Building Bespoke Weather Forecasts

Published Jul 7, 2022, 4:05 AM

Shimon Elkabetz is the founder and CEO of Tomorrow.io.

His problem: How do you build a weather forecasting company from scratch? The company already sells weather intelligence to companies like JetBlue, Uber and the NFL. 

Their next move: Send the first private constellation of weather satellites to space (without running out of money).

If you’d like to keep up with the most recent news from this and other Pushkin podcasts be sure to subscribe to our email list.

Pushkin. Here's the thing I did not know about weather forecasts until very recently. They basically all come from the government. Sure, you may have your favorite weather app, your favorite TV weather person, but their forecasts are almost entirely driven by data that's collected and analyzed by government agencies. And you know, it makes a certain kind of sense. Gathering the data you need to make a useful forecast has traditionally been a huge expensive undertaking, and having a reliable forecast is really valuable for lots of people in lots of different settings. So it's good that the government does the work and makes forecasts freely available to everybody. But the government is the government, and we shouldn't expect it to tailor forecasts for different businesses, or even to build forecasts that are really useful for people who live in other countries, in countries where the government can't afford to produce its own forecasts. Now, imagine what a private weather company could do. A company that relied not only on government data, but that went out and collected data on its own, A company that came up with forecasts that would not have been possible before. I'm Jacob Goldstein, and this is What's Your Problem, the show where entrepreneurs and engineers talk about how they're going to change the world once they solve a few problems. My guest today is Shimon Alphabets, co founder and CEO of tomorrow dot Io, a private company that plans to put a constellation of weather satellites into orbit in the next couple of years. Shimoon's problem, how do you build a private weather company from scratch? We realized it's from a jew political and cost effective way, and all kind of the only way to solveet is to go to space. Shimon and his co founders launched tomorrow dot Io in twenty sixteen. The company hasn't launched its satellites yet, but it already provides weather related advice for companies like Jet Blue, Uber and the NFL. Even before he thought of founding the company, weather was a big deal for Schimon When he was in his twenties. He was an officer in the Israeli Air Force. Whether it is obviously a huge deal for pilots for planes, and Shimon was constantly on the phone with meteorologists. The solution was, Hey, let's talk to a meteorologist three, four or five times, a day, Hey, what's going to happen here, what's going to happen there? And then I take the data, I analyze what it means for me, what it means for the organization. And I have to do it several times because the weather forecast is constantly changing, and I care about multiple locations and I care about multiple parameters. So that was a very archaic way of addressing challenges at scale. Calling the meteorologist and deciding what each plane should do does not scale. So in that universe you're not at scale yet, not scale, not efficient, not automatic. When there's a human in the loop, there will always be an error. Huh did you make mistakes? Of course everyone makes mistakes. I made a few of them. You know. I have colleagues that unfortunately lost their lives due to whether they did accidents. It was very unfortunate. It's just been there, you know. But I didn't think I'm going to start a company around it. Shimon moved to the US to go to business school, and one day a few of his friends, military veterans who like shimone wanted to start a company, started talking about the weather, and we started talking about past experiences, and everybody We're like, oh you also feel this way, Oh you also experienced that. And when we started looking at it, we said, okay, there is something here. We need to start looking into this. And let's try and understand whether you know how the technology how forecast is being generated, why is it limited in accuracy? And now let's look at how businesses make decisions. Do they do it in the same way we did it in the past, or is there a better way to do it? And what we found out led us to start a company. Well, what did you find out that led you to start a company? All right, that's where it's becoming interesting. So the first thing we found out is that climate change is here. That was in twenty sixteen. It wasn't cool to speak to speak about climate change back then. I mean I think it was cool. I think it was cool to me in twenty sixteen. Trust me, when I spoke to investors back at the time and you spoke about climate change, they were like, give me some sass solution. Don't talk to me about climate change. Okay, fair sass software as a service keep going. So we understood that the problem of managing whatever the challenges are is going to get bigger climate change equals weather events become more