 Well, hello everyone and good afternoon, good evening or good morning depending on where you're joining us from today. Welcome to Engineering for Change or you foresee for short. Today we're very pleased to bring you our alias installment in the 2018 webinar series on the topic of improving water and energy service delivery with IoT solutions. My name is Yana Aranda and I'm president here at Engineering for Change and will be your moderator for today's webinar. The webinar you're participating in today will be archived on our webinars page and our YouTube channel. Both the URLs for those locations are listed here. Information on upcoming webinars is also available on our webinar page. E4C members will receive invitations to the webinars directly. If you have any questions, comments and recommendations for future topics and speakers, please contact the E4C webinar series team at webinars at engineeringforchange.org. And if you're following us on Twitter today, I invite you to join the conversation with our dedicated hashtag, hashtag E4C webinars. Now before we move on to our presenter, I'd like to tell you a bit about Engineering for Change. E4C is a knowledge organization and global community of over 1 million engineers, designers, development practitioners and social scientists who are leveraging technology to solve quality of life challenges faced by underserved communities. Some of those challenges may include access to clean water and sanitation, sustainable energy solutions, improved agriculture and more. We invite you to become a member. E4C membership is free and provides access to news and thought leadership, insights on hundreds of essential technologies in our solutions library, professional development resources and current opportunities like jobs, funding calls, fellowships and more. E4C members also enjoy a unique user experience based on their site's behavior and engagement. Finally, the more you interact with our site, the better we will be able to serve you resources aligned to your needs. For more information, visit our website to learn more and sign up. Now a few housekeeping items before we get started. We'd like to take everyone through a practice run of WebEx by starting with a simple question. Tell us where you are. So in the chat window which is located at the bottom right hand of your screen, please type in your location and I'll go ahead and get us started today. So I'm in New York, you should all see that right now. There we go. If the chat window is not open on your screen, try clicking the chat icon in the top right hand corner. You can use this window to share remarks during the webinar with your fellow attendees and just share more information with us. We have some folks here. I already see folks answering questions. I also see you all answering in the Q&A window. That's reserved for questions for our presenters. But I do see in the chat window that we have folks from Nairobi, from Atlanta, from DC, from Orlando, Colorado, Rwanda, Indiana, all over the United States and the world. We have attendees from China and LA, France, welcome, welcome everyone. You can, if you have a technical question, feel free to send a private chat to the E4C admin. During the webinar, please use the Q&A window, which is located below the chat, to type in your questions for the presenters so we can keep track of those. Again, if you don't see it, the Q&A icon is at the top right hand corner of the screen. If you're listening to the audio broadcast and you encounter any trouble, try hitting stop and then start. You may also want to try opening up WebEx in a different browser. E4C webinars qualify engineers for one professional development hour. To request your PDH, please follow the instructions on the top of our professional development page after the presentation. The link is available on this slide. A special note about today's webinar, we will stop 15 minutes before the hour or approximately 7.45 a.m. Eastern Standard. But we will be sure to set aside enough time for Q&A. If we don't catch your questions, you can certainly email them to us. Now, all right, so I'd like to take a moment to tell you a bit about today's webinar and our presenter. Service providers worldwide are working to deliver clean water, sanitation and energy to underserved communities, but a gap remains between quality of service that providers intend to deliver and the actual impact over time. IoT Solutions may name bulletproof services by measuring and reporting on service delivery. Today, we're so excited to be joined by Evan Thomas, who is the Associate Professor at Portland State University and Oregon Health and Science University and co-author of a new world-wide support on innovations and wash impact measures. He will provide us insight on IoT applications based on findings of sensor studies from programs in Rwanda, Kenya, Ethiopia, and Somali regions and align us to the possibilities of remote sensing. Now, Evan is incredibly well accomplished, so we only have a bit of his bio here, and I encourage you to read his whole bio on our platform. What you can see here is that Evan happens to hold a page in aerospace engineering from University of Colorado Boulder, is a registered PE and holds a master's in public health from the Oregon Health and Science University. His research has been funded by National Science Foundation, the World Bank, UNICEF, the United States Agency for International Development, or USAID, the United Nations Foundation, A9 States Centers for Disease Control and Prevention, among others. We are so excited to have him here with us. And at this point, I will turn it over to Evan to share his insights. Thanks, Yana. Okay, hi, everybody. Thanks for joining us this morning. Yana, can you confirm you can see my screen? I see your screen. You're looking great. Okay, great. As Yana mentioned, I'm going to have to take questions at the end, because I can't see the questions coming in when I'm showing the PowerPoint. This presentation today, as Yana mentioned, is going to cover a few example applications of the Internet of Things for improving water and sanitation interventions in developing countries. Some of the work that our team has done, and I'll also briefly introduce the book that Yana mentioned that we just published with the World Bank and the Inter-American Development Bank on Technologies and Methods