Overcoming the Market (Failure), with Owen Barder of PxD – Part 2

This episode is the followup to Part 1, which you can find here. 

Transcript

Jim Fruchterman [00:00]

Welcome to Tech Matters, a biweekly podcast about digital technology and social entrepreneurship. I’m your host, Jim Fruchterman. Over the course of this series, I’ll be talking to some amazing social change leaders about how they’re using tech to help tackle the wicked problems of the world. 

We’ll also learn from them about what it means to be a tech social entrepreneur, how to build a great tech team, exit strategies, the ethical use of data, finding money of course, and finally, making sure that when you’re designing software, you’re putting people first. 

This is episode two of season two, the second half of my conversation with the CEO of Precision Development, Owen Barder. 

Owen Barder [00:43]

Precision Development is a social enterprise that provides information to smallholder farmers through their mobile phones, either through voice messages or text messages. We work in 11 developing countries, low-income countries, and the aim is to increase the productivity of farmers to close the yield gap. But we’re not only focused on agriculture, we’ve been piloting an education program, we’ve done some work in nutrition with good effect, and we’re looking at climate change and environmental services. 

Jim Fruchterman [1:22]

As I noted in the prior episode, we covered so many big picture issues in tech for good, the reason it needs to exist, market failure, exit options, data scientists, that we decided to break this episode into two parts. 

So just a recap of Episode 1’s key points. Owen talked about his journey, starting with being an economist inside the UK government and being seconded to South Africa to work in Nelson Mandela’s first government, helping them set up the same kind of treasury functions Owen had done in the UK.

Owen Barder [1:55] 

This is the Mandela government, this was at a time when we all wanted the Rainbow Nation to succeed. And, you know, I suppose I took about half a breath before I said yes. 

Jim Fruchterman [2:07]

And his work as economic advisor in #10 Downing Street, working for Prime Minister Tony Blair. 

Owen Barder [2:14]

There’s that Hamilton song about being in the room. It was obviously a fascinating time for me because I was in the room. Again, I was supporting others who were making the decisions. But it’s an immense privilege to see how decisions are made and have some influence over them.

Jim Fruchterman [2:32] 

We also discussed the issue of market failure, the work of Nobel Prize winner Michael Kremer and his analysis that says public goods, like a malaria vaccine that helps the poorest of the poor, don’t tend to get done. 

He also discussed how he worked with Kremer and others to develop the Advanced Market Commitment, the idea that governments and philanthropy would promise pharmaceutical companies to buy their product if they built an effective vaccine and how we worked on the very first one of these. 

Owen Barder [3:03]

We persuaded a group of countries and the Gates Foundation to come together and make a promise, not for malaria as it turned out, but for pneumococcal disease, that if somebody developed a version of a vaccine that was for pneumococcal disease that was good for the developing country versions, the strains of pneumococcal disease across the developing world, that that group of governments would buy one and a half billion dollars’ worth of that vaccine. 

That idea, that policy has saved somewhere north of 7 million lives so far, probably nearer 20 million. This is an example of something where there’s a big upfront cost of developing the thing and then very, very low cost per user of selling it or delivering it to millions or hundreds of millions of people. 

Jim Fruchterman [3:53]

We also talked about the necessity of the government exit option, because the sustainability model for something like agricultural extension, which is what Precision Development primarily offers, that’s been the province of government in the rich world for a very long time and also in the developing world. So if he and Precision Development can find a way to make ag extension a hundred times more cost effective, that’s a very sensible approach for government to adopt. 

So let’s resume my interview with Owen Barder, discussing where precision development is today. 

Owen Barder [4:27]

We’re currently serving five and a half million smallholder farmers in 11 countries, which makes us one of the bigger agricultural extension services. We have a cost per farmer, if you take all our costs and divide by all our farmers — and these costs vary from service to service — but we’re down at around $1.25 to $1.50 per farmer per year. So that’s pretty cheap. It’s sort of two orders of magnitude less than a classic, a bit more than two orders of magnitude less than a classic in-person agricultural extension service. 

