Resources

Company resources:

• OPower at Oracle Energy and Water used by OPower

OPower Pilots and Programs:

• Southern Maryland Electric Cooperative: SMECO sees 5X the rate of adoption for their Home Energy Improvement program
• Pecan Street data research and product testing in the energy, water, transportation and agriculture sectors
• CPS Energy (San Antonio): 2021 STEP Annual Evaluation Report
• Arizona Public Service utility: APS increases customer satisfaction by 13% with Behavioral Load Shaping
• Baltimore Gas & Electric Reshapes Peak Pricing Programs

Studies:

• The Customer Action Pathway to National Decarbonization report by the Brattle Group

Career Resource:

• Clean Energy Leadership Institute

Guests

Climate Rising Host: Professor Mike Toffel, Faculty Chair, Business & Environment Initiative

Paul McDonald, Senior Director, Opower Product Strategy & Marketing at Oracle Energy and Water

Transcript

Editor’s Note: The following was prepared by a machine algorithm, and may not perfectly reflect the audio file of the interview.

Mike Toffel:
Paul, thanks so much for joining us here on Climate Rising.

Paul McDonald:
Thanks for having me, Mike. It's good to be here.

Mike Toffel:
So, let's start with an introduction. What's your role at Opower?

Paul McDonald:
Sure. My role at Opower is to help energy providers, utility companies, to reduce their emissions and generate value for their organizations by influencing their customers to take a wide variety of actions. And when I say their customers, I mean energy consumers like you and me. To get specific, I lead product strategy and marketing for the Opower organization at Oracle Energy and Water. I get to lead an incredible team of people that handle things like market research and investment planning analysis. We do incubation of and commercializing new OPower products and services. And we get to help our clients tell the stories of what they're doing with power.

And that last bit is honestly, I think that's the most fun and fulfilling part of the work I do. We get to help our clients just show the whole industry what's possible by putting artificial intelligence and behavioral science to work in the experience that they provide their customers. We get to spend time imagining and planning for what we can do and what we should do next just to make a bigger impact for our clients and for their customers and for the planet.

Mike Toffel:
Terrific. What did you do before Opower and how did that lead you to Opower?

Paul McDonald:
I spent the first chapter of my career consulting for federal agencies on their financial operations. I found Opower back in 2012. I went looking for clean tech companies where I could make a real impact. I had been doing some reading about the concept of the Smart Grid that just completely fascinated me and I was looking for something that I could just connect with on a more personal level. The skills I had at the time were technical skills and project management skills, and when I initially applied to work for Opower, it was for a pretty technical position and I crashed and burned.

I applied again for a project management position and that was the kind of role that my skills were most aligned to. I made it to an in-person round of interviews. and then the coolest thing happened a few days later, a package arrived on my doorstep and inside it was a bottle of champagne and a hand signed note that said, welcome to Opower, this is your ticket to the party we're going to throw when one of two things happens. Either we go public or we reduce U.S. energy consumption by 1%.

And I thought that was just an incredible way to be welcomed to the company. And it turned out to be just one of many, many, many stories about how Opower as an organization has shown me that we have and do just make people and culture a big priority. So, since then I've done a little bit of everything at Opower. I've done project management, I've run parts of our software as a service technical operations group, solution architecture, product management, and now strategy. But the through line, I think, for all of that, has been working closely with our clients.

Mike Toffel:
That's a really interesting decade plus time you spent there. I've never heard of a welcome package pitched in that way. It's so mission oriented. It's really interesting.

Paul McDonald:
It really set the tone.

Mike Toffel:
Yeah. Let's talk about Opower as an organization and what its mission is, how it got started. So, you've mentioned a little bit about that, but what's the elevator pitch of Opower?

Paul McDonald:
The elevator pitch of Opower is influencing utility customer action at scale. The interesting part about it is one of the things that Opower does as influencing an action is helping energy consumers use less of the products of the company that hires us to do that for them.

Mike Toffel:
It's like inverse marketing. Here's why you should use less of our product.

Paul McDonald:
Exactly. Yeah. Another way to say it is Opower uses artificial intelligence and behavioral and science within the utility customer experience in order to influence people to take actions that are valuable for them, valuable for the utility and valuable for the planet.

Mike Toffel:
So, before we get into how you do this, using behavioral science and artificial intelligence, let's just make sure everyone understands why. Why would a utility hire someone to convince their consumers to use less of their products after all they make money when people use more natural gas or more electricity?

Paul McDonald:
The simplest answer I think is twofold. One is in a variety of places around the world, but predominantly in North America, utility regulators incentivize utilities to provide this service to their customers, helping them use energy in smart, effective ways. The practice is called demand side management and it's been a norm in the utility industry for decades now. The way it works is utilities get made whole for that reduced consumption and revenue by their regulator. And in plenty of places, regulators actively create financial carrots and sticks for running cost-effective demand side management programs. The second reason is all kinds of utilities, whether they're investor owned or municipal or even retail electricity suppliers like we see in places like Texas or other parts of the world.

