Graph for Good: Empowering your NGO with Graph Technology

 

If you are more of a visual type, you can also watch the presentation that Mario gave at CDL 2019.

 

 

 

— Transcript —

 

Welcome to the Connected Data London podcast. Brought to you by the connected data London team. Connected data London is the leading conference for those who use the relationships, meaning and context and data to achieve great things. We have been collecting data people and ideas since 2016. We focus on knowledge graphs, linked data and semantic technology, graphs, databases, A.I. and machine learning, technology use cases and educational resources. May graph technology improved the deployment of humanitarian projects? The goal of using what we call graphs for good and action against hunger is to be more efficient and transparent, and this can have a crucial impact on people’s lives.

 

Is there common behavior factors between different projects and elements of different resources or projects be related? For example, security incidents in a city could influence the way other projects run in there. The explained use case data comes from a project called Keit for autonomous cash transfer and humanitarian emergencies Catch, whose goal is to deploy electronic cash transfers in emergency situations when no suitable infrastructure is available. It also offers the opportunity to track transactions in order to better recognize crisis affected population behaviors, understanding goods distribution network to improve recommendations, identifying the role of culture in transactional patterns as well as most required items for every place.

 

Mario Bastande

 

Hello everyone. Thank you for coming and thank you to the organization for inviting us to explain our work. And those are specific to this graph theory project. My name is Mario Astande. You can see here I’m in charge of data and action against hunger and also the gas and water project area where this plane that’s later how this works. And we’ll be talking about graphs for good in general and how we applied it in our work.

 

So this is the agenda that we were talking about. What are we doing in this organization and also what kind of approach we have in terms of graphs and what all this pilot project with guys and our production project that is more focused on the finance of part financial execution and then like good practices, if you want to do the same, the same process and the same kind of process. This is the countries where we are working right now.

 

We have five excuse distributed around the world and we read all these all these countries in the world. That depends on my belief that our crises there, that will be natural disasters or or Maine or some other crises made by men and whether we are all around the world and we have and know little about context everywhere. This is the kind of place we we do a in the next set against Hungary, Spain. We have different kind of projects, but mainly nutrition, livelihoods, water and sanitation.

But we also give support in terms of shelter and protection, access and different kind of of issues. I’m going to play a little bit how our team works in terms of data and in in the HQ, we have better managers and scientists. We all work together with the media department that is monitoring and evaluation, accountability and learning, business analytics and ICT, ICT team. So we are like the and they’re more focused on data than we have that one analyst and they are in their bases in their countries and we give them support and we all work together for the end user support application application users that are our audience in terms of products as we are an NGO and some of you may come from the private sector and is quite similar in some ways, but especially in the NGOs, is quite difficult to work with data.

 

So that’s why our history is always trying to take us to the to that difficult point of going to data because of funding, because of measurement. We talk about it later. But this is the situation. We are now trying to get into the data analysis and that’s our approach.

 

So why we go for graph theory? Well, I have some background working with graph databases. I was working in a previous organization called DOCS, working with documentation, human rights documentation somehow connected. We can talk about this later once once we finish the presentation. But this is an example of graph we have in the organization. We have the country, the Palestinian territory, also some bases, some offices we have there, like Guiseppe. And this is an example of a contract we have.

 

This is a very simple example, but this is the kind of graph theory we want to apply depending on the knowledge based, of course, in the expertize. We have really good context of the data and we know what is happening with this data, but we need to represent these in a way to prove and to be content in terms of what we are going to do. Uh, we have all so many factors involved in our issue. So we want to mix different sources and mix different databases.

We have these contexts. We have these qualitative analysis as well. That is not only the data we put in the graphs, but is some knowledge that we have in our mind.

 

But we are not able to represent the demographic always in that decision making focus. And what I wanted to get with this project was to give the user the freedom to explore the graph themselves, because we don’t have time to be answering their questions every time. So we want them to be able to do it themselves. So this is the main reason I wanted to to use graphs how we made all of these.

 

It was easier in terms of we have some data sources that mainly coming from exercise, but we also have some information, internal information systems that is very basic. So we convert it to that model. And we went to the graph.