for Improving Water and Sanitation Measurements in the Sustainable Development Goal Period. So here we are. We all recognize this place, of course. What's pretty cool about this image is that it's actually not a graphic. It's a series of thousands of images of every point on the earth, on a cloud-free day. So this looks familiar. Here's another way of looking at the planet. I'm sure many of you've seen this picture before. This is, of course, lights, electricity. It shows us where there are population centers, but it really shows us where there are rich people's population centers. You can see the United States, Western Europe. You can see India and China emerging, and then that's the Nile all lit up in North Africa. But there are billions of people that don't have access to electricity and energy every day. There's a billion people that still lack access to clean drinking water. Two billion people, half the world's population still use their masks on open campfires every day for the daily energy use. And one of the perverse effects of climate change is that the people that are the most impacted by climate change are the ones that are the least resilient to it and the least responsible. So you can see that's where these people live. These are fires. Some of these are wildfires. The most of these are campfires. Almost 5 million children die every year from respiratory disease associated with indoor and outdoor air pollution. And much of this is associated with biomass cookstows and campfires. Another 5 million children die every year from diarrhea associated with water and sanitation issues. And as mentioned, people living in these environments up here in Africa and many other least developed countries are the least resilient to the global effects of climate change. There's only 4% less water in the tributaries of the Nile. Rainfall patterns are changing. Even disease patterns are changing. Historically, there really wasn't malaria in Nairobi because the malaria parasites couldn't live about 5,000 feet. But as the planet warms, malaria is now persistent in Nairobi for the first time in human history. So there's a lot of effects of these challenges of poverty are exacerbated by the global effects of climate change. Now, as many of you, of course, experts and our practitioners in this field, there's a lot of efforts large and small to try to address these persistent challenges of access to safe water, sanitation and energy. But many of them look like this. Money comes in from a donor like the Gates Foundation or USA, the Year of the World Bank, and goes into implementations that often have a technology component, water filters, cookstows, off-grade electricity, water pumps, latrines, that are all designed to impact people's health and livelihoods. That impact is not well measured and it's not taken a lift. About $40 billion a year is spent by the international community on water and sanitation programs in developing countries, $40 billion a year. But at any given time, about half of the water points are broken. So half of that money going in every year will go into infrastructure that's broken in a couple of years and in many cases never gets fixed. Furthermore, about half of the functioning water infrastructure is contaminated. So about three quarters of the gains that we expected to achieve with this money and during the Millennium Development Goal period goes into infrastructure that's not providing access to safe drinking water. And one of the challenges is that we don't continuously monitor impact. We have money coming in from a donor going into an implementation, but all organizations only have a couple of years of budget to implement something to try to get a community or a country or a government to adopt a technology or a solution. And yet, we don't provide ongoing support to incentivize service delivery. So we don't incentivize the actual impact of the intent of our spending. Our team, which is researchers and practitioners and other professionals working in partnership with nonprofits and companies and governments have a few different mechanisms that we've been testing over the past few years to try to address this challenge of feedback and sustained service delivery. One is through business solutions. So we try to create pay for performance crediting systems that incentivize long-term impacts. So things like carving credits, health credits, pay for performance social impact bonds, and trust funds that try to create an environment that incentivizes long-term. And the other is through Mojies, internet of things technologies that can remotely monitor the durability, sustainability function and impact of water, sanitation, and energy intervention. So this is part of our technology. These systems are connected over cellular and satellite networks and monitor these interventions long-term. The environment we're trying to create looks like this. When you have a water pump in a developing country or in a community and you are continuously monitoring it using technology, you're able to use that monitoring to dispatch technicians. So instead of installing a pump and walking away from it or hoping a community is able to sustain their own water pump, you can professionalize services, use monitoring to allow dispatch and to enable dispatch, and then link that maintenance activity to funding. So that funders and communities and governments are paying for services that are delivered rather than infrastructure installed. Our team works across a few of these domains. So our theory of change is that when we monitor technologies and we monitor interventions, then we can link those interventions to impact evaluations that demonstrate the more cost effective service delivery. We can improve revenue collection and create new business models. And we can coordinate external resources, including national governments and international donors. So our team works across this domain of technology, service delivery, capacity building, and business development. Towards the goal of achieving universal and equitable access to safe and affordable drinking water for all, which is the new 6.1 Sustainable Development Goal. I'm going to give a few examples on the next few slides of where we've applied this theory of change. A typical hand pump, hundreds of thousands of these are installed all over Saharan Africa, and as I mentioned, about