A data-focused website creates opportunities for communication
Web developers, as an industry, generally shy away from data because it's not trivial to implement. In many ways website server architecture isn't naturally good with data, and most of the content management systems (CMS) are following suit. That's part of the reason we created byteCMS: data connections should be easy.
Since we have a CMS that's naturally good at data, we have an an opportunity to turn data sets into something easily to manage, like an organization's staff list. Of course it's possible to just type in a staff list onto a page, but then you lose some opportunities like having a featured staffer as a sidebar "bucket", or having the blog authors coming from the same list, or the ability to change someone's name and have it change everywhere it exists on the site. When you start seeing your organization's data as useful to people, and it's easy to manage, your website improves considerably.
But data gets a lot deeper, like the 1.7m item collection of Milwaukee Public Library or the 100k pages in manuscripts for The Walters Ex Libris that we needed to find a smooth and easy way to let people browse, explore and search deep into the content.
How we understand data
When we have a large, relational and hierarchical data set that's as large as MPL or The Walters, we find our first disconnect: our brains. There's no way to explore or understand the sheer amount of it, much less find what's relevant to users. So we need to build a tool: previsualization prototype.
We start each previsualization prototype project by just querying the data sets, and asking basic questions, how many X, how many children of X, how deep does the lineage go, etc. From there we start defining how we'd like to explore it, and make some basic prototyped interfaces and fill it with data -- all the of the subject items without parents, aka root subjects. Clicking on each one finds its children, which we put in a second column, and click on those to reveal a third column. Then we make basic hypotheticals, and let the data show us which are true. Then we affect the previsualization prototype and make more hypotheticals, rinse and repeat.
Turning data to interface
A previsualization project gives our UX designer and strategist to start finding ways to merge the original project's goals with the data we've found, and start sketching interfaces. We're looking to show the right level of abstraction to the user that makes understanding and managing the data easy for people, and the right kind of interface to affect the data based on the already-identified use cases (or user stories).
After sketching out layouts and interface ideas, we'll start honing in on useful patterns and start defining the user paths needed to make a system. Doing iterations of that will start showing systems that work, and give us a chance to make a prototype that we can test against, or we'll try more previsualizations based on our findings.
Our job is to distill meaning and find the appropriate abstraction to let people get the most from the data and its interfaces, and after years of experience, we're often right about users. But people understand and interface data somewhat unpredictably, and even our experience doesn't catch all the issues.
When possible, we build in time for user testing. which can be as simple as asking a few friends to do some site tasks while we look over their shoulder, or as deep as employing focus groups, task analysis and utilizing online tools that track mouse movement and user paths. This data will often have us fix some small nits that get in the way, and sometimes it sends us back to solve some bigger problems. Either way, it improves the product by making things easier for everyone.
- Editorial data visualizations
- Live data visualizations
- Dashboards and business intelligence
- Database design, imports, exports
- Open or closed data APIs
- Data normalization and denormalization
- Populating data sets via APIs
- Merging data sets through multiple sources
Byte is a believer of open data and has long been part of Milwaukee Data Initiative's efforts of opening government, and when possible asking our clients to have their open open data policy. Data is often just another form of website content, and often times an organization's own data can be useful to people in ways we can't predict.