually, if you look at the goal of the company, our goal is to increase the overall efficiency of industrial facilities. We just decided that we should focus on the listening part because, for us, once we establish the trust, it’s easier to sell the bigger products.



OneWatt would like to express its deepest gratitude to the Department of Science and Technology for featuring us.


https://www.facebook.com/DOSTph

https://www.facebook.com/nrcpresearchpod

https://www.linkedin.com/company/department-of-science-technology/