half are broken at any given time. Now you might look at that broken water pump and say, okay, well I'm going to do water pump that doesn't break in this way. Well, everything that anybody builds or designs will break. This isn't a question of those boats shearing off. It's a question of the institutions surrounding that water pump that are supposed to keep it running and right now aren't able to. Our team developed sensors and installed hundreds of these across Western Rwanda a few years ago with funding from UK Difford, the UK Department for International Development. And we compared three different service models of hand pumps. In one service model, it was the business as usual model. So the implementer, which was Living Water International, went around and maintained pumps when they could get to them. When they were able to schedule staff and they had the supplies and in some cases communities were demanding enough to demand a repair. And we compared that against what's considered the best practice in this field of a circuit rider model where there were full-time technicians with vehicles, with supplies, who went around and each technician was responsible for 25 pumps. So they went around in this circle and did preventative maintenance as well as repairs on pumps. And then in the third group, we looked at what I call the ambulance model, where our sensor data actively monitored pump functionality and when a pump broke, a technician was dispatched. So we had sensors in all three of these groups, but only in the third group was the data available during the trial to the implementer to actually respond. And baseline, when a water pump broke, it was broken for more than 200 days, so more than half a year. And this is what other studies have shown is about the level of performance we have right now at rural hand pumps up to here in Africa. In the circuit rider model, without the benefit of sensors, but with the benefit of full-time staff, the implementer was able to get this down to about 60 days and move average functionality across the group of pumps from about 50% to over 70%. And then in the ambulance model, the model where the sensor data was actively used, we were able to get that service interval to about 20 days, and average functionality to about 90%. And it's also more cost-effective this way. As you imagine, if you bought a new car every time you got a flat tire, it would be an expensive way to get around. And yet that's kind of what we do with the hand pump and water schemes across the sector. If we implement repairs and service delivery, it's a more cost-effective way of delivering water. Now, 90% is good. Above 90%, you accrue disproportionately positive health benefits. Now, that's assuming the water is clean, which is also another whole area of research. And above 90%, communities are more willing to pay for water services. When your water only runs half the time, I don't think any of us would pay our water bill. But above 90%, you can unlock new revenue from communities who want reliable services. We're trying to drive that 90% to 100% now. With support from the National Science Foundation, we're applying machine learning tools to actually predict when a water pump is going to fail. But if it's about to break, here's what that looks like. In this first section, you see the sensors working, the pumps working, people are getting water. And then you see a pump that is misbehaving. It's what I call the tire pressure warning light for a hand pump. We get almost a week of data where the pump seems to be acting off nominal. So it's delivering water, but something's changed. With machine learning, we don't necessarily know exactly what's changed. It's not a prescriptive algorithm. It takes into account a whole bunch of different features, including the age of the pump, the actual sensor data we're getting, which is vibration, as well as water pressure, the season, a lot of other features that help us predict when a water pump is about to break before it actually breaks. So this is our magic window. We want to deploy a technician during that period of pump in need of service before the pump actually breaks so that we can maintain 100% of our delivery. And we're getting closer. This shows some of our performance of our algorithm. We have really high sensitivity and specificity of being able to identify a currently broken pump. So we know if the pump's broken now or not. With a one-week horizon, we have about an 80% positive predictive value, which means that 80% of the time we've correctly identified that a pump is about to break within the next week. Now, our sensitivity is only about 50% right now in our current threshold of 50% confidence, which means that we're only catching about half of those failures. But when we catch a failure, it's usually right. We're in the middle of this work, and we're trying to get this to the point where we can cost-effectively maintain 100% uptime predictor. One example where we've used sensors on hand pumps. Here's another example where we've used inside a very large household-level intervention. This is a typical cooking practices in Rwanda. About 80% of Rwandan families are rural subsistence farmers, and almost all of them use biomass. So fire what the daily energy needs, which is both an environmental and health hazard. And as already previously identified, most people get their water from these types of water sources, and they bring them home in jerry cans, and most of those cans are contaminated by the time that people consume that water. So clean water and clean air are two of the most critical environmental and health issues related today. Our team, which included our research team, our government of Rwanda partners at the Ministry of Health and the Rwandan Environmental Management Authority, and our private sector partners of Destegard-Francine, EcoZoom, and Del Agua Health, together, our team distributed over 350,000 cookstows and 100,000 water filters over the past few years to the poorest quarter of households in Rwanda. This is a Ube de Hei one or two household, the poorest quartile of households in Rwanda. That's the cookstove you see on the right side of the picture, and the water filter in the front. This is our coverage area. So in 2014, at the end of 2014, we covered the western province of Rwanda, 103,000 households, about half a million people with