A tiny parish beat the city centre once the fundraising data was normalized
Raw totals showed the city centre raising the most for a local cancer campaign, but normalized by population a tiny parish came out ahead and parish presidents started competing. Filipa Castro of the Data Science for Social Good Portugal lead team shows how volunteer data projects create that kind of shift for NGOs. How does a short project get from a vague problem to results an organization acts on?
Learn the five-step journey of a data for good project, with real NGO examples
Filipa Castro, part of the lead team at Data Science for Social Good Portugal, walks through how a volunteer data project for an NGO runs, from meeting the beneficiary and scoping to recruiting volunteers and delivering results. She illustrates it with a Rotaract cancer fundraising case study and describes ongoing NGO projects and DSSG Portugal's public COVID-19 data repository.
“And also because I think it's when you start to actually work that you understand that your professional skills can also help you outside of work.”
“So instead of showing a table with numbers in a presentation, this is much more interactive and impactful.”
What you will learn
Plan a data for good project in stages: meet the beneficiary, scope, recruit volunteers, run and deliver.
Write a scoping document that narrows an NGO's many problems to one target problem for a three to six month project.
Staff small volunteer teams with a project manager, a technical lead and data scientists or data journalists.
Normalize results by population so visualizations reveal insights that raw totals hide.
Turn hard-to-use public data, such as PDFs, into an open, tested repository a community can build on.
Details
Level
Beginner
Format
Talk
Technical depth
Introductory technical
For
Student, Early career, Mid-career
Length
21:17
Published
July 2020
Questions this talk answers
What are the steps of a data for good project?
Meet the beneficiary, scope the project in a document iterated with them, recruit volunteers, run the project while managing the relationship and deadlines, then deliver results and collect feedback on impact.
How is a DSSG Portugal project team made up?
Usually one to five people, always with a project manager and a technical lead, and the others data scientists or sometimes data journalists depending on the scope.
How can data persuade volunteers and local leaders?
Objective metrics showed older volunteers how much they contribute, and fundraising results normalized by population made parish presidents competitive about raising more.
What did DSSG Portugal do with COVID-19 data?
It turned the government's daily PDF reports into a public CSV repository, first by hand and then with an automated pipeline and tests, and wrote an open letter offering help to the government.
How do volunteers extract data from news?
They write a scraper that collects articles from a news website into a CSV file, then clean and explore the data to get value from it.
About this talk
I'll bring my perspective on the potential of data science for social impact. I'll mainly share my experience on social data projects done with Rotaract and Data Science for Social Good Portugal . The presentation will mainly cover:
- the journey of a data for good project, from scoping to delivery of results;
- the impact of data analysis on social good, with a case study;
- the current projects of Data Science for Social Good Portugal with different kinds of NGOs;
- the amazing work done by Data Science for Social Good Portugal on fighting COVID19.
Normalize by population, find the real winner (62 s)
Freeing Portugal's COVID-19 data from PDFs (86 s)
Hello everyone. Good morning. It's still 9am in Portugal. So my name is Filipa Castro. I'm from Portugal and today I'd like to bring my perspective on how we can use data science for social good. So first things first, I will just explain how I got interested in data science for social good purposes. So around 10 years ago, I joined RotRact. This is a global organization of young people that get together to create positive change in the world.
There are a lot of clubs everywhere in the world. So if you are interested, you can look for it and we'll certainly find a RotRact club in your town or at least in your country, I'm sure. Later, I started my integrated masters in biomedical engineering in Porto, in Portugal. And this is also relevant, of course, in the journey, because this is where I got interested into data and learning from this data.
So I started to follow this passion. And once I finished my degree, I started a data related job. And this is important because it's basically a daily opportunity for us to learn more about the field. And also because I think it's when you start to actually work that you understand that your professional skills can also help you outside of work.