One of the projects, and approaches that we’re especially proud, of is we’ve just handed over a 2 million farmer service to the government of Odisha in India, in the East of India, where it was a Build, Operate and Transfer Agreement, where the Gates Foundation, to my point about financing models, funded the setup costs, our costs of getting it going. The government of Odisha funded from day one the operating costs of running this service. 

Jim Fruchterman [5:43]

Now wait a minute, isn’t Odisha like one of the poorest states in India? 

Owen Barder [5:46]

It absolutely is, very agricultural. This is tens of millions of extremely poor farmers. Agriculture extension is run by state government rather than national government in the federal government in India. But it’s worthwhile for them because this is such a huge part of their state’s income, the income of their citizens, to spend a modest amount of money paying for a digital extension service for their farmers. 

We set that up with the government of Odisha, trained their staff, built the system, and a couple of months ago they took it over to run it. 

Another way in which we generate impact is in a different country in Ethiopia. We’ve worked with what was then called the Agricultural Transformation Agency, it’s just changed its name, to help them run their service. This was a service that they had set up and which wasn’t having as big an effect as they would have liked, by which I mean really wasn’t working at all. 

We brought in data science expertise and the fact that we were running services elsewhere, so we’d thought a bit about how these services might work, and we acted as a consultancy to them to enable them to improve their service. 

This wasn’t a Build, Operate, and Transfer, it was a pure B2G consulting service that enabled them to- 

Jim Fruchterman [7:23]

Business to Government, okay, now I’m guessing that’s what B2G is. 

Owen Barder [7:28]

That’s exactly right. The issue here was that this was a service, these services work well when they’re customized, when you’re getting farmers specifically to information that’s relevant to them. To do that, you need to know quite a lot about them. You need to know where they are, what they’re growing, whether they’re using rain-fed agriculture or irrigation, what kind of- kind of seeds they’re using and so on. 

Jim Fruchterman [7:47]

… cultural norms for what you can and can’t do. Yeah. 

Owen Barder [7:50]

For example, we had an experience where we were sending voice messages to a group of farmers. I’m not going to name the country, but this could be several countries, where we were finding that a listening rate for women was lower than for men. Data can tell you that. It doesn’t tell you why. So we went and asked them why. And the answer was, in some cases, not all cases, but in some cases, it was, I’m a married woman. And this was a man I don’t know talking to me on my phone. I mean, it was a recorded message, by the way. It was a man talking to me on my phone. It’s not right for me to be listening to a man I don’t know on the phone. So I hang up. 

Jim Fruchterman [8:33]

Something you might not have guessed would be the reason for losing people. 

Owen Barder [8:38]

Right, so we see there’s an asymmetric listening rate between men and women. And then we go and, you know, like any good marketing company or product company or tech company would do, we go and try and find out why. And it turns out both to be completely understandable and pretty simple to fix. All right, then we start recording the messages for women in women’s voices. 

It turns out that men don’t have the same qualms about listening to the whole of a message from a woman. But it isn’t clear in some countries that men respect the advice they get from a woman’s voice as much as they do from a man’s voice. And you can experiment with things like, what if the person says, “I am a government scientist and you should do X, Y, and Z?” Does that give you a better adoption rate or a lower adoption rate? What if you say, “I’m from the UK government?” 

Jim Fruchterman [9:36]

So one of the things that I think is a breakthrough from what I’m hearing is we are used to the RCT kind of model, right? That there’s a malaria vaccine and it either works or it doesn’t. And once it is proven to work, you can scale it up. And what you’re describing is the equivalent of RCTs, but it sounds like you’re running a whole bunch of them. And I’m guessing a lot on faster timetables than traditional RCTs. 