both cookstows and water filters. You see these sectors in the dark blue filter in 2014. These were control sectors. So we ran what's become the largest environmental health randomized controlled trial for either water filters or cookstows, and definitely the combination. Looking at adoption rates, so are people using the stows and filters, impacts on the environment, so air quality and water quality and personal exposure, and then both household level, self-identified, self-reported, respiratory distress and diarrhea, as well as clinically diagnosed respiratory and diarrhea, comparatively between the households in the yellow group, the didn't have stows and filters, against the households in blue, which did. And then through 2016, we also distributed stows to an additional 250,000 households in the eastern province and several northern districts, which is highlighted in green. This is the cookstove. It's a rocket stove. It reduces biomass use by about 50% to 70%, and air pollution or emissions by 50% to 70%. Really, that's not enough. That's not enough to have a positive health impact unless you combine it with moving the cookstove outside. So a big part of our behavior-change messaging on an ongoing basis was getting people to cook outdoors, exclusively with the ecosystem cookstove. And then this is the water filter, the LifeStraw family 2.0. It treats about 20,000 liters of water to the highly protective WHO rating. It has a five-liter input bucket and five liters of storage. This is a very effective water filter, but the challenge, similar to the cookstove, is getting people to use it exclusively, getting people to make sure they clean it and backwash it, getting kids to take a bottle of water in a clean container to school, or getting families to bring things really not sufficient without that ongoing behavioral messaging. This is what our distributions look like. We had about 2,000 community health workers working with us across 7,500 villages over the past few years. We had skits and plays and songs on the radio. The program is branded the AONISA, which means to live well in Kenya, Rwanda. All part of our behavioral messaging that we use on a community level, national level, and a household level to get people to adopt and use the stoves and filters exclusively and properly. And this is some of the example messaging that our community health workers provided. We also monitored this deployment and the ongoing monitoring with IT and ICT. We had all of our community health workers and our supervisors had smartphones reporting data on household demographics, use of stoves and filters. People signed contracts with us on the phones. We took GPS coordinates and pictures and all the rest of it, which let us monitor the program in real time and is also now our database for ongoing analysis. But we also used our sensors. One of the biggest challenges with environmental health interventions at the household level, whether they're water filters or stoves or latrines or bed nets or a chlorination of water, lots of different types of household level interventions, is that we can prove in a lab environment or in a highly controlled environment that a water filter works or a cook stove works so that a latrine is a good idea for reducing open defecation or contamination of surface water sources. And they have pretty sophisticated tools for estimating diarrhea or respiratory disease. And there are some biomarkers for exposure to cryptosporidium or respiratory health. But there's a messy middle, a lot of people are actually doing in their homes. And a lot of literature out there has really contradictory results. You can have the same water filter in a very similar context. Based on reports and observations, people seem to be using it. 80%, 90% adoption of a water filter, a stove or a bed net, and yet widely varying health outcomes. And there's a lot of different reasons why this could be the case. But one of the challenges is that some people, some research suggests that this implicates the technology. What we've identified is that observations and interviews are really poor indicators of actual behavior in households. If you ask somebody if they're using a water filter, they're going to have a few different biases. One is a recall bias. It's hard to remember how much water you drank yesterday. Another is a courtesy bias. People know that there's a right answer here, so they're going to give a favorable answer. The answer that they think that the interviewer or enumerator wants to hear. So to try to dig into this a little bit, we put sensors inside of our stoves and filters, and we compared a few different things. One, what people told us they did using the best available survey tools, what they did when they knew a sensor was there. So we had this interaction where we told people the sensor was in their water filter and stove and what it was doing. And then the third was a blinded group where people didn't know the sensor was there. They didn't know a sensor was monitoring. Now I'm sure some of you are asking about the applications there. Households were already consented with the RCTs, so they had a lot of monitoring happening. They had a lot of air pollution in their homes. They're getting blood samples, so they get consent to myriad monitoring. In this particular subsample, they didn't know that there was a sensor there on that particular day or particular week. And here's what we found. Using the best available survey tools, people tell us they use about a liter and a half of water per person per day, which also happens to be the right answer. When the sensor is there, it's only about a liter of water, and when the sensor is hidden from view, which is an available estimate of actual behavior, it's only about half a liter of water. And this persists for more than a month. These are aggregated weekly rounds. We actually collect every single incidence of use of the filter, but this is aggregated weekly. And you can see two levels of reactivity here. In the open arm, where people knew the sensor was there, shown in blue, you see a dramatic increase in use of water. And that persists for more than a month, beyond four weeks. But you also see that both arms decline over a month period. So there's a reactivity to the introduction of the filter in the first place and the reintroduction of the filter through a visit by an