To do other things and to, for instance, achieve social impact. In 2019, Data Science for Social Goods Portugal was founded, there were already other chapters in other places of the world. And even though I had a lot of the world. And even though I was not one of the founders, some of my friends were. And so we decided to meet because I belong to RotRact. And the goal of DSSG is to meet with other associations and to help them achieve more in their projects, in their social projects, through data. So we decided to meet and start a pilot project that I will give more details after this.
And in 2020, I also ended up joining the SSG. And that's mainly to talk about our work that I'm here today. So now I would like to go more in depth about the journey of Data for Goods project, how we really do this, all the steps. So the first step is to meet the beneficiary. And the beneficiary here is an association, it might be an NGO, for instance.
And here it's really important that you meet with them, you understand the problems they might have, you understand the data they might have available or not, sometimes, and more important than anything, I would say, you ensure both parts are equally committed and motivated to do this project. So in this case, it was a meeting between the SSG and RotRact. And then we start what we call the project scoping.
So this beneficiary might have a lot of different problems to solve, but our projects are short. So normally we have a project of between three and six months. So we decide to focus on the target problem at first. And we start to write in a document, what we call a scoping document, the problems we want to address, the solution we are proposing, and the main goals of this.
We send these documents to the beneficiary and we iterate with them. So until both parts are satisfied with the project, basically. In the case of RotRact, we decided to focus on one of our projects, which is fundraising against cancer. So basically, this is a big campaign, which involves more than 100 volunteers, but we didn't have any kind of insights on which were the best locations to ask for money, the best shifts, or the best practices.
And we decided this would be a great project as a pilot. So explore this data that we had about the campaign, look for new insights and for new recommendations for future campaigns. And once we have this scoping well-defined, we start to gather volunteers. So the SSG has a lead team, which I'm part of. But the lead team is just responsible for organizing this logistics of meeting the beneficiary, finding the volunteers, communicate our projects, but we are not the ones who actually do the social projects.
So when we have a new project, we open a call to volunteers. We share this call in our newsletter, in our social networks, and then we run interviews with all the candidates. So normally, one project has one to five people. We always have a project manager, a technical lead, and the other ones will be data scientists, but they can also be like data journalists, it will really depend on the scope of the project we are talking about.
And then the SSG has this responsibility of engaging this community of volunteers, so we try to organize some learning events, networking events, to motivate them. So in the case of the pilot project with Rotaract, we have three volunteers, which were Daniel, Miguel, and Tiago. And once everything is ready, we actually run the social project. So the main goal and the main role of the SSG here is to manage this relationship between the volunteers and the beneficiary.
So every time the volunteers have some questions regarding the data, every time they have some questions about the scoping or if we need to adapt the scoping, the SSG will be there. We also need to ensure that deadlines and milestones are met along the period of the project. And once this is finished, we are the ones who deliver the results to the community or to the organization, and we collect feedback from the beneficiary to evaluate if this really has an impact on society or the organization at least.
In the case of our pilot project, the results are the output of the project was an interactive report that I will show next. So just to give you a bit of overview and more context about this fundraising campaign we are talking about. In my little town, we have like 20 different locations, which you can call parishes, and we have more than 100 volunteers that during two days of the weekend, Saturday and Sunday, they are in different places like supermarkets, churches, etc. asking for money.
And for these two days, we normally raise around 20,000 euros just in this small town against cancer. And so now I would like to show you in more details some of the results that this Social Data for Good project had. So I will try to show this. Okay, I think you can see my screen. Yes. Okay. So the first one is an interactive map.
So the fact that it's interactive is really nice because when you go to presentations, sometimes we have meetings with the presidents of the parishes to organize this fundraising and to present them with the results of their parish. And in a glance with this interactive report, you can understand which were the most successful parishes. So in other colors, you have the parishes that are worth more money per population.
And it's interesting to see, as you expected, that this town, which is the central one, would raise more money. And it's interesting to evaluate how this evolves for the suburbs. So instead of showing a table with numbers in a presentation, this is much more interactive and impactful. Another one that I have is the distribution of the amount collected by each volunteer according to their age.