Owen Barder [10:02]

So we aspire to be doing what tech companies do, to be agile, to use A-B tests to try things. We are, you know, our internal conversation is about using the appropriate level of rigor. So we would not go to a government and say you should invest tens or hundreds of millions in this program until we had some pretty rigorous, large scale, you know, that’s much more RCT territory, right? Just as you wouldn’t give regulatory approval to a vaccine until you’ve done a large scale clinical trial on it, right? If it’s a big stakes decision, you want lots of evidence. If it’s a relatively low stakes decision, where you’re just trying to figure it out is a man’s voice better than a woman’s voice, and you try them both and you keep course-adjusting based on that, you can do that pretty quickly. And you know if it turns out, down the road, that actually you’ve made a decision on not quite enough data, and your confidence interval was in retrospect bigger than you realized, then you can always switch it back, right? You’re not stuck with something that is set in stone forever.

So depending what the decision is, you need to use the the right kind of of evidential tool for making the decision. What you shouldn’t do is say it’s all or nothing, right, either you have to do a big RCT or you’re stuck with the thing you first thought of. That’s an absolute disastrous place to be. But i don’t think we are yet great at being as agile as we would in Facebook, say. I mean I have friends in Facebook who can change the configuration of something they’re working on, tell the system, run an AB test on this, and they, you know, go off to their rather fancy campus, come back afterwards, and they’ve got some results from an AB test, you know, from tens of thousands of users about which worked better, right? 

We haven’t got to that level of automation. And it’s partly because too many of the questions we ask are a bit different from each other. It’s not like Facebook, where you basically have a web page and all you’re doing is making changes to a web page. And the thing you care about is, how long did the user spend on the web page, right? 

Jim Fruchterman [12:19]

Whether they click the ad or not. Pretty limited number of outcomes that are interesting. 

Owen Barder [12:24]

Right. And it’s just much harder if you’re, you know, in our world. But I think we could go a lot, you know, I would like us, and I think we all want to go further in automating and reducing the cost of and so, you know, making more Agile and reducing the cycle of it, reducing the iteration time to make those things better. 

But it’s not because we’re sort of, conceptually, we don’t want to do that, we don’t want to make choices. You know, conceptually, we’re in the right place. We just haven’t yet got enough of the tools to make that work really well. 

Jim Fruchterman [13:00]

Well, but I mean, I think you’re apologizing for the fact that market failure exists when you want to serve 90% of humanity. And the fact that you’re not having the same level of staff investment and activity that a Facebook or a Google has is like, well, okay. But something tells me you guys are head and shoulders above the status quo in the social good field. And so I wouldn’t quite apologize as much as you just were. 

Owen Barder [13:25]

I’m British, we apologize, right? 

Jim Fruchterman [13:27]

I’m so sorry that I’m only helping 5 million farmers instead of 15. Okay. All right. Oh, please make good on that. 

Owen Barder [13:35]

But part of our philosophy is that we’re always learning, right? And we’re learning driven by evidence and data and we have some ways to go. One of the things that we’re learning, by the way, is that there’s more to do than simply providing up-to-date and accurate information to farmers; that many farmers face constraints that are not just to do with things that they don’t yet know, but if they did know, would make them better farmers. We can advise people to buy a better seed, a drought-resistant seed, for example. There’ll be some farmers that don’t know that those exist, don’t know where to get them, and we can advise them on that. 

But there’ll be some farmers who have good reasons not to buy a drought-resistant seed, and that’s the reason why they’re not buying it. That might be that their local agro dealer doesn’t sell them, or it might be that the yield from those seeds is a little bit less. Or they think the yield from those seeds is a little bit less in a normal year, even though in a drought year, that is much higher. Or it might be that the cost of shipping those seeds to a remote area for a single farmer is too high. 

So we’ve begun to think beyond the, you know, as it were, a sort of curriculum of things that farmers should know. And again, thinking a bit like a tech company, starting with, well, what is stopping the farmer from doing this thing that looks on paper like it ought to be a good choice for them? Rather than sending the message saying, “Do this, do this thing”, right? Go and figure out whether it’s a good choice for them. And if it is a good choice and they’re not doing it, what is it exactly that’s stopping them from doing it? And then see if you can solve that problem. And one of the things that we notice is that farmers are operating under more uncertainty than they need to operate under. And there are sort of three obvious categories of uncertainty. 

