 Okay, good afternoon everyone. I think we'll start now. Welcome to the Open Data Institute. My name is Anna Scott. I'm the editor here. And I'm delighted to welcome Andy Hamplet and Simon Rafer for today's Friday Lunchtime lecture. Andy joins us from AAM Associates and Simon from Capellia. And today they'll be talking about the work they've done for the Trus or Trus Food Bank Network and how data might be the latest weapon in our fight against hunger. If you'd like to join the conversation online you can use the hashtag ODI Fridays. And there'll be some time for questions at the end if you have any. And for those listening online, if you have questions, you can also ask with the hashtag ODI Fridays and we'll be sure to ask them at the end. Thank you for having us here. And a little wave to, I believe leads ODI are in here as well. So a little wave to everyone leads. We're here to tell you the project we've been doing in about the past year, which is looking at data within food-back networks. I'm going to take you through how we got to data. We didn't actually start out either as an open data project or actually an analytics project. So just take you through those early formative parts of the project. Mainly so we can learn a bit about the main client, which is the Trus or Trus. So the project was actually to, it was a research project in partnership with Hull University, funded by RCUK. And it was to explore the potential of digital technologies within food-back networks. So we got the money and we approached the Trus or Trus to say, can we play around with this? And the trust itself is a food-back network. It's the largest UK food-back network. They work on a social franchise model, so they have an operation. And mainly churches locally decide there's a need in their area for, there's a big area around food poverty. And then they engage with the trust to use their models and use their brand and use their networks and all the support they give to enable local delivery to happen. But generally their mission is to empower local communities to combat poverty and exclusion. And when we get to the later parts, those two points there about poverty and exclusion, you'll notice they don't actually say food. It's not about food, it's about fighting poverty, which becomes really important later on in the project. And just a second thing on the trust, those of you who've been following the press, it's been kind of like a touchstone for the austerity crisis in the UK, where food banks have risen exponentially over the past five or six years. That's two years in a row now, the latest stats there on March 2015-16, where they've provided over a million food parcels a year. So that's not a million individual people, that's a million parcels of three days' worth of emergency food. So there may be some people who've just been there, but still an astonishing number from those who've raised up where they'll be. And then organisationally, because we're here to find help when we're as an organisation to use data and use technology to really fight what they're going to do, they've also gone through a really, really rapid growth, well under a million to well over seven million in four or five years. And it's quite a good story for them actually to be able to be on top of that and grow as a charity and keep it on top of it. So rapid growth, trying to counter poverty in that locale. So we did two things as part of the early part of the research. The first one was with Giles Hindall and Richard Vigeon from the Hull University, and they used operational research methods to do some business modelling. So we did loads of workshops, talked to the Trussell Trust and the local food bank managers, many of whom are volunteers, just to work out what is it you do, how do you progress through food, through your networks, and mainly we were looking for the holes in the road, so it won't go into too much detail, but that was it. It was a business modelling about just to get under the skin of what the charity is all about. And then we, on the other side of the research, did loads of technology research. So we were looking for technologies which were being used or being explored in food bank networks globally. We found about 60 or 70 of them, and some really interesting stuff in there. There's one, it's a very simple online volunteer management system where I'm a volunteer, I can log in and just book the sessions that I'm going to come in and person those sessions rather than having to go through a volunteer and do diary management. So simple technologies were there. The point of doing these tippies research was to look at mainly the holes in the road that they were in their operation. Where are you struggling? Where are you stretched? Where isn't it going brilliantly? What's happening over in the technology world that might be able to plug those gaps to make a big difference, to sort of fit them in, and then we had some money to actually go and build something. But what we found actually is the two things. One, they're a really well-run network, they have pallets and pallets and tons and tons of food to the network, so that's well done. They are well separate volunteers, they're growing. So there wasn't anything jumping out immediately to say, actually, technology is your answer, let's do it here. So we started then to look at their broader strategy. And as any good charity would be, they all started thinking, well, how do we design ourselves out of this? We're here to meet an urgent need and crisis which we don't particularly want to be there. So what is our role in actually helping people come to us with food once and get that emergency supply, but then not come to us again? So they're starting to explore their more than food initiative, which is about understanding them as one player in a connected network of people trying to fight poverty. So that's where it is. And then we started to think, well, if that's what you're trying to do, what