These organizations have just as much of an incentive as any other service provider that you and I work with, Mike, to just provide valuable service in a great experience to their customers. It helps with customer satisfaction, it reduces service costs and it grows adoption of value add products and services that these organizations want to offer. And providing that kind of service, from my experience, it's part of the utility industry's DNA. There's a real public service mindset that utilities bring to their work.

Mike Toffel:
So, you don't see this in other utilities, you don't see this in telephone or you don't see this in cable. And I imagine this is partly because of the enormous fixed costs associated with having to add more generating capacity, for example, online. I've always thought that what they're trying to do is maximize the utilization of their existing capital stock while trying to slow the pace at which they have to make new investments, which requires regulatory permitting and new transportation lines and so on. And so their view of maximizing profits is actually to try and operate near the hundred percent utilization frontier rather than having to make investments because they have a duty to serve, which these new investments will at least at first really not get much utilization in which case it's costlier to operate. Is that what's going on here in at least the electricity sector?

Paul McDonald:
That's pretty close. The point that you mentioned about needing to make the best possible use of these capital investments is spot on. So, demand side management as a practice got its roots in providing as much economic value as possible to energy consumers. The whole name of the game is the utility canon should only run a demand side management program that reduces energy consumption

However, in recent years, the growing focus on the climate crisis and the need to reduce emissions in the energy sector has started to change the reasons why regulators ask utilities to run these programs and even how they run them, measure, verify them, report out on how they're doing.

Mike Toffel:
Got it. And this is where you're saying the policy makers are stepping in not only to try and reduce the need for more infrastructure, but also to facilitate carbon reduction goals that the public and government have set for themselves as well?

Paul McDonald:
Exactly. 75% of energy consumers in the US are now served by a utility with a net-zero commitment on the books.

Mike Toffel:
And those net-zero commitments sometimes are about pipeline leaks, for example, and not really considering the end consumer's use of the power source, which would be a scope three for them. And they're saying, well, we're going to get it to our customers in a way that minimizes the transportation and generation's piece. But anything that happens with the consumer, if they need electricity, it's up to them to buy green power if that's what they want to do. In some cases.

Paul McDonald:
You're right. For a very long time, the prevailing worldview was the way to decarbonize utilities is by cleaning up energy supply, whether that's pipeline leaks or swapping fossil fueled energy supply with non emitting resources. For a gas utility that could be renewable natural gas or hydrogen for electric utility. Everyone imagines wind farms and solar arrays. But if you step back and think about that for a second, that is massive critical infrastructure on a global scale that we need every day to live our lives and operate our economies. It's concrete and steel and complex machinery and rebuilding all that is going to take decades of careful planning and trillions of dollars in capital investment that has to happen.

But in the meantime, utilities need big, fast, affordable pathways for reducing their emissions in order to make progress towards their goals that they're setting at the behest of their leadership, their investors, their largest customers and the governments of the states in which they operate. So, yeah, they're looking to customer action as that pathway for big, fast, affordable emissions' reduction. There was actually some recent research we conducted with the Brattle group that showed that by 2040 utility customer action can reduce nearly two times more emissions than rebuilding energy supply infrastructure alone.

Mike Toffel:
The Brattle study refers to consumer behavior changes and it compares that to the supply side. Of course, both really need to happen, and I know that some of the products and services you're working on actually touch on many of these things. They touch on how do we improve energy efficiency? How do we shift electricity demand from periods where the grid is dirtier to when it's cleaner? And all this is happening in a moment where there's an electrify everything movement going on, whether it's transportation, people moving to EVs or people shifting their use of natural gas and, for example, in homes to electrification through stoves and through heating systems.

So, all of the actions we're talking about in efficiency might well be overwhelmed by the increased demand for electricity due to this electrify everything movement. And I know you're working sort of in all of this space, so let's talk through some of the products and services you offer. Maybe let's start with, I think the area where you've been working longest is on the energy efficiency side with your home energy reports which you deal with utilities and you sort of help them figure out how do we communicate to customers to get them to adopt more efficient behaviors. So, why don't you walk us through that history and any other products in that space.

Paul McDonald:
Happy to. The Opower's original invention was the Home Energy Report. This started as a piece of paper that was mailed monthly to energy consumers showing them how they're using energy, making personalized recommendations for how they can use less of it and motivating them to use less of it primarily with behavioral science techniques, the workhorse of which is one called Descriptive Norms. It compares a customer's energy use to similar homes around that customer. And over time that product has evolved dramatically. When I was joining the company, we started using digital versions of it, personalized email has turned out to be a really effective communication channel. Since then, we've layered on a digital interventions like high bill alerts if a customer's trending toward receiving a high bill, and that can be predicted with machine learning that determines a customer's sensitivity to changes in weather patterns. Or it can be predicted with analytics that look at the trends of a customer's metered usage throughout the course of the bill period.