So mainly for someone which is not into this project, it's just, okay, maybe older people with more than 30 years collect more money. And it's true. This is normal. And it can be explained because older people normally are more popular in town. They know more people or they are more, they have more experience in this campaign. But for us, this visualization is really nice because we know that in our club, we are a group of young people, but we also have older volunteers.
But every year, it's more difficult to convince older people to help us because it's like they have to stay, stand up for four hours in a row asking for money. So when you have like 50 or 60 years, this is not the most healthy way to spend your time. So we try to convince them, but we use this kind of data, really objective metrics to show them, okay, you are really important.
So we try to remind them that they are important for this campaign and to help us reach higher value of money. And here I have another one, which is basically a bar plot with the amount fundraised. Maybe I can zoom in for you to see better. With the total amount of money that each pair is collected. So here you can see that like the gray one, the gray bar is always the one with more money.
And this is not surprising because this corresponds to the center of the city. So this is not very impactful. So what the SSG did was to show the same results, but normalized by the population. And here things become interesting because you see that, for instance, here you have a very small town, which is this one in pink, that actually can collect more money than the main city.
So in terms of population, they are better. And when you are in a meeting with the politicians and the presidents of fish parish, they really become competitive about this, which is nice, because they will try that in next year, they will fundraise more and they start to be competitive with these small cities. Okay, so going back to the presentation, I already talked about a bit about each of these visualizations.
And now I'd like to talk about other projects that are ongoing projects at the moment. So one of them is with this public pet hospital, where we are trying to optimize the waiting times for appointments or for treatments. So basically we'll try to understand which kind of patients or group patients in terms of and understand the correlation with the waiting times so that we can optimize the schedule and that we can open it to more clients each day.
Then VOST is also a volunteer organization and basically they have a lot of data from the government and a lot of data from civil protection. So here we decided to focus on one project, which is the characterization of the impact of a hurricane that happened in Portugal. So basically we think and they think that their data is not totally accurate about the consequences of this hurricane and impact.
So we want to cross this data they have with some other sources like news that were at the time of the hurricane to better characterize with a more complete characterization of this impact. And this will hopefully help to prepare for better for other disasters for future disasters and to better prepare for them if we can understand what was the real impact and the consequences.
For instance, if there were some buildings that were not prepared, we can understand why and you can prevent this for the future. And also KAIS, so KAIS is an association that works mainly with homeless people. So they have this magazine that they produce every month. Then they give this magazine to these people and they can sell it.
So they actually work and then they can keep this income from the magazine and use it for their food or for the rent. And what we want to do, because this magazine already has 25 years, is to get all the data they have about these people and about their selling and try to use storytelling with data to tell the story and impact of KAIS to then engage the community and raise more donations and increase the impact of this association.
And last but not least, just to give you some context about what we did regarding COVID-19. So basically in Portugal, the government shares every day the data about new cases or test cases or deaths in a PDF file. And this is not very nice for data scientists and for data exploration because then you cannot extract the numbers.
You need to do this manually. So we decided to create a data repository online public for everyone to use. And we started to manually extract this data every day and put it in a CSV file so that scientists could use it for their analysis. And then we started to gather some volunteers to help us and now we already have an automatic data instruction pipeline that collects this data from the PDF every day.
So we just need to click a button now. And we also have some automatic tests that we use to make sure that the data makes sense because sometimes there are some mistakes. So when there are, we warn the community about this. Since then, like three months, we are receiving a lot of contributions from the community, a lot of dashboards, a lot of analysis.
This has been great. We also wrote an open letter to the government. Basically, with our experience from this project, we did a list of the gaps we think there are in the information they are sharing. And also we offered our help for free, our skills, in case they want some help in improving these methods. So maybe I have some time to show you this page I'm talking about.
So in case you want to visit, I think, yeah. So this is the page, the repository I'm talking about. And actually, what we did in the last weeks was, as we had some ideas on how to improve this repository, we created some mini projects, which you can check here. So this is, for instance, a project for sentiment analysis during the period of the pandemic, so that we can correlate the sentiments of people with restrictive measurements from the government.
