It Takes a Village: Helping people make decisions now, using data visualization.
My May 2019 SWD Challenge Submission
As I was reading this month’s SWD Challenge about Artisanal data, I was at work, re-opening a survey for a second time (In real life, this happens even with most well-intentioned strict survey close dates-someone important inevitably misses your close date). The next day I was about to hop in a car with this data and drive 4 hours to a city in West Texas to co-lead a strategic planning session. We were hired to help a child abuse prevention community collaborative find out what the service gaps are in their community so that they can begin to funnel their energy and funding into that direction.
As Mike Cisneros says in this month’s blog about the challenge: “When, then, can you, as the data visualizer, be perfectly confident in EVERY SINGLE FACET of a dataset? Only when this is true: when YOU are the collector.” While I do think in some cases that is true, especially if you are collecting data that is about yourself, in a strict uniform fashion, utilizing tools to ensure the integrity of your quantitative claims, there is still room for error. We are human after all. Anyone working in social sciences knows the pitfalls of self-reported data, social desirability bias is a big one, but there are many other ways the data you collect yourself can go awry.
The survey I created, had many fill-in-the-blank questions- you know the ones, notoriously known to be interpreted widely by the survey-taker, but it was the only way we could at least begin to capture the information we needed without making the survey 100 questions long. These are busy people, and we didn’t want people taking too much time on the survey. Right before I was sending it off to our client a co-worker went in and made some edits. In one of her edits, she expanded upon the services question. I had asked people a simple “yes’ or “no” about a small group of services they offered with their programs. My co-worker expanded the type of services we asked and added whether or not they made referrals internally/externally for the services. It was a great edit by my co-worker, but it also left me wondering what I was going to do with the data from this question. This question was the heart of the survey, it was where people might be able to see the gaps they needed to see. It was also one of the few non fill-in-the-blank questions and I had planned to develop a more simplified version of this:
From a few years ago, wouldn’t try and put this all on one page again ;)
I wanted to create a 1-page viz where people could see every program and what it offers on one page, but now, with the extended array of services and inclusion of referrals this would have become a color-coded mess, a data viz nightmare, kinda like the image above, but worse. I’d need at least 11-colors or icons for all the types of services measured. I don’t think so!
Why Referrals?
Referrals are important to capture. While a program may not offer a service, in fact, there is no way a program could offer all the services listed, but if they had a referral to send the family somewhere else to receive the service, that says something about the connectivity and collaboration of the organizations in that area. This was important to this discussion because there may not be any clear service gaps in the community, it may be more about gaps in collaboration and information. Something a community collaborative ought to know and address.
On top of the last minute edits on our end, we also had to extend the deadline to close the survey. In my initial plans I built in 4 days with the data to parse through and visualize. Unfortunately, I ended up having more like 4 hours.
Because I had very limited time with the data, I made two quick vizzes to show the data from the services question. A sad cluster bar chart and 6 page charted matrix-which provides the answers of every program that entered information into the survey = nicely formatted raw data. I formatted it as best I could using excel and word, printed it, and hit the road. Was it the best representation of this data- absolutely not, but did it get the job done? Well, maybe.
Sad cluster bar
a 6-page services matrix- organizations and programs de-identified.
For the record, the day was a great success. The group was highly engaged in our meeting, fully participated in the planned activities and at the end of the day we had started to understand some of the gaps. My vizzes were never meant to be the only thing people were to infer from and discuss. This is a group of people who bring a lot of expertise that could never be replaced by a survey. These people live and work in this community and can provide insight into how things really are. But as the experts worked in small groups, I saw them flipping through the matrix, and taking notes on the cluster-bar graph, so at best, I think the extra information certainly helped them in some ways.
I used this month’s challenge to take my basically automatic default presentation of this data and bring an artisanal flavor to it. I decided to try and combine my sad graph and chart into one viz. I tried to evoke visual metaphor by presenting the array of services, in an actual array, using the circles I used in the matrix. I like how the dots are gathered around the map of Texas almost like the programs, ready to serve the people of Texas. (which I will change to a county map, for the real report, just trying to keep things anonymous here). I also reordered the services by how many programs actually offer the service, so you can see that Home Visiting Programs was the most common type of service among the survey-takers and Legal Services was least common. I didn’t find this to be as clear in the original clustered bar.
Child abuse is not prevented by one program or type of service, it’s prevented by a strong network of services and community-oriented conditions and resources that serve children, parents and the community. This viz gives people an opportunity to see what kind of services are engaged in this community collaborative and who they may need to engage more, it also allows them to see how referred a service is. For instance, while few of the engaged programs offer housing and employment services, they do seem to know who to refer families to, while legal services are less referred.
If I had more time, I thought about adding call outs and annotations to each service circle, from other parts of the survey, Like information about the population that is served by the program, or the breakdown of how the programs are funded, but for now I’m sticking to data from this particular question. This is still a work in progress.
Could this data viz stand on its own for the general public? No, but it could be a useful tool for the people of this engaged community collaborative. This was fun, reconceptualizing very rushed work into something more visually appealing. It also helped me think about data viz and decision making and what people need to know in the moment when they are in a room together.
And as always, if anyone from SWD made it this far into the blog, thanks for the opportunity! Looking forward to next month’s challenge.