enumerator. And so constant reinforcement of healthy behavior is necessary and has been identified in many of the other studies. But we also see a reactivity where people know a sensor is there so they are reminded in one way or another to continue to their use of the filter. So when we aggregate all of these different indicators we have for this program, we were able to identify a 92% increase in the number of households with an indoor air pollution among cooked outdoors. We were able to get them to move outdoors. This translates to a 3% reduction in diarrhea among children under 5 and a 29% reduction in acute respiratory illness among children under 5. And extrapolating based on available models, we're able to estimate 90 averted deaths in children under 5 per year. Over 7,500 averted disability-adjusted life years. That's a mouthful. Averted disability-adjusted life year might be one person not sick for a year or 12 people not sick for a month, et cetera. It's an aggregation of all illness, in this case, avoided illness. This figure, this characteristic of an averted disability-adjusted life year, we're trying to monetize now, create into a commodity. Our colleagues at the University of California of Berkeley have already been successfully doing this with cook stove interventions where you're able to monitor indoor air pollution and personal exposure or translate it using these models to averted disability-adjusted life years. And the gold standard, a voluntary commodity trading body that got started with carbon credit, has now adopted a dailies as a new commodity that can be traded. We're excited about this because it, again, brings us a little bit closer to financially incentivizing impact instead of coverage or promises. All right. Moving on to a further example in the sanitation area. Sanergy is one of our partners in Kenya. Sanergy has franchise toilets where they service toilets every day and they collect the waste and they turn it into fertilizer. And one of their challenges was a day in how they could optimize that service. We use our sensors, which were motion detectors combined with RFID tags to monitor the use of the toilets to allow franchise owners to request services and then to allow Sanergy staff to log that they've actually performed service. So during this study, we were able to help Sanergy improve both their quality control and optimizing their servicing activities. Using a similar technology, again, our motion detectors, we've been able to monitor thousands of toilets in Bangladesh and India, mostly with, as part of, Gates Foundation's funded epidemiological studies trying to identify the health impact of latrine interventions. This is just one particular figure from one of our studies, but it makes the point that we've seen over and over again, which is basically what people say is not really what they do in terms of use of sanitation facilities. In another study, our colleagues showed that the only time that sensor data correlated to observational data was during the window of time that observers were actually there. So observers were faithfully recording use of latrines, but people were using latrines a lot more because the observers were there. So we've, again, identified that surveys and observations are not a really robust measurement tool for latrine use of sanitation use. In another study in West Bengal, India, we tested 7,000 water sources for 12 parameters, and we also conducted what's called a sanitary inspection. So sanitary inspection is basically technology-free conductive survey at water points to see if there are certain hazards, like, is there a latrine nearby, or is the concrete slab look like it's in good condition or good repair? And we used it to create a hazard index that is advertised as being predictive of water contamination. We tested 7,000 water sources and conducted sanitary inspections, and fortunately, in most cases, across three different categories of bacterial contamination, the sanitary inspection was barely better than chance at predicting water contamination. All right, moving on to some of the current work our team is doing, we're now looking at scale with water systems. This is a program in Kenya called Kenya Rapid, led by the Millennium Water Alliance that has a lot of other partners involved, including Catholic Relief Services and CARE and IBM. We're working across five northern counties to develop technology that helps us monitor the service delivery of water borehole schemes. So if you look on the left-hand side picture, this is a failure of a water point that our sensors detected. If you look at the lower right-hand corner, this is one of the boreholes that is currently a level of disrepair. These systems are in all sorts of levels of functionality. Some are very old and are being held together by the hero operators. Some are brains-taking new, but don't really have a service model around them yet. We developed a technology that lets us monitor all of these electrically driven mechanized borehole pumps. Each of these are distributing water to an average 5,000 people per person at a time. We're also doing this work in Ethiopia. Lowland Wash is another USAID activity. We're working across a FAR region and Somali region. In this case, we're instrumenting all of the water schemes for a FAR region for about 2 million people and about half of the water schemes in Somali region, about another 2 million people. We currently have about a million people's water supply under surveillance now between Kenya and Ethiopia, and by the end of this year, 2018, we'll have about 5 million people's water supply that we're monitoring. The technology is satellite-connected, so it runs off of a very small two-watt solar panel. You can see in the picture there. So from anywhere in the world, we can monitor the functionality of these systems, and that data goes back to a bunch of different institutions. Local operators, regional water bureaus, water and sanitation companies, ministries of infrastructure, national drought management authorities, and international donors, all of whom have a role and a responsibility in funding associated with keeping water running. This is what our dashboard looks like. This is Northern Kenya. We have about 100 systems for about half a million people under surveillance in Northern Kenya today. You see in green functional water points, in