When customers get those high bill alerts, we use another behavioral science technique called Agency to show customers there's something they can do about it in order to control it and offer them, again, personalized recommendations on how they can reduce that upcoming bill. Another tool that we've added to the kit for delivering behavioral energy efficiency is a product called a weekly energy update. And Mike, we originally were very cautious about communicating with energy consumers on a weekly basis because one of our design principles is customers do not care. There's a lot of people that have much more interesting and important things to do than even think about their energy use. So, we originally offered that kind of intervention as an opt-in thing for customers who were interested, they could say yes, I do want to see my week-over-week comparison of how I'm using energy and how that breaks down by hour and by appliance and what I can do about each specific appliance.

But in recent years we have tested and found for ourselves that when you enroll massive numbers of customers into receiving that weekly communication, they open it, they love it, and they save more energy as a result of receiving it. So, there's a variety of tools that Opower uses now to run a specific kind of demand side management program for utilities called behavioral energy efficiency. To date, our clients have helped their customers save 35 terawatt hours of energy, which is the equivalent of 17.2 million metric tons of carbon emissions. And to make that real just imagine a pound of coal, you could probably hold that in your hand. Now imagine 19 billion pounds of coal that did not have to get burned to produce energy. That's how many emissions that we've abated to date.

Mike Toffel:
Let me ask you two questions about these home energy reports based on what you were just saying. So, one is you mentioned that you have the ability to give, say, weekly alerts. I'm not sure if people understand how often their electric bill is being read by the utility. And it used to be that trucks would drive by quarterly or monthly to take readings, but it sounds to me like you're operating in areas that have much more frequent readings, maybe even real time readings. Is that right? Is that your sweet spot in places that have these smart meters that are reading instantaneous and feeding that back to the utilities?

Paul McDonald:
That is where we are able to make the most of the technology that we offer utilities. There's even places in the world that only read the meter once a year and there's only so much you can do with that data. Most of the clients that we serve at least read the meter once every month or once every other month. But a significant portion of our customers also have smart meters. And these meters record energy consumption typically aggregated up to 15 minute or hourly intervals. And over the past 10 years, that has created an explosion of data to be managed in the utility industry. And that's one of the things that Opower does.

We import about a billion of these meter reads every single day. We're managing about 4 trillion of them on behalf of our clients. And what that enables us to do is train machine learning models and run predictive analytics that do two things. It shows all types of different energy consumers the thing that they might need to see next to be motivated to take some action and the actual recommendation for an action that they should take. All of that starts with all of this data being read right off the meter.

Mike Toffel:
Yeah. So, you mentioned earlier about recommendations of what they can do. I think the simple thing to do in the summer is, if you have air conditioning you turn up the thermostat so you don't cool it quite as much and in the winter you say, well, I'm not going to be quite as warm, I'm going to turn it down. But it sounds to me like you're a lot more precise than that and you may even be recommending replacement of equipment or maintenance of equipment. So, what are the types of recommendations that you're making and how does your system getting, say 15-minute increments of electric use, how do you convert that into these types of recommendations?

Paul McDonald:
There's a standard library of hundreds of recommendations in the same way that Netflix has an enormous library of content that you might want to watch, and part of the magic of that service provider is determining what's going to be most relevant for you. In a nutshell, that's what Opower does using data, is collecting meter data, digital interacting data, merging third party data in order to continually predict what each customer is going to need to see next and do next and then offer them that piece of content.

So, it could be things like changing your thermostat, it could be purchasing a smart thermostat and enrolling it in an automation program. It could be adopting a new pricing plan that will reward a customer for shifting energy and saving. It could be upgrading an appliance that's using an unusual amount of energy in the home. The variety here is pretty large and our clients use that to very great effect. I'd love to tell you one story about an East Coast client that just recently did that.

Mike Toffel:
Great.

Paul McDonald:
So, Southern Maryland Electric Cooperative there's a long-standing client of Opower’s and they've been running behavioral efficiency programs. The state of Maryland is constantly looking for the next thing for utilities to do in order to change and upgrade and deliver more impactful to man side management programs. So, SMECO and Opower implemented an experience within their home energy report program that showed customers, here's how much you're spending on these specific appliances and here's what you can do about it. You can take advantage of a SMECO program that will send a technician into your home, evaluate what's specifically going wrong and actually do a tune-up or recommend a device appliance purchase for you.