One is weather. You know, this is not stuff they could have learned from their father and their father’s father, right? What tomorrow, what this week’s weather is going to be. This is stuff you, you know. 

Jim Fruchterman [15:52]

But for an American farmer, there’s a bunch of companies that are happy to sell you that data so that you can decide when, you know. So the rich world farmer has that kind of intelligence, but that’s not the average farmer. 

Owen Barder [16:03]

You can’t make money selling weather data to smallholder farmers in developing countries. People have tried, there just is no business model. And so that’s one thing. The second thing is about pests and diseases, right? That if there’s an outbreak of a particular pest, it would be good to know that quite quickly and to know what is the right pesticide or the right thing to do to protect your crops. And this is something that we do do some of. And our farmers absolutely love it. When we ask our farmers what would help you, weather and pests are the two things that come 1 and 2 every time. 

The third thing is market conditions. Where could I sell my produce? Where can I buy an input that I need most cheaply and least likely to be counterfeit? 

Those three types of real-time information about what’s going on, about weather, about pests, about market conditions, are things that, if you could reduce the frictional cost to a farmer of knowing that stuff, they could make profit-maximizing choices much more easily. They’re much more likely to be able to sow at the right time, harvest at the right time, put fertilizer down at the right time, get good fertilizer at a low price. It isn’t just about telling farmers “this seed is better than that seed, switch to this seed”. It’s about giving them up-to-date information that enables them to optimize as they’re going through the growing season. 

Jim Fruchterman [17:41]

So, Owen, you described how the power of data can benefit the individual farmer to make better decisions that get them the information they needed. But the other thing that comes to mind is, especially with your background as a policymaker, I somehow think that you are able to know a lot more about the farmer in a given country that would be really interesting to policymakers. 

I mean, people make policy all the time with terrible information. So, have you had any governments go, hey, could you tell us more about if we shifted this one agricultural policy, what might happen? Do you actually have that policy kind of conversation with them? 

Owen Barder [18:18]

So, we don’t yet. There are some conversations around, you know, how would you predict production each year? You know, if we had enough information from enough farmers about who’s growing which crops, you know, what are the import needs going to be of the country, you know, what’s wheat production and maybe information that we could produce during the year that would help the government anticipate what the market conditions are going to be like and maybe how they need to respond to that. 

The other kind of information that we’re already generating for governments is about pests and diseases, because in some places we have call centers where people can call in and ask for advice. We can use that to identify if there’s a disease outbreak. 

Jim Fruchterman [19:08]

And now it’s a prevention being worth a pound of cure kind of thing. 

Owen Barder [19:11]

Exactly. So, we do do some things like that. There is a holy grail here, which is if you had enough information about every farm, every farmer, their soil type, what seeds they’re using, what fertilizer they’re using, the weather, their behavior, and you could measure their yield remotely, then we might be able to get some big data insights into which agricultural practices work best in which circumstances. 

And we don’t. That big data doesn’t exist in the developing world. We have the benefit of agricultural research and research institutes that run crop trials, but we’re not far from a world in which there’s enough data to check whether the advice we’re giving and see whether there’s heterogeneity. 

We’re not just interested in average effects, we’re interested in, for each farmer, what do they need to do in all their circumstances to optimize their production and to optimize their well-being. We have one trial where we do have plot data, where we’re using satellite data to see which farmers are underperforming relative to their neighbors. And then we target that farmer for advice. 

Jim Fruchterman [20:31]

Well, this is the whole idea of mass customization. And in the past, it hasn’t been possible to do that with farmers, but it sounds like you’re already trialing some of these capabilities and meshing geospatial data with, “How do I help this individual farmer based on the observables?” That’s incredible. 

Owen Barder [20:56]

And there’s huge potential. We are one part of an ecosystem of researchers and agricultural companies and data and technology companies. And what we’re trying to do is make sure that this human knowledge is applied to the problems that face the world’s poorest people, because somehow we’ve set up the world not to do that. We’ve set up the world to find cures for baldness and not to design vaccines for malaria. And we’ve set up the world to help a farmer with detailed advice in Ohio, but not give basic weather information to a farmer in Ghana. 