frequent and more volatile in any given ear in every part of the world. More hurricanes, more wildfires, more heat waves. Doesn't really matter where you are, there's some extreme phenomen it is going to happen more frequently. Okay, So that's one thing we'll learn. The second thing we'll learn is that the technology of forecasting weather, meaning what's responsible for the accuracy, is generated and dominated by government agencies. And as folks who served in the government for many years, we understood that there must be a way to privatize and innovate faster. And just to give an example, you know, NASA for decades innovated and paved the way to space right, But today you have SpaceX, who's augmenting the capability of a private company doing what the government has done for decades. Yeah, maybe not inventing the rocket from from scratch, but definitely taking all the decades of research and adjusting it to a commercial use case. What we found out is that there is an opportunity to create a SpaceX of weather. So SpaceX built rockets. You want to build weather forecasts. How do you do that? What do you need to make a weather forecast, you need three ingredients. On a very high level. You need observations that describe the atmospheric conditions in real time, the temperature, the wind pressure. Yeah, then you have a good real time description, right. The next thing you need is a model, an equation, a set of equations. Physical models doesn't really matter. The point is that you take the observations and you assimilate them into a model, and then the last thing you need is a computing power on which you process the model. The output of the model is a weather forecast. And what we found out is that there is an industry of weather forecasting companies, you know, big brands. You know, a blue logo, an orange logo, Acy Weather. Right. The point is that these guys are here since the sixties, seventies, maybe eighties. They just repackaged the forecast that the government agency or the government agencies publish every day, every hour whatever. So I'll say, I know, the weather forecast is like a classic thing to complain about. Oh the weather man said it would be Sunday and we had a picnic and it rain. I do feel like weather forecasts are pretty good, and clearly they've gotten better. Was there a particular weakness or failure or set of weaknesses or failures that you really thought you could improve. First of all, you set out to do this. If I may ask, where do you live, I live in New York. I live in New York City. Okay, so you're privileged because I am privileged. I'll be the first to say I'm in Boston. I'm as privileged as you are. Most of the world doesn't have Noah, the US government agency, the big rich country government agencies that do a pretty good job of forecasting exactly. And if you're a private company and you try to provide equally accurate forecast for the rest of the world, you're limited. You cannot provide it. So there is a global problem. So one thing you want to do better is provide better forecasts for people businesses who don't live in rich countries that have big, fancy weather agencies like NOAH in the US. That's one thing. The second thing is that even within the US, you know, the agency's main job is to save people's lives, Okay, it is not to optimize businesses. Right. That also seems reasonable, like absolutely, that's what I want them to be optimized for. Absolutely, we're on the same page here. But with some scientific improvement you can help businesses have better outcome, improve their top line, their bottom line, their safety, their efficiency. So there is a lot of room for improvement. The other element of it is that remember that I said, you know it's it's one thing to handle the forecast. The second thing is that once you improve either on the observation, on the modeling, on the computing power, and you get to more accurate forecasts, there is the are part of it, which is how you make decisions and how do you do it at scale? Right, So this is something that the government almost doesn't address at all, except maybe like when you need to evacuate a city for a hurricane or something, right, very rare circumstances, exactly exactly. Now, I'll give an example. A lot of company like utilities or airlines. They work in a very similar way to the way that I described in my military service. Some guy calling some other guy on the phone and being like, what should we do. You go to a metrologist, You speak to that meteorologist or get a report of road data, and then you have to do a full analysis of what it means and make a decision it's not scalable. Or if I have thousands of trucks driving in the country, or if I have thousands of miles of railways track, or if I have many airplanes in the air and I care about hundreds of airports globally, it's very hard to rely on five or ten even meteorologists on staff. Right. I've heard you say that a mistake you made early on was optimizing for accuracy, and that's really interesting to me, and I want you to tell me what that means. So at the beginning, we thought that if we just create a more accurate forecast, that's it. It's done. Deal with serve like that's a hugely