Now the way we did this is we collected SMECO smart meter data, ran it through machine learning models that predicted each customer's appliance level usage and then automatically selected the right recommendation to offer each customer based on their appliance level usage profile. And we ran those experiences as campaigns over the course of the year. What we found is that customers who received that kind of experience from SMECO adopted these home upgrade projects at a rate five times higher than a statistically equivalent group of customers that did not get that experience. And we could say with certainty, by using data and using behavioral science in that way, we caused five times more SMECO customers to make upgrades to their homes.

Mike Toffel:
That's certainly a profound impact. The piece that sounds like magic that I'd love you to just unpack a little bit is if you're getting only one signal of how much electricity is being demanded by a household every 15 minutes, how do you decompose that to the appliance level? How do you decompose that into, oh, this must be coming from the refrigerator freezer, this much must be coming from the air conditioning, this much from lighting and so on. You're only getting one signal.

Paul McDonald:
There's three ingredients to getting that prediction right. The first one is data. You need mountains and mountains of training data in order to make a machine learning model that's going to be useful. The second is you need a really effective data science organization. Our team of data scientists, when building these predictive models, they consulted over 2000 public research papers into load disaggregation. It's also known as non-intrusive load monitoring. And they identified the current state of the art in using smart meter data to make these predictions. And then they found ways to improve upon it. So, the third ingredient is the machine learning technology that you use to start with, and our team uses deep learning. These are deep neural networks and massive amounts of computing power that train to make as accurate predictions as possible. Now, how do we know that they're accurate is the real question?

It requires ground truth data. So, there are plenty of sources of this in the public domain where Pecan Street is one example of a project where homes have been sub-metered down to the appliance. So, there is data available where you can check your predictions against that ground truth data. One of the things that Opower does is encourages customers to provide information themselves about their homes and their home energy use. And millions and millions of customers have provided tens of millions of data points about their homes through our practice of digitally engaging with them that serves as a pretty unique and really powerful ground truth data source. So, our data science team is able to create and train models and then check them against these sources of ground truth data that were not used to train the models, they're unseen homes. And with that kind of rigor, we're able to determine with certainty the accuracy level of these predictions.

Mike Toffel:
Got It.

Paul McDonald:
And that's about as far as I can go in terms of data science detail without getting way over my skis.

Mike Toffel:
That's terrific. I think that's as far as we need to go into the technology piece. I mean I imagine some of the signatures of the demand are recognizable. For example, recharging an electric vehicle that sort of has a signature that you can detect and so all of a sudden some houses have that, some houses don't. You can get a sense of who has EVs and when they are charging, other things must be correlated with temperature, the heating and cooling systems of a home. Other systems must be correlated with time of day. So, around the dinnertime hour, you see a signature draw, you're like, oh, that's probably their stove and the additional lighting and maybe some fan use, things like that. And you're putting all this together along with the modeling that you're talking about and the enhancements to the modeling to sort of disaggregate and then use all that to generate recommendations. Does that sound about right?

Paul McDonald:
Exactly. Yeah, you got it. The thing that our data science team loves to say is all machine learning models aren't wrong. Some of them are useful, but when you have really large volumes of data, really powerful machine learning technology and an incredibly experienced team, you can work those models to accuracy, precision recall levels where they are incredibly useful. For example, some of the easier to detect signatures, we can disaggregate that energy use with accuracy up to 99%. Some other end uses, they're a little less prevalent, a little more hard detect, like a pool pump may be only accurate to 92%. But 92 to 99% accurate is pretty darn good for showing a customer going on in your home and here's what you should do about it

Mike Toffel:
Right. So, this has all been talking about energy efficiency and recommendations on energy efficiency. Another piece of the story of Opower products and services are load shifting to cleaner energy. And I mentioned EV charging, which I imagine is one candidate for that. When you come home and your EV has 50% and you want to bring it up to a hundred percent, maybe you're not going to use it for another 12 hours. So, you have flexibility in when it charges. Right now, if you just plug it in, it'll charge according to the battery demand, but it's not necessarily very smart. Or similarly with if you run a load of laundry and you turn it on when you go to bed and you just kind of want it done by seven in the morning, you don't really care which 90 minutes it runs in. And my sense is that some of your products that you have are sort of helping guide customers either with recommendations or even taking over some of the equipment and saying like, we're going to run this now versus later. Can you talk about this whole line of load shifting?

Paul McDonald:
Yes, this is incredibly important to reducing utility emissions and doing it in such a way that keeps the transition affordable. So, I'll talk about one specific client that’s been running an Opower service called Behavioral Demand Response. So, CPS Energy is a municipal utility in San Antonio. They for the past 10 plus years have been running a demand side management program that is a little bit unique. Most of these programs over the course of decades have been geared towards reducing energy consumption. Energy efficiency has been the purpose of it. But CPS operates in Texas in the ERCOT market where prices can be very volatile for transmission, especially when demand is really high. And they have an imperative to keep the price of energy affordable for their customers and to deliver that service reliably and as cleanly as possible. So, they had been running a demand side management portfolio that was specifically geared towards reducing demand, and it was for those reasons, reliability, affordability and cleanliness of service. Coming up at the end of this 10 year plan going into 2020, they were a little bit short of their goal.