you need to do more is understand the people who come to you in time, and that's what led us to explore data. And we took a few examples to them around where data analysis, in very small ways, has started to help charities understand what they're doing. So Butler UK is a funder in the UK and a great organisation you'll know called DataKind did this work, I think. And so they provide small grants to people in crisis. They did a text analysis survey of all of the applications that came into them, and they found some really interesting things using some data science methods. One outcome was, two outcomes, one they realised that if bedwetting is mentioned in one of those applications, it's a real strong indicator that these guys might need to be fast-tracked, that's something there that seemed to suggest that they would do it. And then they started to tweak their application process where they could see really early on these might be some sort of people who are really in need. So just again, an early stage understanding of data analysis. Very simply visualisation, you know, what I should have said earlier on is the trust have been really, really good at collecting data. From a very small charity, they set up their lines in the sand quite early on to say we're going to collect data. It's been really, really helpful for them in the policy agenda to try and take this out. We should say those million numbers, they travel far in the press every twice a year when they bring stuff out. But we just took this little example. This is someone just using a commercial data visualisation platform tableau, and it was a teenage pregnancy advice group in the States. And they just said it just transformed the way we could work as a team looking at not just our data, but other data out there. So rather than have someone sitting in a back office looking through the Excel spreadsheets and putting it out, everyone could play with it. We could look at it in meetings and actually sort of do stuff with it. And then finally, obviously on the edges, just to start to tease about predictive modelling and this is a fairly well-known example from New Orleans. And they piped a load of open data into helping them design an outreach strategy. I think the two things they looked just reminded myself. So they found the houses that were least likely to have a smoke alarm in their house and the most likely to suffer fatalities because of fire. We've got limited resources, that's where we're going to put them. So fairly simple data analysis which comes with some quite interesting stuff. And so at that point, we thought, hang on, you've got data. You know what your mission is, what can we do? And this is where Simon came in from. Thanks Andy. Yes, I'm a statistician. Well, data scientists as we're now widely known. But I was brought in by Andy to probably sort of inspired by those maps that you just saw particularly around the predictive analysis that those guys were doing with those maps to see if we could replicate or sort of take forward the work that had been done there but using the trust or trust data. As Andy says, the data there was very rich. It was both sort of temporarily rich. I mean, they'd been gathering data daily since for the last three or four years and spatially very rich as well because they'd logged the locations of the people using the food banks and they'd classified them by electoral ward. So there was fantastic data to be done there. With our eyes on the idea of developing this tool, we first did some preliminary analysis of the data. And this is, as I'm sure you realize, it's an important first stage in any of these projects because in order to understand your sort of direction and what can be done, it's important to get this kind of overall general picture. And very often, organisations can be very good at collecting data but they don't necessarily use all the tools that they can to see it. So one of the first things we did was some quite simple stuff. We just broke down their vouchers by the crisis because people logged their crisis on there. They'll say if the use of the food vouchers was due to benefit delays or homelessness. And then we just kind of, for example, in this graphic here, we just cut it by region. I know I'm going to trip over this. So I don't know whether you can see here. It all looks kind of the same to begin with, but when you start to look carefully at what's going on, you can see some quite interesting patterns. So for example, in the southeast, low income is much more of a factor than in other areas. And in say the north and the east, there are many more issues to do with benefit delays. So that for a start was quite interesting for them. We presented a number of different slides on the things that were going on, including time series analysis. Now, this is a very simple thing to do, but they actually hadn't seen it in this level of detail. So for example, you can see the growth over time and you can see the pattern that occurs around Christmas time. So that was immediately food for thought, because what seems to be happening here is people needing to use much more, many more food parcels prior to Christmas, and then there's some kind of slump. Did they identify the reason in the end for that slide? I think that all of these things prompted them to go and look at it in greater detail. But they thought, A, you don't want to be without food over Christmas, secondly. But then they also thought that they wanted to look at the opening hours. There's quite a lot of people, quite a lot of food banks which are volunteer led. Obviously they want to spend time with their families as well. So that mix of people knowing the food bank is going to be closed probably led to that spike. And then there was the general levelling off, which was quite interesting. And lots of questions arose out of this. What was behind the initial expansion? Was it just the food banks opening up to increase the overall volumes, new