valuable thing, right, even if you're a little bit more accurate, that's worth a ton of money to an airline or the NFL or any any number of really big companies. But what we found out, but we learned that most of the businesses that are impacted by weather do not know what to do with a weather forecast or with a weather data and they need the full loop, the translation to insights and decisions. And that's what helped us design our platform and the way we're operating today. So nobody understands hands how to read a weather forecast. Basically, how do you think from a business perspective? Yeah, so you realize from that that like, providing these people with a better weather forecast isn't actually going to help them solve their problem because they don't because they're not experts in analyzing the meaning of a weather forecast. Yeah. I mean you've named a lot of your clients publicly, right, I mean whatever, Delta and jet Blue and what Uber and the NFL. I guess the NFL isn't playing now, Like what do you telling JetBlue today? Like what do they want to know today? So you know, it's almost summertime. In the summer time, as you know, in the biggest hubs like JFK or Boston Logan, you have disruptions related to lightning strikes thunderstorms. Well, well exactly, they'll shut the airport down for hours and everything will be a total mess exactly. So instead of someone looking at a model and whatever, we're just basically providing a weekly calendar that says, expect the disruptions between dead time to dead time. Here are the recommendations to do. ABC staff more people here, staffless people there. So we actually go into the operational recommendations as a result of the expected disruption as a result of the weather forecast. Now, listen carefully to what I'm saying. It's as a result to the weather forecast. So if you're not relying on an accurate forecast, the business insight is useless. It's actually damaging. So there's no way to get around the need to improve the accuracy. Specifically, what are you better at forecasting than anybody else right now? Precipitation data. We provide global real time and now casting data that is providing a kind of like minute by minute forecast for a range of about six hours on every point on Earth, which is quite useful. For example, you have some sixty minutes minute by minute works that you have on some phones, but it's only in the US and in the UK. Yes, I do find I have that on my phone and I find it's pretty good. Dark Skies. I have Dark Skies on my phone and it's good. But you're saying, if I if I left the US, if I went on vacation to Mexico or something, it just wouldn't work. It's not available. It's just not available. And we created this thing on a global scale with longer time horizon. That's one example. And the other example is like quind, we forecast twind in higher accuracy for the next day, two days, three days, which is very useful for farms. I think, so okay, and are you just better at that because you're more focused on it and you've trained the models more than and it's more important to your clients than it is to say, a government agency, so you have an incentive to figure it out exactly that. I'm sure that if Noah wanted to double down on that, specifically Noah the government agency, they would have been able to do that. But they have no incentive, and you know, the pace of making a decision in a large organization, it's just not enabling them to move fast enough. The next problem Shimona's colleagues are trying to solve, how do you predict the weather for people who live in countries that can't afford a big national weather service like Noah to do that. Tomorrow, dot Io is going to go to space. That's the end of the ads. Now we're going back to the show. Tomorrow. Dot Io's next big project is putting a constellation of weather satellites into space. And there are two big questions I had about that. What problem will it solve and what's it going to take to make it happen. So the first thing I'll say, what motivated us to get to space. The main motivation was how do we optimize forecast and make it more accurate? And when we looked at the blend between okay, we have observations, we have models, we found out that the lack of observations on a global scale are the main reason why we cannot improve whether focussing significantly on a global scale. So the problem wasn't the models. The problem wasn't the computing power. The problem is just there's just not enough data when you get outside of what outside of the US, Europe, Japan, basically the data quality falls off. Yeah, okay. And the most important weather sensor that we identified and I think is agreed on all the community from NOAH to NASA to others, is Doppler radar. A Toppler radar, just to be clear, is it the one where you see a color like if it's raining really hard, it's red or something that stopple? Correct? Okay, Now, radars are looking out in the sky and they help us know where is the training in real time, how the cloud formation looks like. It gives you some kind of three D description of the atmosphere. Okay, now what we found out is that five billion people leave outside of radar coverage. Five billion. He goes