And that makes no sense, right? These are the people who would benefit most from this information. And the technology now exists to get them that information at no cost to the world. That’s our mission. And if there’s anybody listening to this who would like to do this or help us do this, we have no ambition for our own growth. We’d love it if the world served this problem, and we could go and turn our attention to the next most important problem. But for now, this seems like a really important solvable problem that the world could get its arms around, and it could make a difference to hundreds of millions of people’s lives. 

Jim Fruchterman [22:23]

And congratulations to you and your team for building, essentially, a digital public asset that makes these things possible, and with the investment of not billions of dollars, but millions of dollars in helping millions of farmers. 

So I think this has been an outstanding story and Owen during our conversation I think we’ve hit a lot of the lessons that I I thought you had to teach about the power of how to use data to this, you know, larger impact; how to do AB testing; how to work with governments because they’re essential players in so many of these things; and and that the highest and best use of our efforts is to make other people able to do these things without our help. And that’s incredible. 

So do you have any final words to to our listeners about, if they are aspiring to work in applying technology for social good, is there anything else that you’d like them to keep in mind as they go out there and try to do things that are as exciting as you’ve had the privilege of working on for the last few years? 

Owen Barder [23:28]

Well, as you say, it’s a privilege and I absolutely don’t want patronize people. 

One thing I’ve learned from my team and from the work we do is the importance of listening to the end user and spending time understanding what it is that they are constrained by, what problems they are trying to solve, what it is they’re optimizing for, and not assuming that we know all of that, not assuming… There is a real tendency for us to sit on the West Coast or in London and think we know what the problem is and what service people need. Very often we don’t. I think this is something that the best companies in the world do really well. Because if they don’t, they go bust, right? They don’t understand their customer, they will eventually fail. It’s not something that government services always do well, and it’s not something that all nonprofits do well. 

Sometimes people who switch from the private sector to philanthropy switch off that humility about listening to the customer. Please don’t do that. I think that’s what we all need to do. We need to listen to use data and evidence to understand your user. But it’s that, it’s iterating, it’s building prototypes, seeing what works, seeing if it solves problems, seeing what the unexpected consequences are, understanding what the problems you don’t foresee are, and then building out from that and recognizing that every community, every problem is different constraints and different needs. 

I’m very proud that nearly all our staff are out in, nationals of and working in their own country. I think that’s really essential to provide an effective service at scale. 

Jim Fruchterman [25:40]

And I think you guys have built the sort of technology that makes easy to do mass listening and mass customization to the needs of the individual rather than one size fits all, which we know isn’t optimal if you have any other option to work on. 

Owen, this has been terrific. I think this is a very exciting story, I think more people need to know about what Precision Development is doing. And I think the world needs another thousand people like you and your team, organizations like yours, out there solving these kinds of social problems, using the same kind of philosophy and technology savvy to really tackle systems level problems. So thank you very much!

Owen Barder [26:18]

Jim, thanks so much for having me. 

—

Jim Fruchterman [26:23]

To hear more interviews like this one, be sure to follow the Tech Matters podcast on Spotify, Apple Podcasts, your favorite platform, where we’ll be publishing new episodes every two weeks. 

For transcripts and related articles, check out our website at techmatters.org. 

And is there something that you found particularly insightful about this episode? Anything maybe you disagreed with? Something you want to know? Be sure to let us know by sending an email to podcast@techmatters .org. We’ll pull together all of the thoughts from our listeners and try to address them before the end of the season. Thank you so much for listening, and see you next time. 

—

As we wrap up this episode, and every episode, of the Tech Matters podcast, I want to thank the following funders for their generous support of Tech Matters, our nonprofit, noting that the content of this podcast is of course solely the responsibility of me and Tech Matters:The Patrick J. McGovern Foundation, Okta for Good, Schmidt Futures, The Skoll Foundation, and Splunk.