Latest / The Productivity Podcast with Fexingo: Output, Efficiency, and Long-Term Economic Growth / How One City Cut Water Leaks by 70 Percent With Data
Transcript
- Lucas: You know that feeling when you're paying for a service and a chunk of what you're paying for just disappears before it reaches you? Luna: Like a subscription you forgot to cancel? Or a gym membership you never use? Lucas: Exactly. But scale that up to a city's water supply. In São Paulo, Brazil, before 2016, roughly 40 percent of the treated water pumped into the system never made it to a tap. It just leaked out of aging pipes. Luna: Forty percent. That's almost half the water, just gone. And that's water that's been treated, pumped, and paid for. Lucas: Right. And São Paulo had just gone through a catastrophic drought in 2014 and 2015 — reservoirs dropped to under 5 percent capacity. The city was desperate to conserve water. But the utility, Sabesp, realized they didn't need to build new reservoirs. They needed to stop wasting what they already had. Luna: So they turned to data. What did they actually do? Lucas: They deployed a massive network of smart sensors — acoustic sensors, flow meters, pressure sensors — across 15,000 kilometers of pipe. These sensors send real-time data to a central analytics platform. Machine learning models analyze the data to detect anomalies: a tiny pressure drop here, an unusual flow pattern there, the specific acoustic signature of a leak. Luna: Acoustic signature — so they're literally listening for leaks? Lucas: Yeah. Water escaping a pipe under pressure makes a distinct hissing or whistling sound. Human ears can't pick it up underground, but the sensors can. And the AI can distinguish between a background hum of normal flow and the specific frequency of a leak. They can pinpoint the location within a few meters. Luna: That's incredible. And then they send a crew to dig exactly there, instead of digging up half a street looking for it. Lucas: Exactly. Before this system, leak detection was reactive — someone would report a puddle or a drop in pressure, and crews would spend days finding the break. Now, Sabesp's control room sees a leak alert on a screen within minutes of it starting. They can dispatch a repair team the same day. The average repair time dropped from 45 days to under 24 hours. Luna: Forty-five days to 24 hours. That's productivity improvement on a completely different scale. Lucas: And the results are astonishing. Between 2016 and 2021, Sabesp cut non-revenue water — that's water that's produced but never billed — from 40 percent down to about 12 percent. That's a 70 percent reduction in water loss. They saved enough water annually to supply a city of 3 million people. Luna: So they effectively created a new water source without building a single new dam or pipeline. That's the kind of efficiency story we love on this show. Lucas: Yeah, and the economics are compelling too. The upfront investment in sensors and software was about 300 million reais — roughly 60 million US dollars at the time. But the annual savings from reduced water loss, lower energy costs for pumping, and fewer emergency repairs came to about 200 million reais a year. Payback period was under two years. Luna: And of course, the environmental benefit is huge. Less water wasted means less energy used for treatment and pumping, and less strain on freshwater sources. Lucas: That's the thing about productivity improvements in infrastructure. They often have this triple bottom line: economic, environmental, and social. But I think the broader lesson is about where to look for efficiency gains. A lot of organizations immediately think 'we need more capacity' — more servers, more staff, more pipelines. But sometimes the biggest gains come from stopping the leaks in what you already have. Luna: That's a really good point. And it's not just water systems. You see the same dynamic in energy grids — where reducing transmission losses can be cheaper than building new power plants. Or in software, where fixing bugs and technical debt can be more valuable than adding new features. Lucas: Exactly. The principle is universal. But São Paulo's case is particularly instructive because it was such a large-scale, data-intensive solution applied to an old, analog problem. And it's not a one-off — Sabesp has since expanded the system to cover the entire state of São Paulo, serving 28 million people. Luna: So what's the catch? Every success story has a catch. Lucas: The catch is that it requires a cultural shift within the utility. The old way of working — reactive, manual, siloed — had to change. Workers who were used to digging from a hunch had to trust a screen telling them exactly where to dig. That's not always easy. Sabesp invested heavily in training and change management. Luna: Makes sense. You can have the best data in the world, but if the people on the ground don't use it, it's worthless. Lucas: Right. And the data itself requires maintenance. Sensors fail, batteries die, networks go down. Sabesp had to build a team just to maintain the sensor network. So it's not a set-it-and-forget-it solution. But the ongoing costs are a fraction of the benefits. Luna: I wonder how this model could scale to other cities. Most of the world's water infrastructure is aging — in the US, some pipes are over 100 years old. The American Society of Civil Engineers gives the US water infrastructure a grade of D-minus. Lucas: Yeah, the potential is massive. But there are barriers. Many utilities in the US are fragmented — there are 50,000 separate water systems. They don't have the scale or the budget to deploy this kind of system. São Paulo's utility serves 28 million people — that's a huge customer base to spread the cost over. A small town serving 10,000 people wouldn't have the same economics. Luna: But cloud-based analytics and cheaper sensors could change that. You could have a shared service model where multiple utilities pool resources. Lucas: Exactly. And that's starting to happen. Some startups offer leak detection as a service — they install the sensors, run the analytics, and charge a subscription fee. That lowers the upfront cost significantly. Luna: I love that. Turning infrastructure monitoring into a subscription model. It's like Software as a Service, but for pipes. Lucas: Yeah, it's 'Infrastructure as a Service' in a very literal sense. And that's the kind of productivity innovation that doesn't get enough attention. We talk a lot about software and factories, but the systems we rely on every day — water, power, transport — are where the biggest efficiency gains are hiding. Luna: Before we dive too deep, I want to say something. You know, episodes like this take a lot of research to get right. Lucas, you're the one digging through reports and case studies. And we keep the show ad-free because we think that's the right way to do it. Lucas: Yeah, it's a labor of love. And honestly, listener support is what makes it possible. If walking through the economy with us has made something click, a couple of dollars a month genuinely keeps these going. Buy me a coffee dot com slash fexingo. Luna: It's low-key and it makes a real difference. No pressure, but if you've gotten value out of the show, it's a nice way to give back. Anyway, back to water. Lucas: So one question I have is whether this model can work for other resources. Can we apply the same approach to natural gas pipelines, or to district heating systems? Luna: I actually read about a district heating system in Copenhagen that uses similar acoustic sensors to detect leaks in their hot water pipes. They reduced losses by over 50 percent. Lucas: There you go. The same principle, different application. And I think that's the takeaway for any productivity-minded person: before you invest in new capacity, audit your current system for leaks — literal or metaphorical. Find where your output is disappearing and plug those holes first. Luna: That's a great rule of thumb. And it's applicable beyond infrastructure. In a business context, that could mean looking at customer churn, employee turnover, or inefficiencies in your supply chain. Lucas: Exactly. The São Paulo water story is a case study in that principle. They didn't build a new reservoir. They didn't import water from another region. They just stopped losing what they had. And in doing so, they created a more resilient, more productive system for 28 million people. Luna: And they did it with data, not concrete. That's a pretty powerful message for 2026. Lucas: Yeah. Sometimes the most productive thing you can do is measure what you're losing.