out of the border to Mexico, all the way to southern South America, and basically you don't know where it's raining in real time, say for Africa, India and many other places. That is surprising to me. Maybe I'm naive, but like, were you surprised when you learned that, No, because I came from a place where it was not Oh you didn't have it either. You didn't have it either. Yeah, it was pretty broken most of the time. And it's not a new technology, right, it's a decades old. It's not a new technology. But the implication of not having it is huge. You cannot provide flood alerts. Pilots when they fly, for example, to Cancun, they don't know the weather in the route. It's a huge problem for the economy. The next point is that the oceans and the seas are not covered with radars, and every time, for example, a hurricane is formed over the Atlantic. The US government is flying airplanes over the eye of the storm to scan it with a radar so we can send it back to the model that as an understand if it's going to be category one, two or three, when and where it's going to eat, and whether we should evacuate Miami or New Orleans. So rest assured, nobody's flying any airplane over a typhoon or a cyclone in the East. So this is a huge in Asia. In Asia, they're not going out in Asia to get a really accurate forecast of where it's going to go. They can't afford to do that. But you can do it from space. Is that where this is going so exactly, So we realize that from a geopolitical and cost effective way and all kind of the only way to solve it is to go to space. The problem is that radars are pretty big. We actually the world has one radar in space today. It is called the GPM. It's a program by NASA with the collaboration of the Japanese agency. It's more than a billion dollar program that created one radar in space, a very sophisticated one. We have one radar in space today. That radar cost about a billion dollar if not more, and it samples every point on Earth every three days, So it's not very useful for hurricane forecasting because imagine you just sample the hurricanes moving too fast, or general weather forecasting. So what we were trying to do was to say, how can we take this huge radar and minimize it so we can put many of them. But we are a small company, we don't have a billion dollar How can we actually do it in a way that will be cost effective? And the goal is, of course, to monitor every point on Earth with a radar in almost real time, because when you do that, you are going to improve weather forecasting dramatically. You're going to improve hurricane cyclone typhoons, you are going to be able to provide flood alerts for every point on Earth. And it will improve also climate science because now climate scientists will have better understanding of what actually happen. No, I'm sold on why it would be useful. It seems like the hard thing is how do you do it exactly? So how do we do that? The first thing we did was to focus on the sensor. How can we build a sensor that we'll keep most, if not all, the characteristics of the radar. We looked at and how can we make it small enough so we can launch it a not a nano or micro satellite, but something smaller than you know, the stationary satellites, a low orbit. And bottom line, we've finished the development of the radar and in a few months we're going to launch the first satellite out of a constellation of about thirty And our constellation is going to have two types of sensors. One is the radar, the second is a microwave sounder. The combination of the two is going to provide a very good scientific result for every point on Earth. You sound very confident, like are you at a point where you know it's going to work or is it the kind of thing that you hope is going to work. No, we know it's going to work. The question is, okay, will it take us more time? Will we fail in the first lunch? Will we need to reiterate between one lunch to a number? But it is feasible, it is working. It is And how much is it going to cost you to get roughly thirty satellites up and monitoring the weather. Our early estimations, which so far given the inflation, are still are still in the same ballpark. We're looking at around one hundred million dollars for the entire constellation. So that's a big cost reduction. That's compared to what is it saying a billion for an existing one that only does once every three days? Yeah, what are the things that might go wrong? I mean, it seems like a quite hard thing that you're trying to do. I feel like, as you're describing it, it's like, oh, yeah, now, all we got to do is get these thirty satellites up into space and we're going to go But so imagine it's still going to be quite hard and lots of things can go wrong, of course, So you want me to give you examples of things that can go wrong? Yeah, what are you worried about? Okay, the rocket can explode in lunch. Sure. Classic second thing is that you know, we may have some communication malfunction. We may have some when we build our satellites and the radars. We may have to wait for longer than expected for chips to arrive or all kinds of chips. Radiance like supply chain, supply chain, supply chain issues is something pretty