So, they needed a way to reduce demand at scale and quickly. Now this point is a little bit of a nuance, but I'll just say it can take time and expense and effort for utilities to build demand flexibility portfolios that are device-based because it requires getting customer after customer after customer to opt-in and enroll, to allow their device to be controlled to reduce demand. CPS asked Opower to launch a behavioral demand response program because it doesn't require that opt-in. We can automatically enroll customers into an experience that helps them reduce demand on peak days. So, the way it works is when the conditions in ERCOT are right, when it looks like CPS is heading for one of their peak periods for the summer, that will set prices within the following year, they'll call an event and within a few hours Opower will dispatch communications to hundreds of thousands of San Antonio customers that tell them, hey, tomorrow we're going to play a game.

They call it power players at CPS. And it shows customers, here's how you ranked against other similar folks in your community during the last power players' event. Here's when it's happening and here's what you can do in your home as a personalized recommendation to reduce your demand during the event. The day immediately after the event, Opower follows up and shows customers how they did. Here's your new rank. The person who is number one gets a little bit of a badge and they get some more recommendations on how they could improve their performance in the game the next time around. And what CPS got as the key metric that matters out of this program was well over 20 megawatts of demand reduction each event in the summer of 2020 that put them over the top of their demand side management goal.

20 megawatts is about the size of a utility scale solar array, and those can runon the order of a million dollars per megawatt to develop and bring that capacity online. CPS got 20 megawatts without having to do any of that in a much more cost-effective way through Opower. So, it's one example of a utility using behavioral demand response to create a new gamified experience for their customers and reduce demand in a big way.

Mike Toffel:
And it points out the value of shaving the peak to avoid investments in new infrastructure simply by changing behavior.

Paul McDonald:
Exactly. Exactly.

There's another service that's becoming increasingly important in the industry as time of use rates become more prevalent, these are financial incentives into the pricing plan that you and I might pay for energy that encourage us to use less during those peak periods where the system is most constrained and the generation supply might be at its dirtiest. We created a new product a couple of years ago called Behavioral Load Shaping, and what it does is continually educate customers about how their pricing plan works because, again, most customers don't understand how this pricing plan works, even if they select it. And it shows customers on a weekly basis, here's what you're using and spending during that peak period, here's what appliances are costing you the most money, and here's what you can do to reduce your spending away from that peak period.

Arizona Public Service is one utility that's been running that service for its customers for a few years, They found that overall customers receiving those weekly behavioral load shaping communications from Opower agreed that they were overall satisfied with Arizona Public Service at a 13% higher rate than customers who weren't. And just as powerfully they saw dramatic reduction in customers who were reporting that they were dissatisfied within Arizona Public Service. Over the course of one summer, they shifted 250 megawatt hours of energy use off the peak from a relatively small group of customers, about 40,000 customers. But we got our first look at how that product performs at scale and in extreme heat last summer. Mike, you may remember the heatwave that rolled through California last September.

There's a large West coast utility that had defaulted a large number of customers on the time of use rates just recently, and they deployed behavioral load shaping at scale to almost 800,000 customers. And they deployed it for the same reason that APS initially did to help customers just understand this pricing plan that they were being moved onto. But what they found was that customers getting that experience in total reduce their demand by 14 and a half megawatts every single day during the heatwave. So, that showed us that this can be a valuable grid resource, especially as climate change creates more extreme heating patterns and the system still needs to operate with resiliency and reliability.

Mike Toffel:
Yeah, so you're closing the information gap between those who are on variable pricing schemes, but then don't find out how much they have to pay until the end of the month. You're giving them, it sounds like, weekly updates and perhaps even more frequent. Have you ever moved all the way to real time or daily price awareness? If you send an email to someone or notification on their phone, it says, today you spent $18 on electricity, that's $8 more than usual.

And for some things like cable TV where it's sort of the same price every month, well, that's not a big deal. But here, especially if there's variable pricing to find out at the end of the month, wow, this month I just spent twice as much as normal, how am I going to scrape up the money to pay for that? That's a whole different ballgame.

Paul McDonald:
It is a very different ballgame. It's kind of funny that this is one of the last variable monthly bills that we have and it can be the most unpredictable. And what makes matters, even more acute right now is the rate at which energy prices have been rising over the past 12 months and will continue to rise as these infrastructure investments get made. We have not gone all the way to notifying customers every day or in real time about their energy use and spending. However, some utilities offer prepay programs that will include more of that ongoing communication to customers so that they don't overrun their balance.