locations, et cetera? Or was there increased usage of the food banks? Lots of different projects going to span off that initial analysis. So, yeah, the point there was just to say, Oh, I haven't done one of these things with builds before. I don't really use PowerPoint. So the point there was just how important that initial preliminary analysis is. We then formulated our goals for this mapping tool. And one of the most important thing obviously was to make use of maps to understand the geographical locations. So a very powerful tool for seeing patterns in data. And they're not nearly used enough. We then also wanted to make use of all the open data that's out there. Particularly because as you'll see, we would like to relate some of the demographic data captured by the census to the food bank usage data. But more important, I mean, those are things, this is sort of general objectives, but we also wanted to create something that was useful to the trus or trus, both at a high end sort of at the head office level where they could see the overall patterns of usage in the country and also for the individual food bank managers so that they could see locally what was going on. And we also wanted to see where it's being used and also how far each food bank had an influence. So what was the furthest direction that somebody had come from to use a particular food bank. Obviously you can see that that could have big implications for the way in which they expand their network. People are going a long way to use a food bank that maybe they needed to open a new branch. We wanted to make sure that it was usable through a browser because browsers these days are probably the most effective way of delivering anything and also means that we could ensure that the tool was usable all the way across the trus or trus because everybody's got a browser. And we wanted to ensure that whatever solution we built was easily extendable. So we used lots of conventions that people developing software used to ensure that somebody come along and plug in new components into the work. So here we go. This is the tool that we eventually built. It's a prototype but it was fairly smooth by the time we got to it. What you can't get from the static image but you'll see from the video is it's very much zoomable so you can zoom right into a town, see the streets and zoom out again and the uses patterns will be there. You've got, well I'll explain, you've got a control panel down here and I'll explain what this stuff goes over here on this side as we go in through the next slides. But you can slide down the opacity of the heat map and see the towns and villages underneath as well. So here we go. So just to talk a bit more about the functionality. We've ticked actual usage here. Now we are highlighting in darker blue those electoral wards where usage is highest per head. So they can immediately see where their zones are. It hasn't come out so brilliantly here but the orange circles are the actual food banks as well. So you can see the usage around a food bank. You can also click in and select, I think we've just truncated the picture here but we can select a different crisis type. So here we've gone for, which one was it? Was it something like homeless now? I think it was, yeah. So we can toggle in the different crisis types and see the different maps for the different types of problems that people were having when they're using the vouchers. And then here you see some functionality which we use where you, this is the beach stuff. So we are selecting a particular food bank and we're seeing that, here it is right there, on Ponzi or something like that and this is how far people have come to use that food bank using the actual voucher data. And finally is a bit of an experiment. We use the Google API to work out the travel reach from each food bank so we can see how far people, using different types of transport whether they're walking, whether they were using trains or driving, how far they had to, how far they could come in 30 minutes, absolutely. Sorry for that, sir? I don't know, no, but so I mean that would be interesting to look at actual distances and go and reverse it. So to take the actual people and say how far did it take you to go? That would be a brilliant thing for our next phase, but this was kind of the other way around. This was how far can you get to in 30 minutes if you do what I mean. So that's a really good one to look at next time. So, but the real, I think the really, the most powerful and interesting from our data perspective use for the trussel data combined with the census data was our predictive tool. We pulled a whole bunch of different measures from the O&S site, about the census from 2011, stuff like deprivation indexes, employment, the percentage of the people who are in part-time employment or low-paid work, health stuff, you know, amount of people who are having to care for others, any indicator that we thought might have some kind of relevance for food bank usage. Then we used that data and we married it up because we got ward level data, electoral ward level data for the trussel trust and electoral ward level data for the census, married it up and then built a model, not a perfect statistical model but a fully functional usable one, which tried to predict which areas would have high food bank usage from the indicators. That way we could then use the model and say which areas do we predict would have food bank usage but don't have food bank usage. So it was a very useful contrast for the trussel trust. They can toggle between actual use and predicted use. If you like, the predicted indicator was just a very sensible index of those measures from the census that are highly predictive of food bank usage. So is it back to you? I think so, yeah, thank you. Cheers, cheers. So all this is very interesting and there were lots of ooz and ars when they first saw the map, the trussel trust stuff were with us obviously all