big right now. There are so many things that can happen. But are you sure the thing you built is going to work? All the things you've described as like, oh yeah, the rocket could blow up, that's not really our faulter, the chip oncome, that's not really our fault. Like, is it is it at the point where it's like, oh, yes, this will definitely work. Is it like that or is it possible that Okay, it's gonna work. It's gonna work. The question is is it going to be more expensive than we thought, It's going to take longer, and there might be you know, business implications on tomorrow. But it is going to work. It's not a question of science business like like might you run out of money before you can get it going? When you say business, everything can happen in that context. But this thing is working. Pending one hundred million to put a fleet of satellites into space, is it's still a lot for your business? It is a lot. And the market is very bad. It's probably the toughest market in the last twenty years for tech companies. The market for raising funding you mean, yeah, yeah. The investors are not very happy to see businesses that waste money or spend money or invest money, depending on how you look at it to build a solution, and it's definitely a challenge, and I just hope that, you know, the investment community will keep supporting us. We'll get to the lightning round in a minute, but before we do, I just want to say that what Shimone talked about in this episode is actually a really good example of a big idea that came up in an earlier episode of the show. It was the episode where I interviewed the founder of the company Rocket Left, and I was going on about how making rockets and satellites cheaper was a big deal, and he made the point that the big breakthrough is not just that they're cheaper. It's that cheaper rockets and satellites enable people to do big new things, things that just did not get done before. And Shimone's plan tomorrow dot Io's plan is a perfect example of that idea. Even a decade ago, it would have been prohibitively expensive, but today it's possible to put a constellation of satellites into orbit to improve forecasts everywhere on the globe for a price that is affordable for a startup. As long as they can get a few more years of funding, we'll have the lightning round with Shimone in just a minute. Now, let's get back to the show. Okay, I know you have to go soon, but let's do a quick lightning round. What is one piece of advice you'd give to someone trying to solve a hard problem. Focus on the problem and not on a solution. The solution will be obsolete. There are many kinds of solutions, but if you're focused on the problem, you're going to objectively look at what's right, what's wrong, and you'll be able to ditch something that doesn't work and find out something that is better. Focus on the problem. What do you prefer really hot weather or really cold weather? Hot? Okay hot? Could there really be a shark nado like in the movie Shark Nado? I don't know. What's the most underrated weather hazard? Most underrated heatwave? Lots of people die from heatwave annually, strokes, health issues, heart attacks. I actually testified in front of the Congress in the summer of twenty one on this topic, specifically, other than weather, what's the domain where people should use probabilistic thinking more finance, for sure? I mean, how do you manage your investments? Although I feel like bad. Use of probabilistic thinking was a major problem in the run up to the financial crisis of two thousand and eight. I don't know if you remember, but people kept saying like this is a one in ten thousand year move in whatever you know race, Like, clearly it's not your model is wrong? Right, yeah? Yeah, I'm thinking more in a personal level. On a personal level, like a household, how can a household manage their risk and everything? I think they should think about all the scenarios, all the probabilities, and I think people don't do that enough. So whether it's like this classic way to make small talk, you know, when you don't want to talk about work, right, So what do you talk about when you want to make small talk and don't want to talk about work? Oh? Football, I mean soccer? I guess that's the other classic, right sports? Yeah? Yeah, but I'm very passionate about it for real. I mean I can we could talk about it for an hour. Who's your team? What do you say? Who's your club? Who's your club? My club is in Israelity team called mac It's uh, how's doing and learning? Will if everything goes well, what's a problem You'll be trying to solve in five years how to reduce carbon emission with our solution. That will be probably the next step and will be the most impactful thing we can do. But we'll try, okay. Simon Alphabets is the co founder and CEO of tomorrow dot Ido. Today's show was produced by Edith Russelo, edited by Robert Smith, and engineered by Amanda ka Wong. I'm Jacob Goldstein, and I'll be back next week with another episode of What's Your Problem.

What's Your Problem?

Every week on What’s Your Problem, entrepreneurs and engineers talk about the future they’re trying  
Social links
Follow podcast
Recent clips
Browse 143 clip(s)