 

OK, um, in this case, we have the context of that time, although so we know how they are related. It’s not complicated and we have this knowledge beforehand. So different sources from operations, finance, look, logistics, security, all of then we makes them by using some ETAs like transformation processes to get these in these kind of graphs and other tools in visualization. The learning curve was hard, as you imagine. We don’t have expert.

And these we were only two people working on that. So it was quite hard to to get into the into the appropriate. But we finally got it for a pilot use case.

As I told you, I’m in charge of these distribution projects. Gotcha. Is is the need for autonomous cars solutions. And we may be at smaller banks in remote areas where no truck transactions are possible. So they don’t they don’t have and they don’t have access to to the money. They don’t have access to to cash. So we go there because now is like a new training for humanitarian action that you don’t go to some place with rice or with shelter and you give it to them and you try to empower these people, not only the affected people, but also the market system itself.

 

And you try to empower them by giving them guys or by hiring them as vendors, for example. OK, so this is a new trend in humanitarian action. This is why the structure that way.

 

So that’s why I that I started with that, because I didn’t want it to fail slowly. And we have Maylee relations among people and I’m on track transactions. So how people least like a bank, OK, so how people buy things and how people say things, we see that, that and whether or now this is data specially for Lebanon, for Lebanon. But we did it for Mali as well.

 

So this is the model in our minds. Some main points we have the people like AFET affected people and the ones we hired as have vendors. So they will have they are like street shops or regular shops, different kind of vendors and of course, the people. So transactions are also important in this model. And then we have some other factors.

 

This is how it looks in that graph database, a little bit more complex because we wanted to have everything related. But it’s all about relationships is exactly the same thing we saw before. So we started to make some questions to this database. And we got these kind of results, for example, how people in general like vendors or people affected move. And that’s very important for us in terms of some security. So happens in one place. So why they move, you have to anticipate that and why people, not only vendors, but the affected people, how they move because of some reason.

 

So you you catch this and you go to to to get the reason what is happening. For example, we also have these kind of of results that is taking the vendors, the people and where they are doing different transactions. So this is giving us information about where they buy things on the way up. And also what’s very important for for us, the fraud, because they normally spend all the money at the beginning, like we give that we put the money in there, in the card, in the credit card, and they spend this money in the first few days.

And we wanted to understand why. We wanted to understand if it was fraud, if they were getting cash out to the money instead of buying the the the items. So these were so interesting what items the people get if they have access to these two, the commodities to things they want to have access will see how how to get to that. Some recommendations later, but also like mainly interested in the profiles of these people, like depending on the gender, how they buy things, depending on their gender, depending on the nationality.

And you equal even a group people, uh, some people, groups of people and understand why they go to Dubai, for example. So you get closer to the to the affected people, things to the surface. But algorithm, we were able to recommend some people, uh, to go to different places to buy items that they don’t have the area. So this was really, really useful. So it’s a very simple algorithm that you apply automatically.

And while he’s also talking about the availability of these of different items, if you can recommend to vendors to buy near Whitlams new commodities, but we got some recommendations to use them, not like the ones we are used to with Google and so on, but something similar.

And of course, how people relate to these groups we make with people, how people relate to each other. We we got to know this. These users already, they have some problems, some security problems to go to the benders. Why? And these questions and of course, in terms of that quality, we realized we had some problems. We have a lot of problems bringing in in the part of quality that we were we were solving later on.

So this is the part of, gosh, this was a pilot project. But then we said, OK, this is working and we can answer some questions. So why not go into some more specific question that was much more important for us. And we went to the financial execution and financial execution was a problem for us. And we tried to to put it in an aircraft, in a model and.

Yes, to try to solve it. So what’s going on with the rest of the organization was we try to mix all of them, different sources coming from different databases. And we got we got some results about it.

This is a model we thought was going to be use a very simple model. I don’t want to go into into the details, but you have here information about operations, finance, logistics, because there are different reasons for this financial execution to not be done in a proper way. So we already identified that it was easier for the data, more that we went find where the problems were coming from. So we decided to go for this. That’s a model that’s later on.

And there was a little bit more complex and we realized we could get some data from activities like different stuff. So it’s not always the idea you have in your mind. Then when you go to practice, of course, you find different, different issues.

 

So what are some of the results about these in these projects already in production and already used by different end users?