the way through, but how can this be useful to them? So first of all, just on location and again you can't see all of these, I'll put all the three locations up. So all of this obviously sits in an Excel spreadsheet but until you see it like this they couldn't actually, you would never think, oh that post goes a long way away. So all of these things are trying to make them understand are we placed in the right area. So for example in that first one, they're right up at the top, loads of people seem to be coming from a long way away. So there's that one. Other ones that we looked at as well were also looking at why so many people passing other food banks to get to us. Is that a communications issue, is that they don't know about that or were they just here visiting friends, what is it? So let's look at that in greater detail to see how are we as a network working together to allow people to receive the support they need as close to home as possible. And similarly in that middle one, so you can see the greater usage is way up in the sort of north, in the northwest part of that little grid. So all of them, it was great, we presented this at the National Food Bank Conference and there was a big queue of people wanting to play around with their area and go oh I want to see where people are coming from and how can we change. So that's where the interesting bit comes in. And then you've got some other ideas to look at. So again, this is up in the northwest of England I think. So again, they hadn't realised when they looked at this quite how much of an issue homelessness was. So this is where you start to move into them as strategic operators. So if homelessness is a big issue for the people who are coming to us, what do we do with that? How does this fit into the modern food thing? Who do we need to engage with? What other homeless agencies would die for this information? What is it that we do to play a part as a strategic partner rather than just giving people food and then sending them off again and hoping they don't come back? So again visualising this is really important to actually what's the greatest need and areas of crisis in our area that we should be looking at in greater detail. Similar there. And then you start to think about them as strategists and not just about thinking through providing food but anyone who looks at that and thinks school holiday meals, kids can't get fed in school because obviously they get free school meals during the school term. But then once you're out of that you cut off and you're on your own. That is a really powerful lobbying tool to go to local authorities, to go to local funders, to go to everyone and say, look this is what we're up against in the summer holidays. Can you help us? Can you provide fundraising? Who else do we need to work with? So again, just the visualisation of the trouble they're trying to sort out is really really helpful. But as Simon said, I think the predictive need area is where it should possibly have the greatest strategic input for them as a corporate body. Because go back to the earlier slide, their whole model was based on a social franchise network where local churches see the need, start to meet that need, and then it sort of comes up this way. What they've got now is a different view of the country they're operating in. But they can see really big blotches of dark blue where they think we're going to be needed here. What do we do? And then actually you start to think, well is the thing to just go and rush there and start at food banks? And I carry around in my head a shocking statistics. We talked to a few of the American food banks, and they're absolutely enormous, but I didn't realize how big. Feeding America, for example, turned over $1.8 billion last year, which is enormous and shocking and incredible and hard to get your head around. But just tracking that story back, in the late 70s food poverty was massive in the States. And so this raft of charities stepped up to the plate. But now they're part of the system and they're embedded. So you couldn't possibly take them out now because what would all those people do? Does the trus or trus want to do that? Do they want to rush to meet that need? And then actually they're there forever, and they're part of the state provision almost. So they've got some really interesting strategic issues to work here all aligned to their more than food initiative, which is okay. We attract people because they need food because they're in crisis. Once we're here with them, then what? So they've now commissioned more university research to think about their more than food initiative, to think how data plays a role into it, and actually think about how they can act as part of a network of anti-poverty charities that take stuff forward. So that's the really exciting bit. And then you can just sort of step away and happy to talk to everyone. We've got loads of ideas about how to take this forward. We're going to continue working with the trus or trus themselves to take Simon's brilliant prototype and get it into a working tool for all the local food banks because once it's in their hands, they'll find a million things to do with it. We know who will take that forward. But I'm perhaps more interested in this as a model for what other charities might do. I just pulled up two slides there. So that's where we are. And those are all the local charities in the area where they're based. The wealth of information that they're sitting on and not looking at is absolutely enormous. And already, you can see, two or three of them in one area have postcode data related to the crisis type they've got, put all those together, and you've got a really rich picture of poverty and crisis from the bottom up, which you can then align with all the open data. Stick Hackney into the London data store and you've got 56 data sets. That's something, I think, really, really important for most charities in the middle there. Let's work