This, for instance, is very simple. It's just, every time we update the data, we want someone to make an automatic script, so that we publish this information on our Twitter. And we have been receiving a lot of help from volunteers. So please check it if you are interested. If you have any questions, please ask in the chat, I will try to answer that.
You can also reach me with LinkedIn. Here you have the contacts for me and also for the SSG or even Rotaract. So please get in touch. If you have, if you are motivated to create your chapter of the SSG in your country, please get in contact with us because we can help. And yeah, and I think my time is almost over.
Okay. I have one question I can try to answer. So I feel it for most. How do we extract the data from the news? So here we can also try to share, for instance, for COVID, what we did in the repository that we have, is that here we have some data about, okay, noticias, which is news in Portuguese.
So it's possible to go, for instance, to a website of a journal or a magazine and to write a script, which you call a scrapper, that basically collects in a CSV vial all the news. For instance, if you choose news from the 10 of June, it will collect them. And then it's a bit of data cleaning and data exploration to retrieve some value from this data.
But it's basically that. Then I can give you some more details later. Yeah, I think the time is gone. Thank you, everyone. Okay. I'm seeing here people that created dashboards. Yes, if you are interested into it, you can also see here in our repo, it's mainly about Portugal, of course. But here we have some applications that people did.
So if you open, for instance, this one, you will see the dashboards. So every time someone does a dashboard, we share it with a community, for instance, this one. Okay, I will share the links. Okay. So this one for the record story. And in case you want to know more about the association, it's dshe.pt. And the website is in English.
So I think everyone can understand it. In terms of contacts, if you want to reach, I will leave this here. This one. Okay. So you also have my contacts here. And thank you very much for your presence here. I will now be on other sessions. Thank you. And please reach me on LinkedIn. And I will still check some questions here and answer if I can.
Okay. Bye.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Article based on this talk
How do volunteer data scientists deliver real impact for NGOs?
Filipa Castro of Data Science for Social Good Portugal shows the journey of a data for good project, from scoping to an interactive report that changed how a town fundraises.
In short
Volunteer data projects work when the problem is narrow, both sides are committed and someone manages the relationship from start to finish. Filipa Castro, part of the lead team at Data Science for Social Good Portugal (DSSG Portugal), explains the five steps her organization follows and shows what they produced for a Rotaract cancer fundraising campaign, several NGOs and Portugal's open COVID-19 data.
Key takeaways
Plan a data for good project in stages: meet the beneficiary, scope, recruit volunteers, run and deliver.
Write a scoping document that narrows an NGO's many problems to one target problem for a three to six month project.
Staff small volunteer teams with a project manager, a technical lead and data scientists or data journalists.
Normalize results by population so visualizations reveal insights that raw totals hide.
Turn hard-to-use public data, such as PDFs, into an open, tested repository a community can build on.
From volunteering to data for good
Castro's path into this work started about ten years before the talk, when she joined Rotaract, a global organization of young people who get together to create positive change. Her integrated master's in biomedical engineering in Porto sparked her interest in learning from data, and her first data job showed her that professional skills could serve causes beyond the office.
"And also because I think it's when you start to actually work that you understand that your professional skills can also help you outside of work." (1:14)
DSSG Portugal was founded in 2019 by friends of hers, joining chapters already active elsewhere in the world. Because she belonged to Rotaract, the two groups met and started a pilot project, and in 2020 she joined the DSSG lead team.
Meeting the beneficiary and scoping the work
The first step is meeting the beneficiary, usually an association or NGO. Castro stresses that this meeting is about more than understanding the problem and the available data.
"And here it's really important that you meet with them, you understand the problems they might have, you understand the data they might have available or not, sometimes, and more important than anything, I would say, you ensure both parts are equally committed and motivated to do this project." (3:02)
Because projects last only three to six months, the team then picks one target problem and writes a scoping document describing the problem, the proposed solution and the main goals. The document goes back and forth with the beneficiary until both sides are satisfied. For the pilot, the focus became Rotaract's fundraising against cancer: a campaign with more than 100 volunteers but no insight into the best locations, shifts or practices.