 

So, for example, we were able to identify the and the contracts that were not going in a good way at the end of the contract, which is like the main point. If we are finishing some contracts, we need to go to that contract and take a take attention on that. And we found out where and why. And all these reasons and factors that I was telling before, like depending on different factors, have really identified by the users, by the team.

 

We were able to answer these these questions. For example, they don’t delay in the signature of this of this contract was something we knew at that ERISA and a factor to be late and to not go for the financial execution in a proper way. So we were able to be identified by this by this factor. What are the main donors creating this these delay? And we prove this way not only in the minds of the team, but in a way that is represented on this story.

 

Magrath so we had some other conclusions that were not we had the hypothesis that the partnership was going to be another factor, that kind of this financial execution. But we got that is not that reason for for this problem for us. So not always. You can prove what the team. No, but you can always you can some like remove cancel some hypothesis that they made beforehand, OK?

 

And we finally solve for the the financial execution. And this is our production product. Then the users have available and they. All these later, thanks to to the to to the project. So what are the next steps in terms of how will you continue with this with this project? So, of course, how commoditization of these meters is the transformation of data. That was really interesting for us weighting the relationships, because not all all of them are so important for for the team.

So then we can have some, like optimization optimization algorithms that you can maximize or minimize the impact. Of course, some other systems like for this contrast we we saw before. So something is going wrong. So get and get an email about it. This is something, the easy things to a lot of visualization systems we got at team really collaborating on this tool. I’m getting insights about it, which was my main main objective. But it of course comes with a lot of training, with a lot of time like spending time with them because they don’t have the technical knowledge.

 

We would like to get some predictions. We admit at that point we are really improving the quality of things to go for the technology and we’ll go to do some or their problems with our financial solution. Um, OK, their conclusion. Yeah, OK, Sam, conclusions like we went only for the research focused, we went to go to to answer some questions that said, of course, we improve decision making. And for me, the democratization of data was the most important point.

But of course, you have this finding on time and lack of vision to long term from the Leykis Supercheap in the in the NGOs or so in the private sector, but especially with actually in our sector. So these are the two main problems you can find when you go for these kind of projects. So thank you. I don’t know if we have time for questions. Yeah, OK, perfect. Thanks. Yeah. So I is not see any kind of game.

 

Question from audience

 

Issue. OK, how to use the system to attempt to define. Why don’t we try so that’s me, because this is our little environments, such as a functional environments, also use their ability to resolve the conflict is actually just a pilot program in the U.S. so that we can be better by targeting walls and the ceiling. So the amount of cash kind of on is based on the user consumption of a second chance. First question. The second question is how big is this device?

Is there a way that we can actually recycle relational understanding of the process of Cubavision side, where see decisions on the basis that are necessary and allowed to monetize as the issue makes sense that while, you know, some of the emergency seconds is the way, I mean, that’s used to be quicker. It’s just kind of a broader question. Yeah, okay. For the first one, there are different things over there like that. Technology is the first.

 

We cannot rely on the information systems we have. And it also happens in all the organizations like you have no data or you have data that is cannot prove anything. So this is the first one. The second one is that record of all these data. So if we cannot go for predictions, if we don’t have data from a few years before. So that’s the second thing also related with the second question like this lack of beachum. I’ve long term, so this is it seems possible to have a team of two people doing that, one technical and one with more with activity approach and to be doing developing this kind of brilliance without the support from the direction, without funding, without all this stuff.

 

Mario Bastande

 

So for that was really struggling to slow to train this team that was working with this application. And you will find this you will face this guys in this kind of living in the country offices. But this is something that will be that. And first, you need training. You need, of course, this second factor I said like that long term vision. So they don’t have time and they don’t have time. So if nobody pushed them to get this and they will not care about it because they have enough enough problems, enough work.

 

And that’s also, as you know, very difficult question in an NGO sector. Yeah, I answer.

 

Question from audience

 

I’m sure that one of the biggest in the project, but also is in some ways, yes. So I would like to ask you, where is the source of truth in this instance, if everyone is of what happened to your face? How do I start doing that? Yeah, you could use the database to go to Switzerland.

Saw the need for one of the girls, you know, very serious asset classes that have.