out what your data is, who else can share some stuff with you, and then where's the open data on top of that. So that's all we've got, too. And we're very, very pleased to take some pictures. So you can see this working. I'm just going to play this in the background so you can see it zooming in and out and doing it to think. Thank you very much. Cheers. Ian Duncan-Smith got really angry with the Truswell Trust that he said they were politicising the issue. And I wondered whether, as part of your model, you could factor in future legislative changes so if there was a proposed future cut in benefits or something like this, whether you could model what effect that might have on a particular region or a particular area as a sort of strategic tool for opposing these sorts of changes. Do you want to say anything? Yeah, sure. I mean, from the... I'll talk about the modelling, but you can do the politics. Yeah, we actually... What we didn't mention is we had a few spin-off projects where we looked at a sort of dynamic model of the growth. We had planned to look at things like the type of causes that you'd mentioned, looking at the impact of different policies to do with the supply of benefits. But we found in the end that we first wanted to answer some difficult questions about what was fuelling the growth in the food banks as a system in general. So it was quite difficult to just look at whether or not growth was due to the food banks increasing, the number of food banks increasing rather than just to the demand. If you can imagine that's quite... That's a model with two moving parts and you have to separate the two of them. So we didn't get on yet to the... looking at the impact of the policies on the time series. I think that is a very good area to look at and I think... and it's probably some ideas about whether or not we'll be looking at in the second phase. I think we're okay with this one. Oh, there we are. Oh yeah, why am I doing this? Sing us a song? Sing us a song if you want to spy. We've got a lot of feedback. I think two things on that. Yeah, we were poking around what was really interesting in this as we were presenting loads of really exciting pictures to them. As with a lot of these kind of projects, it kind of confirmed stuff that kind of half knew already. So for example, related to this, not directly, when there were a few spikes in one of those charts, they were, why is that? And they thought maybe that is because actually we're part of a referral network. I said earlier, everyone gets referred to them and they have to take a voucher and do all that kind of stuff. So people can only get to them if those referral agencies are still there. And a lot of advice bureaus, we're closing down. So things started to call down on the ground and I think, why is this? But seeing those spikes where they are, first of all said yes. And then they said, well again started to think exactly also how do we prepare ourselves for the future? What is it that led to those advice and was closing down and then any number of things? If we know stuff's happened here, how can we just get on the front foot and prepare? And I think the second part of that is what we're hoping to do next is engage with government agencies. This isn't about bashing any government about austerity. This is about saying, this is a problem that seems to be real. It's in your best interest to help and sort it. It's in ours, it's in everyone. So how can we come together and with your deep understanding of the data you have, how can we work together on this and that's what we're hoping to do in the next few months? My name is Isabelle and I'm a data journalist and I was wondering what is the, when did I start collecting the data and did you find any interesting findings through times or thinking to aging of the population or that sort of thing that would be reflected in the characteristics of families or like people asking for food packages or anything really? Yeah, okay. Yes, we did do some, we did look at the, there is an issue in the fact that the demographic data that they capture could be richer at this point. So there's a limit to the kind of questions we could ask of it but I think they're planning on expanding it. But we did see things like different regional patterns where perhaps the winter had a greater effect in the north than it did in the south, stuff like that. There was also some interesting relationships between the population density and the food bank usage. So we found that extreme ruralness was an issue. So the big outliers in terms of usage, those places where the demand was very, very heavy seemed to be in the rural areas. But like all research things, our research ended in a more investigation is needed and a provider. Because we're really doing this as a stepping stone on to building a tool and to open up questions. I think what I just said to that is they've been greater collecting the data and they were already thinking about what more data do we need to collect but this might give them a little bit of a push. So they're now moving to an e-referal system where it will automatically pipe into the stuff and they're starting to work. Well, how can we, rather than just saying are you over 18 or under 18, what kind of divisions can we get in there? And I've worked in charities a lot all my life and what's interesting, data collection, it's not particularly sexy or liked by anyone and you fill stuff in the other cold face and you fill it in and you send it off and never see that again and the funders want it and that's it. What's different about this is the local food banks can see what it's going into and it comes back to them in a virtual loop I think it's really important for everyone to want to collect more data. I just add as well the variables that correlated with the food bank usage when you start drilling to them they're very interesting too. For example, like Andy says they tend to confirm what you think but the very