red broken water points, in blue water points that have been reported as under repair, and in yellow, you see ones that are low use. So maybe okay, but ones that you keep an eye on. And our value proposition here is to try to address both service delivery where there's a 40% to 50% downtime, so about half the time or a little bit less water isn't running. And then even when water is running, 40 to 50% of that water that's delivered is not billed for. So there's 50% increase in revenue to be achieved by water and sanitation companies by just cutting by a quarter their downtime and by a quarter their non-revenue water. So this is where we're trying to institutionalize the use of this technology and data. And we're also applying our machine learning tools here. This is just one snapshot from one take upon. We're actually learning tools are able to identify a broken water pump in red, a intermittent use pump, so maybe it's working but it might be in trouble, and then a highly functional pump, water is being delivered every day. We also see a clear impact of fetum. So in East Africa you can see that we have northern Kenya covered, northern Ethiopia. In that right hand, the eastern part of Ethiopia is the Somali region that we'll be covering this year. And this is the East Africa Valley. Where over the past 30 years, there's been 20 to 30% less rainfall than historic norms. It's called the East Africa Climate Paradox. Most climate models assume that there's going to be or predict more water in East Africa. In fact, there's been 20 to 30% less water in East Africa over the past 30 years. And we see how that impacts use of groundwater. In the chart on the right, we see rainfall two seasons in the region, and clearly two groups of use of our pumps. So we're able to predict use of our pumps and extraction from these pumps based on rainfall. And this is something we're partnering with NASA and remote sensing experts at RCMRD in Nairobi, Kenya to look at how remote sensing from satellites can help predict and inform use of groundwater resources. Just to wrap up here, here's a few of our recent, as we published a year ago, we're all years ago actually, is about the business model. So helping credit paper performance, social impact bonds from a variety of different partners, looking at how we can incentivize services instead of promises. And then our second book was published just last month by the World Bank and the Inter-American Development Bank, Innovations in Wash Impact Measures, identifies how surveys have some gaps in terms of actually measuring long-term impact and functionality of water points and sanitation interventions, and we highlight innovations from a lot of different partners, including our own work. One of the exciting things about the Sustainable Development Goal Period is this new service ladder approach. This has been identified by the Joint Monitoring Program. We're also co-authors in our new publication, where we're no longer saying a water system or sanitation system is unimproved or improved. We're identifying a ladder from surface water all the way through safely managed or open defecation all the way through safely managed. But one of the challenges is we don't yet have institutionalized and agreed upon metrics for each of these ladders, for each of these rungs in this ladder. Now, sensors can be part of that solution, but we're not going to put a sensor in every latrine in the world. So we have to learn what we can from instrumentation and combine it with improved surveys and other models to incentivize and monitor long-term performance. And this is why all of us that work in this development sector, Global Health or Development Engineering are inspired by pictures like these. We use these in our reports. This picture is a promise. It hasn't been anything. This picture is already about eight years old, actually. It's proof that these kids had clean water the day this picture was taken. We need to move away from pictures as promises to actually measuring performance and incentivizing performance over time. All right, that's what I have. I'm happy to take your questions down. I'm going to switch over so I can see them. Thank you so much, Evan. This was incredibly insightful and certainly a rich presentation. And we've had a number of questions come in. So I'll kick off with you here. So some of the folks asked about the center specifically. I did share a link indicating some information. But one question that came in is regarding the maintenance of the sensors. How do you ensure maintenance of the sensors and data transmission devices? Specifically, what percentage of the sensors should we expect to be dysfunctioned, i.e. break or have a breakdown or not be able to access the internet? Apparently, this particular user is trying to monitor meteorological data with IoT and facing technical challenges. So again, how do you ensure maintenance of the sensors and what percentage should be expected to be nonfunctional at any point? Yeah, that's a great question, of course. I think there's two parts of that answer. One part is, of course, the design. And as technologists, there's a tension in the development sector where Silicon Valley will say, fail fast, fail often, develop your minimum viable product, get a data out there. But the development sector expects, I think, rightfully, more highly functional, more robust technology. So we're all evangelists for seeing how new tech can help in the sector, but at the same time, technology development takes a long time. It's taken us eight years or six truly viable, scalable products. So in terms of that, we personalize the bump of the sensors, personalize contract manufacturing. The sensors that we've designed are purpose-built for this environment so they can handle a far Ethiopia, which is the hottest environment in the world literally. It runs off of a two-watt solar panel so the battery stays charged all the time. You don't have to be a plumber. You don't have to be an electrician to install it. So it's essentially a very easily installed sensor. It only takes about an hour once you're at a site to install it and leave. But then again, everything that builds will break. So our sensors do occasionally break. We have about a 4% downtime of our sensors, which is either data coverage like cellular coverage or an actual breakage of the sensors. So at any given