 

Mario Bastande

Yeah, yes, we only used one one graph database and then we use because they have, like I use for that, some diagrams, open source tools, I also graph databases and visual visualization systems. So this is why they have different the same. But no, we only had one graph database that was in the top of the school system. Database is normal and ones and yeah, that was it.

 

By the way, though, we don’t have different graphs and we actually have these these Gasparilla at the graph in one place and then the financial exclusion in another. OK, so the next step is how to mix them and for example, understanding not only because of their donors delay or some other reasons, but if security agency then happens in in Lebanon, in the place where where the gas prices are running and how to connect these with the financial inclusion and understanding why this contract easily, for example, OK, or why we don’t get to the finance and can afford that concrete contract.

 

Yeah. I mean,

 

Question from audience

 

which is what the users you look like and how.

 

Well, yeah, good question. So this is something that’s always useful to your team, to your organization. They really love it. Like, I want to do this thing I want to be able to do. I know that it’s really difficult to reach somebody that is totally able to to do that in terms of time and availability. So they are not taking at all and they know nothing about technology. You have to go with them step by step and really, really, really slow.

 

And also because everyone I mean everyone, you and me, we’re used to relational databases. We are used to have tables and that’s it. So when you come to these, you don’t understand what he said. We’re up. But then when they start and they get used to that, it’s easier for them because he said, I’m more graphic way to do, to to do and to see the things.

 

OK, so they are mainly working and they have a lot of field experience. So that’s why I was really like a fun seeing them exploring the graph because they know what is next. So they already have this knowledge in their minds and this is far too early to show them the result of that and to prove in a discreet way that that’s all they were thinking about is true. OK, yes or no? Yes.

 

Question from audienceSo your problem is to just have kind of obligation right over here on a daily basis?

 

Mario Bastande

 

Yeah. OK, so there and we have the graph database. That’s for that you need to a lot of coding knowledge. So we were we started to do these questions with this code, but then we, uh, we send it to visualization system that is more like you can ask, you can make some queries without knowing any code. OK, and you show them how to do that. Yeah, exactly. Exactly. You show them how to get the first question, the global one, that you get one or two nodes and then they go just for the different reasons, because these these donors delay these partnership reasons to be late in execution were different notes.

 

OK, so where they were nodes connected to that contract and to the places. So if you’ll go straight to that. No. And expanded, you already know what is what is the program. Yeah,

Question from audience

 

I was curious about the you mentioned the land is basically the process of getting to. And I was going to give you an elaborate on how it relates to this. So a group of about what am going to in the U.S. around the world and you U.S. moves to the database and instead of competing and the U.S. right now is just here to talk to you and Jim.

 

Mario Bastande

 

OK, so I knew this this question was going to pop up. So I’m going to say this like we use a sequel then for the transformation that was in doubt. And this transformation to all of them are open source. I mean, not a sequel, but then we use new euphoria and curious. OK, so this process I already did this process before in that organization, this human rights organization. I told you so. It was not so difficult.

 

But in this case, we had a lot of data. And we are not talking about big data, to be honest, but it really takes a lot of time. And we we spend a lot of time in coding in Benzal how to get this fast because we needed to have it like an everyday basis or something like that. So it was hard because we had to learn how to use Buntao in that case. I have experience in that. And by the way, once you get to use and the code is clear, it was like two years ago.

 

So at that point, there was not a lot of documentation about how to do this integration, this transformation of data. And nowadays there are a lot of tools to do that. Santo’s open source that you take your your data model in a relational database and it transform automatically. So in that case, we really needed to, uh, to see where we put the properties, where we put the labels. And it was like more handmade work. But is that enough to get it together?

 

Yeah, yeah. But it was easy, like there were some tools to transform these data. And once they graph this was validation tool is automatically like bad bet everything. And you do have it in two places at the same time. Very pretty. OK. All right, thank you so much. My thanks to The Guardian. I’m going to give you my thanks to you.

 

We hoped you enjoyed the podcast. To get more of our own material and to keep up with the latest industry and research news from our domain, we invite you to connect with us.

 

Connected data, London has an omni channel presence. Besides all major podcast platforms, YouTube and SlideShare, you can find us on Twitter, LinkedIn, Facebook and Instagram. You can join army troops or you can keep up with our news and special offers by joining our mailing list.