There is also a new piece of technology that's starting to be deployed in the industry. The Moniker Ford is AMI, Advanced Metering Infrastructure 2.0. These are smart meters in the field that have the ability to give customers connectivity directly to them from their wifi network so they can view their usage in real time and they can get notifications in real time if they're approaching their new demand peak. We have a few clients that are deploying this technology right now and we're in the thick of designing the right customer experience that will serve the purpose of getting the customers that are really interested, easy access to that real-time energy information and spending information but not overwhelm customers with so much information that they'll lose the signal from the noise. It comes back to that design principle I mentioned a little while ago that we have to assume that most energy consumers have more important things to do than worry about their energy bill. So, we work hard to design experience that will get them in, get them out, and get them taking action as quickly as possible.

Mike Toffel:
How far off are we from a load shifting perspective where if I had an EV, I plug it in and having already set some parameters about cost and greenness of the power, it'll just figure out when in the next 12 hours it should power it or similarly, the laundry load or dishwasher load. I need it in 12 hours, I don't really care which. How far away is are we from having these things just taking care of themselves?

Paul McDonald:
Utilities are increasingly beginning to run what are called managed charging programs where they will offer customers a financial incentive in order to give the utility access either to their smart charger associated with their home or directly to their car. So, the utility can determine for them when it's going to charge. We're in the early stages, but it's only going to grow from here. If you look at forecasts of smart device adoption in the utility industry, they're called distributed energy resources. They're just growing at an increasingly dramatic pace over the next 10 years. And what's interesting, Mike, is that creates both an opportunity and a challenge for utilities. As customers adopt these devices, if they are invisible to the utility, they can create spiky new loads, they can create situations of voltage sag with intermittent generation, they can create reliability issues. It just makes the distribution network more volatile.

But if utilities do have visibility and the ability to dispatch those devices, they can operate the whole system more efficiently by instantaneously balancing clean energy supply as it exists in that millisecond with demand to match it. The only way they get that is if utilities use their customer experience to become part of the enrollment process

Mike Toffel:
Yeah, that's a really interesting point about their need to understand these predictive and behavioral measures because if customers install them in a way that's hidden to the utility, if all of them get the signal, say charge now because it's cheap, all of a sudden there's a spike at three in the morning which the utility hadn't planned for and now has to sort of address in real time, which is a totally different thing than if they knew and even were instructing the charging stations to charge you at three o'clock, you at 3:10, you at 3:20 and they were planning the supply at the same time as this demand. It's a really interesting point.

Paul McDonald:
The takeaway for me as a professional from all of this stuff is there is massive amounts of change happening on the supply side of the energy industry and there is massive amounts of change happening in the devices and ways consumers use energy and making all of that work together, that is a problem that you can make a career out of solving.

Mike Toffel:
And all of this comes about in a moment of increased electrification. And so I know that Opower's doing some work here as well thinking about using behavioral science and AI for education campaigns on the EV side, helping to inform people about how many of their neighbors have EVs, for example,. Can you tell us a little bit about that work?

Paul McDonald:
Yes. So, this is an incubation pilot. We don't know if we will be successful here, but we're trying with one of our most innovative clients. Baltimore Gas and Electric had a variety of firsts with us, and this is the next one. we're running a pilot with BGE right now where over the course of a year we're using Opower's machine learning to detect customers that do not yet have an EV charging in their home. And we are showing them using those same machine learning models which customers around them in their neighborhood are charging EVs in their home and putting to work a behavioral science technique called social proof.

This is a normal thing to do along with a few different predictive analytics and behavioral science techniques, all with the goal of increasing the rate at which Baltimore Gas and electric customers use BGE's web tools that help them understand their options for purchasing EVs, understand what it means to buy an EV and how they might go about it. We're trying to increase customer awareness and intention to purchase an electric vehicle just by virtue of getting these communications from Opower. And increasing the rate at which customers actually buy electric vehicles in Baltimore Gas and Electric service territory. That is the most aspirational one that we're measuring for in the course of this pilot.

We're very eager to see whether we can actually bend the adoption curve just by using AI and behavioral science and the utility customer experience.

Mike Toffel:
And the way you'll know whether EV purchasing behavior has changed could be, I imagine, either looking at registration data since folks have to register their vehicles or by looking at the change in the electricity demand signatures. Are those the mechanisms that you're going to evaluate this?

Paul McDonald:
Yes. Once we predict it, it's very simple to include a magic link in an email that says, hey, it looks like you're charging an EV at home right now. Is that right? Once they click, yes. That's new ground truth data. It confirms the results of the experiment and serves to improve the predictive power of those machine learning models over time.