time, across our hundreds of sensors monitoring these water systems, about 4% of them are down. To address that, we integrate maintaining the sensors with the same institutions that are responsible for maintaining the water points. So a broken sensor triggers a dispatch record the same way a broken pump does. A technician will go out to a site and swap the sensor out. They don't have to take the sensor or fix the electronics in the field. They're all interchangeable. So we've tried to make that as easy as possible, have as little downtime as possible, but we also have to definitely deal with the fact that sensors break. Yeah, so that's definitely the reality of any tech. So now speaking over to how the information that the sensors collect is being used, one of our listeners wants to know if you know how LifeStraw and its distributors will use the information you have found. Yeah, so within our program in Rwanda, we use this data in real time. So we were able to use this on a dramatic basis. We were able to look at the data as the program is rolling out and use it to improve the messaging. So we identified, for example, that people were not cooking out as much as we thought they were using our sensor data and using some other indicators. So we pulled down on our efforts to communicate outdoor cooking, which was things like impact and exposure impacts that we saw and that I shared. In terms of the water filter, we were able to use our sensor data to improve the design of the water filter. So in fact, what we have out there now is a evolution of the design that we worked on with Bestergaard Franson to address challenges in the use of the filter to optimize the product for that environment. And again, we tried to use the sensor data to improve our behavioral messaging over time to try to get exclusive use of these technologies. Now, I wouldn't say that we're fully successful at that. We don't have sensors out there all the time. We have to, you know, that takes effort and money to have them integrated. We only ever have them inside a sample of the intervention. So we tried to use statistics as our friend to be indicative of the wider population. So we've had some major success with that, but honestly, in a lot of ways, it creates more questions than it answers. It's on that slide identifying the way people are reacting to sensor data. I will say one of the things that we've done further is we're now part of a very large National Institutes of Health and Gates Foundation study that's running in four countries. We're working in Rwanda on cookstows again to try to establish the maximum possible health benefits of clean cooking. The trial will happen. And our role is to develop sensors that monitor indoor air pollution and provide that as active feedback, real-time feedback to households so that we can show them the current air pollution level. And if it goes above the WHO 24-hour limit for PM 2.5, particularly matter 2.5, we trigger an alarm. Not a fire alarm. Continue use of... Please remember to use your LPG stove. That's excellent. On the slide of the model itself, one question has come in a couple of times so I feel really compelled to address it. Have you quantified the economic loss of responding to a false positive and not responding to a false negative in relation to economic gains from correctly addressing true results? And this is noted as a quote-unquote a true world OOP failure. Yeah, great question. So in our hand pump work, we were able to identify that if you use our current algorithm, which has about an 80% positive predictive value, you can get about 99% uptime without additional cost in terms of cost per liter of water delivered or cost per functional pump day. I wouldn't say that that algorithm is ready for prime time yet. One of the big challenges is who's actually paying for that. There's those donors, not donors, but implementers. In this case, the water project has money from donors for putting in new water pumps. They also have money from maintenance. So even though during this trial, we identified how the water project could incentivize their maintenance or achieve better uptime cost effectively, there's still a challenge of how institutions adopt service delivery as a priority. And they were a very strong partner of ours during this trial. We are now moving on to focusing on water schemes where we can identify a much better cost effectiveness proposition. With 40 to 50% of water that's being delivered, not being billed for, and 40 to 50% of the water point, there's an opportunity for an increase from their maintenance. And for those of you who are interested, we have actually coming up in March webinar dedicated to circuit writer programs and other examples of water services if you're interested in learning more and further kind of diving into this and identifying opportunities for improvement. So I'm now swinging over to talking a little bit about the IOT data side of things. So there's a request for you to talk more about the decision process for utilizing mobile versus satellite networks for data transmission. Can you speak to that and specifically the pros and cons for GSM and satellite network transmission? Yeah, so the simple answer is that we use cellular whenever we can, whenever there's coverage, and we use satellite when there is no coverage. The more complex answer is that GSM is obviously very cost effective, but towers are only going to go up where they have at least 20,000 people, I think, using a tower. So they're in urban and peri-urban areas, but many, most of our work right now are in very rural areas where most of the people are pastoralists and they don't have the population density that's driving a cellular provider to put up a tower. So in those cases, we use satellite networks and we have a new partnership with a satellite company that is probably going to be able to be as competitive as cellular over the next year or so. But in the meantime, we're using expensive everyday networks and then when we can, cheaper GSM networks. We have a great partnership with areas communications who provide universal SIM cards for us that work in most places around the world so we don't have to set up local contracts for every different country or every different partner within a country. It doesn't work everywhere in Ethiopia. There's only one telecom provider it's run by the national