Recruiting volunteers and running the project
The lead team does not do the analysis itself. It handles logistics: meeting beneficiaries, finding volunteers and communicating projects. Each new project gets an open call shared in the newsletter and on social networks, followed by interviews. Teams usually have one to five people, always including a project manager and a technical lead, with data scientists or sometimes data journalists filling the remaining seats. DSSG also runs learning and networking events to keep the volunteer community motivated.
During the project, DSSG manages the relationship between volunteers and the beneficiary, answers questions about data and scope, adapts the scope when needed and keeps deadlines and milestones on track. At the end, it delivers the results and collects feedback to judge whether the work had real impact.
What the data revealed about a cancer fundraiser
In Castro's town, more than 100 volunteers spread across about 20 parishes on a Saturday and Sunday, asking for money at supermarkets, churches and other places, and the campaign normally raises around 20,000 euros, as she described. The pilot's output was an interactive report, which she found far more useful in meetings with parish presidents.
"So instead of showing a table with numbers in a presentation, this is much more interactive and impactful." (9:03)
One chart showed that older volunteers collected more money. That helped the club, which finds it harder each year to recruit older volunteers for four-hour shifts on their feet.
"So we try to convince them, but we use this kind of data, really objective metrics to show them, okay, you are really important." (10:12)
Another chart showed raw totals by parish, with the city centre always on top. Not very useful, so the team normalized by population, and a very small town came out ahead of the main city. Parish presidents became competitive about raising more the next year.
Ongoing NGO projects and open COVID-19 data
Castro then described projects under way at the time: optimizing appointment waiting times for a public pet hospital; working with the volunteer organization VOST to cross official data with news reports and better characterize a hurricane's impact on Portugal, to prepare for future disasters; and using data storytelling for CAIS, an association whose magazine, sold by homeless people, has run for 25 years, to engage the community and raise donations.
The largest effort concerned COVID-19. The Portuguese government published daily figures in PDF files, which made analysis hard.
"And this is not very nice for data scientists and for data exploration because then you cannot extract the numbers." (15:04)
DSSG Portugal created a public repository, first extracting the numbers by hand into CSV files and then building an automated pipeline with tests that flag mistakes to the community. Over about three months it received many contributions, including dashboards and analyses. The team also wrote an open letter to the government listing gaps in the information shared and offering its skills for free, and it opened mini projects such as sentiment analysis during the pandemic and automatic posting of updates to Twitter.
In the Q&A, Castro explained that news data is gathered with a scraper that collects articles from a news site into a CSV file, followed by data cleaning and exploration. She invited viewers to contact DSSG Portugal if they want to start a chapter in their own country.
Watch the talk
Data Science for Social Good, by Filipa Castro (about 21 minutes): https://www.womentech.net/video/filipa-castro-data-science-social-good
About the speaker
Filipa Castro is part of the Lead Team at Data Science for Social Good Portugal, with a background in biomedical engineering. She works as an artificial intelligence and computer vision researcher making subsea operations more efficient and sustainable, and says DSSG Portugal gives her the chance to work towards making people's lives better every day.
Questions this talk answers
What are the steps of a data for good project?
How is a DSSG Portugal project team made up?
How can data persuade volunteers and local leaders?
What did DSSG Portugal do with COVID-19 data?
How do volunteers extract data from news?
Drafted with AI from the talk transcript. Editor review pending.
Frameworks and ideas mentioned
Scoping document
Resources mentioned
Data Science for Social Good Portugal COVID-19 data repository
Data Science for Social Good Portugal website
Rotaract clubs
Cite this talk
Filipa Castro. "Data Science for Social Good." WomenTech Network, 2020. https://www.womentech.net/video/filipa-castro-data-science-social-good