Behind the (Technical) Scenes: Failure to Deliver

By Tyson Bird & Mara Corbett

The Story: How the rise of out-of-hospital births puts mothers and babies at risk

1 Problem: The results of more than 150 records requests about midwives and freestanding birth centers aren’t digestible, much less particularly helpful for you. Even one of those results, littered with jargon and complex IDs, can be confusing. And every time a new records request is returned to our data reporting team, it’s more rows in the spreadsheet. More jargon. More complex IDs.

Simply put, it’s a whole lot of data.

Results from early user-testing, marked up and hanging on the wall of our office.

How we did it: A whole lot of data can best be made useful when you decide who it’s for — in this case, moms. So, we asked ourselves “what does mom want?” and then “what does mom not want?” User-testing helped us choose search fields and wording for the database. Since the search field allows you to look for a city or name, we applied a filter in the code that looks to see if your query most closely matches a midwife’s name or the midwife’s city. If it closely matches the city, it returns the city in the search results, and vice-versa. This allows the entire query to be performed from one search box but returns the most relevant results.

We also hosted and served the database with data.world — allowing the reporting team to update the csv files on the data portal without any extra work for the development team (hooray for us!). The data.world API allows us to deliver an experience that looks and feels like the rest of the site while keeping the data in a manageable place.

What we learned: Analytics about the database didn’t surprise us — People gravitated toward the most straightforward actions and even our paired down, tested presentation included unused depth. The surprise in the numbers was scroll depth: People were getting to the bottom of each longform article.

We had two stories average more than 20 minutes on-page — significantly higher than other longform journalism we’ve published. Yet even pages with 5 minutes of use showed people made it to the bottom, suggesting readers scrolled rather than use the menu or leave. With a stylized card at the end of each story, we redirected a lot of traffic back into other stories on the site by providing a link at the bottom of the page to a next story. This is a behavior we can continue to test and harness on future projects.