Mike Toffel:
This sort of reminds me of when you go into hotels and they used to just say, if you want to save the earth, don't have your laundry done every day, and now they're telling you 95% of people on this floor don't have their towels laundered every day. Maybe you should be like them. And there's something sort of similar about providing statistics that show you, oh yeah, that's the group I want to be a part of, and that's the normalization.

Paul McDonald:
Yes, there's a behavioral science technique here that I just recently got exposed to, thanks to a professor at Wharton, Katie Milkman, she told me of a study that showed that if you use those kinds of messaging treatments that you just described, did you know that this percentage of people do this? People will behave more in accordance with that. But if you tweak it just slightly, the name of this other technique is called Dynamic Norms. If you just tell people, did you know that in the past year 50% and more people did this thing, then they did it last year, you get an even more dramatic increase in customers kind of trying to norm to that Dynamic Norm. So, we're excited to put techniques like that to work pretty soon in the utility customer experience.

Mike Toffel:
That's really interesting. So, one last area that I wanted to talk about of the products and services you're offering has to do with ensuring equity and affordability in the energy transition. So, you alluded to this a little bit earlier in our conversation, but I wonder if there's other techniques that you wanted to sort of touch upon?

Paul McDonald:
The problem here is number one, energy is becoming increasingly unaffordable. Number two, there are currently 97 million Americans living below 200% of the federal poverty level. That's squarely in the low and moderate income range. And that limit is often used as a qualifier for whether customers can get assistance paying their energy bills. And number three, I think prices are going to continue to rise as infrastructure needs to get invested in, and there's a real risk that limited income customers could get stuck with the cost of a more expensive, albeit cleaner energy system. Part of the answer to this is helping get limited income customers participating in programs that make their bills affordable and given equitable access to the clean energy programs that are being run in their communities. We've done a lot of research in this area over the past few years and we've been learning about some of the barriers that get in the way of customers taking actions on these programs.

They can be things like skepticism. If somebody's offering something for free, it's like, is that real or is this a scam? Another can be pride is one way to say it. This isn't necessarily for me. I don't identify as someone who could or should get access to this kind of program. So, we have been researching and designing and piloting experiences to cut through those psychological barriers and get more limited income customers participating in really well-funded programs that can help them out. We ran an early pilot with Washington Gas, which is my local utility, last year. They were looking to get more customers in their service territory access to funds that could help pay down an account ballot with a Virginia Energy Assistance program. So, what we did was we applied Opower's predictive analytics first to find these customers. We merged multiple data sets describing areas of need to predict what customers will likely be eligible to participate.

And once we define that customer set, we reached out to them with an experience that used a behavioral science technique to show them that they are not alone. And it showed them a huge number. This is how many customers have gotten access to this program in the past year. And when we user tested it with a small sample size, customers were like, wow, That blows my mind that this many people would be getting this access. And then when we field tested this experience with Washington Gas, the very first metric we got was pretty incredible. Customers in arrears in Washington DC right across the river from us, clicked on that communication to explore this program at a rate three times the industry average, which was just unheard of. That's the highest click rate we've ever seen from any communication of Opower's ever delivered.

And this is exciting for us as a company because this problem is something that we've been rallying around and making real investments in over the past couple of years. So, the early indications are that when we again use predictive analytics and behavioral science in the utility customer experience, we can get customers taking actions that are going to be good for them and for the utility and ultimately for the planet.

Mike Toffel:
That's a really inspiring story to help those who need help with affordability get the information they need in a way that's really resonating with them. It sort of reminds me of college or high school orientations where sometimes you'll see slide decks exposing students to statistics of how many themselves of tutors or of psychological counseling to sort of create the idea that it's normal and that it's perfectly appropriate to seek those types of assistant programs sort of resonates with them in I think a similar way.

Paul McDonald:
Exactly, that's just one piece of the problem. There's a lot of different needs within the low and moderate income sector of customers that utilities serve, and that kind of journey to affordability can start in a moment of financial crisis. You could be facing service shutoffs for not being able to afford your bill for a long time. So, getting them out of that moment of financial crisis is step one. And we've got early indications that we could do that, but it does not end there if we bring it back to reducing emissions and making an impact on climate.

Again, there are really well funded programs to help low and moderate income customers make free upgrades to their homes that will reduce their emissions and make dramatic reductions in their utility bills for the long term. These are things like weatherizing their homes or upgrading an ancient HVAC system that may be using way more energy than is normal for any HVAC system just because it's so far degraded. The goal is to get as many customers through that entire journey, taking multiple actions in order to get them on a long-term path to energy affordability that also reduces theirs and the utilities in their profile long term.

Mike Toffel:
So, what's next for Opower?