government. The Ethiopian government has shut down the cellular service in our area for political reasons, which has meant that we have only been able to get data from the cellular sensors. So we can get a little bit of a hint of some of the other externalities that come into play here. And can you talk a little bit more to the cases where multiple entities, for example, government and GO, donor academia, et cetera, are successfully accessing and utilizing remote monitoring data from a common program area? Yes. I'm going to give you an example of what we have going on now in Kenya. So across northern Kenya, there are thousands of boreholes, but only about a hundred of these have been designated by the county governments as emergency boreholes or ending drought emergency boreholes. These are boreholes that have to work during drought. They used to only be every decade, it's now every year, and about a hundred of these boreholes have to work during the drought, which is the summer. However, they're basically not working any better today than all the rest of the boreholes. They're only working about half the time. We've been able to see this in their data. So what we've done, what we are in the midst of doing, is installing sensors on all of these critical borehole infrastructure, which have been identified by county governments, using that data, providing that data to county governments, water and sanitation companies, and other service providers in these counties. So across five counties, that's about 15 different institutions. And also national government entities, which are the National Drought Management Authority, who provide some budget to county counties for maintaining emergency drought infrastructure, and SAID and UNICEF, and FAO, who all provide money during drought for emergency drought relief. So those are 15 local level institutions, several national institutions, and international donors. We're all using our data now to keep the borehole infrastructure running. And we're running an impact evaluation. Well, we are not. We're the subject of an external third party impact evaluation looking at if this combined effort is actually able to improve access to water during drought coming up this summer. There's not only a humanitarian and security and social benefit to reliable water services, but there's also a direct cost implication. All of these county governments, these national entities, and international donors all pay a lot of money for water trekking to happen when boreholes fail. It's exciting to hear that there's this other impact study that's also assessing your work and demonstrates your commitment to that rigor on both sides of the equation. I'm going to make this the last question if you don't mind, Evan. I know that we were supposed to have a hard stop about five minutes ago, but if you can dovetail that question, which looks at infrastructure at a higher level, the question is, what is the actual effect of water resource sustainability and the scope of IoT from the fast growing infrastructure, for instance, dams, roads and railways. There's an accelerated speed of such instances in Kenya today. Given that this particular listener is in Kenya and seeing these other infrastructure changes, is there a pathway for IoT solutions relative to these infrastructure developments? In your opinion. Obviously, seen from my presentation, our team's focus mostly on water and household entities. We have some discussion around these questions of things like highways. Highways also need maintenance. You can spend a lot of money putting in a highway, but if you don't maintain it, a lot of that investment starts degrading pretty quickly. We're working on a proposal with the Millennium Challenge Corporation with the same question of using accelerometers and remote sensing and other technologies to monitor maintenance and functionality of highways, for example. I would love to work with any of you who are on this call to identify how IoT can help address these challenges and other applications. I just want to say as a closing note, of course, we're technologists and things that our technology evangelists to, but we know that there's only tools. It's no better than a hammer. It's no better than a smart phone. IoT is a way to help monitor the impact of this funding and interventions. And if it's not combined with institutional response and funding that incentivizes services long term, there's no real point in having it that way. I get this question a lot. Who pays for the sensor data? Oh, man, it's going to be $2 a month or $20 a month. We can't afford that. There's really no point in taking data if you're not also budgeting for water point maintenance, which is going to be a lot more expensive on a budget line item basis and is only feasible if you take a little bit of money from installing new infrastructure without this. Apologies, Evan. I had a chop in my feet, so I thought you had completed your sentence, but overall, I think there couldn't be a better note to end on than to advise all of our listeners around budgeting for the long-term sustainability, which includes operation and maintenance in addition to this monitoring. Thank you for your time today. Truly fantastic insights. I would like to thank all of our participants for joining us today and asking fantastic, thoughtful questions. We will have the recording of this presentation available in a few days on our YouTube channel and on our platform for those of you who are interested. Thank you, Evan, for staying a little bit longer. I know you have an urgent meeting and apologies from all of us to those we've kept waiting. For those of you who are seeking professional development hours, the code is listed on this slide that you are seeing now. Feel free to go ahead and apply for those PTHs. If we didn't tackle your questions and they're burning, please feel free to send us an e-mail at webinars at engineeringforchange.org and we will do our best to have Evan respond to those as he is and more about technology-based solutions and the ecosystem in which they live and, of course, to get information on our upcoming webinar next month we will be talking about service provision and encourage all of you to join us. With that, I wish you all a good morning, good evening, good afternoon, whatever you are and I hope to catch you on our next webinar. Bye-bye now.