fact that they do does take the conversation forward. So for example it's the child summer holiday usage. What's it called again child? Child holiday meals. We correlated with people in part-time work or low-paid work so you can see that that was an issue there. Hi, I'm Elahisa. I'm from Brazil, from real ODI nodes and that would be very interesting to be replicated in Brazil but I have some questions what do you think they heard those because not only the tool but the data collection because you said the trust had a lot of data collected so what do you suggest first steps if you want to replicate that in our country is to find out what data is available or because it seems very organized not organized but a lot of data so you organize it in a way we could visualize but I would like to hear a little bit about what you suggest for other realities. What I would say what I found is really useful is just going to people with a range of case studies because data itself is quite scary lots of people don't have the technical understanding neither do I but once you start seeing half a dozen or a dozen things that other people have done like the bottle and the other stuff and just present that to the charities then that gives them a really good routine to thinking we could do that but just giving and the ODI is great at pumping these stuff pumping this stuff out and then that will drive them to think what is the data we collect could we do that, could we do that, could we do the other because one side doesn't have it all I would say pointing to what others have done and starting there would be the best place and I also add that the geospatial staff in terms of the mapping that's available is very well developed for the entire geographical map so the map of the world is fairly straight forward to be able to produce something as visually striking and interactive in that way and then whatever you've got that could be plotted over that data to inspire people so that they could see what the potential is and then that might, as Andy said get people moving in terms of collecting it's not easy and again data has this whether it's open data or closed data within most social organisations for lots of reasons so that thing I was saying earlier about closing that loop where people can see the value of it is really important one other quick example was a great book a year ago about a school over in California and they wanted to be data informed so they collected loads and loads of data but very quickly worked their way around saying we've got to use this and show this and the example they did was they looked across the performance of individual pupils and found that two lads were way down in this subject and they were up here on everyone else so they had built a data culture where they could take that to the teacher and say we need to explore this don't we so he invited a couple of other teachers to view his lessons to try and find their way at us then they came back again and videoed him and he looked and said I'm shouting at them why am I shouting at them, why am I doing that and next stage he was a Hispanic teacher they were Hispanic kids and he wanted to drill in them, you've got to do better you've got to do better he was actually forgetting his best practice in doing that so that is a virtual loop of data where everyone's comfortable with it they know it's there to support you and people are there to get to the good part and that's not easy but that's the kind of thing you mentioned earlier that maps aren't used nearly enough to represent data and see data why do you think that is? I think it's because this might be a total guess but the kind of journals that used to put the statistics that came out in would typically show a time series because it's easy to produce and the software that was previously needed to visualize data on a map was expensive and took time and now that's just broken down entirely because as I said in another talk recently it's all there for anybody to use you just need to have a little bit of JavaScript knowledge a little bit of coding knowledge and you can do it so I think that's the and I think because of that sort of historical chain of events people aren't used to doing it they forget that it's there there's a thing which you have to be careful with maps in that they can be a little bit misleading in the sense that you might get one block of the country absolutely purple because it's a large rural expanse so I think there are statistical reasons for being careful with using maps and that might help people back a bit but I think that's not an obstacle that can't be overcome My view on that from an organisational side would be yes it's becoming a lot easier from what I understand not a technologist but again a lot of social organisations are trapped in the data so scary we don't have the technology we're struggling anyway in the current crisis we don't you know where's the stories on return on investment what do we know so that they're not coming to it easily first of all I'd say and then if you look at maps generally you know even the really good funders they'll do a map or they'll do lots of maps and then it'll be in a report published once a year the difference with this and the difference with the tableau example around the teenage parenthesis is you can engage with it you can look at it, you can play with it and that's actually will drive you to ask more questions and take that conversation rather than reading a report going on what do we do with that Please join me in thanking Andy and Simon for coming Thank you very much Just to let you know that we will also be recording a podcast on this issue later today so we'll be going into a little bit more detail and you are welcome to listen to that when it's online in a couple of weeks our next Friday on the time lecture next week we'll be on thinking out loud which is our data as culture for art exhibition so hope to see you then, thank you