Paul McDonald:
I'll point to two things really quickly. Number one is using more and more of the small touchpoints that utilities have with their customers to influence beneficial actions. So, this is where it starts to become more of a technology problem. Every time a customer visits the utility web portal or calls the utility call center or interacts with the utility mobile app or receives a utility bill, every single one of those moments is an opportunity to get the customer's attention with something interesting predicted and identified about their energy use and make a personalized recommendation for what they can do next. That can be technologically complex. It comes back down to data integration and technology integration. So, we have entire teams of people that continually work on making that embedding of these features into the utility customer experience as easy as possible just so that the cost benefit works out for a utility and they can do more of what we want them doing.

So, that's part of the future for me is using more utility customer experience channels more often to influence beneficial actions. Utilities have a captive audience in the same way that Facebook and Google have a captive audience. Utilities don't necessarily monetize that audience in any particular way. However, if I was an HVAC contractor, or a heat pump technician, or somebody selling smart thermostats, or electric vehicles, or solar panels, any of these devices that are used by people who are served by a utility, I would probably want to use the utility customer experience to accelerate the pace at which customers are purchasing these things. I think there are opportunities within the confines of the utility regulatory compact for utilities to fund the programs that they're running by bringing third parties into it and using the audience that they have just to provide good advice, not to sell them something in a plaid jacket, but to provide great personalized advice of what they can do as an energy consumer.

The other comes right back to demand flexibility. So, we talked a little bit about how, with more devices connected at the edge of the grid, it gets more complex from the consumer side and with more renewables coming online on the supply side of the grid, that gets more complex as well. The task at hand is to balance supply and demand at every single millisecond of every day and do it as efficiently as possible without having to overbuild the system. The answer is going to include artificial intelligence that is at the core of the utility operations' system that manages their network. Continually sensing and responding to conditions on the supply side of the grid and instantly responding by dispatching resources on the demand side of the grid to keep things in balance.

So, for example, when the wind stops blowing or a cloud passes over the utility scale solar array, the utilities network management system needs to sense that and respond by dispatching thermostats, EV chargers, solar panels, batteries, to shed just the right amount of load in the right spot in order to keep things in balance. And the reverse is also true. It's kind of crazy to curtail wind and solar, just not use them when you have excess supply and not demand to meet it. I think utilities are going to need to increasingly dispatch demand, have cars start charging, have homes start pre-cooling to use that clean renewable energy when it's available. We're solving all of these problems for our utility clients and part of the future I think means using the customer experience to get utilities more visibility and access to those devices at the edge of their grid and then using their network management system to operate it.

Mike Toffel:
Terrific. Very interesting future in this space. So, let me ask our final question, which is that some of our listeners are interested in career opportunities in business and climate change in some manner or another, whether it be smart grid and data analytics or other dimensions. What type of advice do you give folks when they ask you how I get into this space?

Paul McDonald:
I would typically give the same advice I give my own team when it comes to career planning, and I think that's the sweet spot for doing your best work and being most fulfilled by it. That sweet spot's going to be the intersection of what you're good at now. What skills and expertise do you have that you can bring to bear? What are you passionate about, what do you want to learn and do next and what's going to be needed from you? It really means thinking real time for some internal reflection. I do mine best with a keyboard at my fingertips so I can go back and read what I just wrote while I was thinking about what it is I'm good at, what I want to do next. But it also means making real time for research.

I mean, scouring the internet. You want to start up, look where VCs are investing in right now, and there's nothing that quite replaces those delightfully awkward informational interviews with people that might be in or near or not even close to your professional network. Those are real learning opportunities for what might be needed from the companies that are out there. I'll raise my hand and say, any one of your listeners can reach out to me and be happy to help them with one of those delightfully awkward conversations.

Mike Toffel:
Thank you. I think what people forget or maybe underestimate is that people generally like talking about what they do, especially if they're excited about their company and their career, and so while you're asking them for a favor to talk to someone they don't know well or who's a friend of a friend, it's actually kind of a fun thing to do.

Paul McDonald:
It really is. It really is. I go back to one of the first things I mentioned. One of the most fun and fulfilling parts of my job is helping our clients tell the stories of what they're getting done with Opower technology. That dynamic exists there. Well, when people are doing work that they're excited about, they will want to talk about it. It's a good nudge to get over any kind of hesitation folks might have. So, reaching out cold on LinkedIn or through some other channel.

I got one more specific recommendation I would just offer people that are looking to get into this space. Myself and my Opower colleagues have been involved in an organization called the Clean Energy Leadership Institute for a number of years. This is an organization that teaches emerging leaders just about the opportunities to lead in the clean energy space and provides great education and great access to people that are doing this work right now. So, it's a really good forum for doing some of the kind of research and learning and introspection that I just mentioned a moment ago.

Mike Toffel:
Great, and we'll put a link to that in the show notes of this episode. Well, Paul, thank you so much. I really appreciate you spending time with us.

Paul McDonald:
This was a ton of fun. Thanks for having me on, Mike.