It was one month until publication, and there was blood everywhere.

My partner, S.P. Sullivan, and I were in the home stretch of an 18-month investigation into the New Jersey state medical examiner, and the entire JavaScript framework I’d built to support our narrative was crumbling. Just weeks before we were set to go live, the CMS—the foundation of everything our readers would see—was just tossing a string of errors.

Nothing else was going especially smoothly, either. An edit of the main story had just come back to me and needed an overhaul. There was still reporting to be done. Our interview with the state was coming up.

I had three options, as I saw it: Build a scrollytelling website from scratch, keep trying to shoehorn my code into the CMS we were using, or give up on our original design concept.

After much hand wringing, I gave up.

And it was the best decision I could have made.

The Story We Wanted to Tell

I began poking around the New Jersey medical examiner system early in 2016. I’d just published an in-depth look at the heroin crisis in New Jersey and employees from within the system began contacting me, suggesting I should look into the system, which they said had been in shambles for decades and was under added duress due to the influx of drug overdose cases.

(Side note: One of my chief sources on this project was actually someone who emailed me to complain about the heroin piece. Go figure.)

After some sniffing around, I made a records request to the state for a database containing every death that had been referred to their system in the past two decades. I knew the database existed, and found out exactly what it was called before I made my request, which might’ve made all the difference.

Still, it took six months of negotiating, appealing denials, and legal pressure before the state coughed it up. It was details of 420,000 people who died under suspicious or unknown circumstances in New Jersey from 1996 to 2015.

With some analysis, it was pretty clear that what our sources had been saying was true. New Jersey was turning away two-thirds of the cases it was asked to investigate. The cases that did receive an examination or autopsy were taking months to complete. And some doctors were performing an excessive amount of autopsies—far more than limits set by the national accrediting agencies.

Through reporting this out, Sean and I discovered a series of horrendous cases. Body parts had gone missing, people who were likely innocent were rotting in jail, and the state was potentially ignoring dozens of child murders, which doctors had been screaming were being overlooked for well over a decade.

The JavaScript that Nearly Worked

We’re constantly looking for novel approaches to storytelling at NJ Advance Media, and I’m lucky to work at a place that allows us to push boundaries, even when some of the ideas tossed around are ludicrously dumb.

Through reporting the medical examiner story, Sean and I had collected at least a dozen personal anecdotes from people who had been beaten and battered by the state’s medical examiner system. Our story was pushing 10,000 words already, so we were having trouble getting them all into a mainbar we knew would have to be cut anyways.

We’d already decided to narrate our story in an unusual way, using the second person point of view—arguably a gamble to begin with—so we decided to play with that. The plan was to use JavaScript to automatically assign everyone who came into the story one of the cases we’d discovered, and as they read through, more would be revealed about what happened to that person.

It’s the same idea employed by the Holocaust Museum to great effect. Here, each person would get a different version of the story: Different lede, sections of text, videos, graphics—all of which would pertain to one case we had profiled.

So we built all the parts, which included the agonizing process of writing seven different ledes to our story. Then came the tricky part, getting several disconnected elements to randomly populate the same thread of information for one of our cases on the CMS we were using: ReadyMag.

We nearly got it to work. Nearly.

We built a script that would generate a random integer from 1 to 7 and use it to trigger segments of JSON that matched that number up and down the entire project. And it worked, until it didn’t.

ReadyMag allows you to insert code, but says to do so “at your own risk and proceed with extreme caution.” Essentially, if you break it, they’re not going to help.

Because of the way ReadyMag fires JavaScript , the script would only populate the JSON for the 10 slides above it, and our project was 20 slides long. Long story short, we were faced with three options:

• Crunch the project into 10 slides or less at the expense of design

• Build a scrollytelling page from scratch

• Scrap it and move on.

Two elements that didn’t make it into the final piece. Top: an auto-populating toe tag. Bottom: one of seven possible second-person ledes.

The Solution We Finally Arrived At

Every fiber of my being wanted our original second-person concept to work. I’d spent months on building the framework in ReadyMag and had enlisted colleagues to help build out the custom content. When it worked, it looked super cool, but we still needed to work out design kinks.

There simply wasn’t time.

We scrapped the second-person idea for the mainbar and focused on simpler graphics and other elements to fill out the story and showcase some of the anecdotes we’d reported out. It was the best decision I could’ve made.

I leaned heavily on my brilliant team: Carla Astudillo, Erin Petenko, and Disha Raychaudhuri. Together, we built out two locally focused tools that allowed residents to see how bad the medical examiner was in their neighborhood and contact their legislators if they wanted to take action.

I used the framework from one of Carla’s projects to showcase five of the custom videos we’d made for the anecdotes. I focused on making the graphics we had clean and seamless, and we published on time—just as a new governor was taking office.

The story got a swift response, was just shortlisted for two Data Journalism Awards, and got a nod from the Headliner Awards, 2nd place for interactive or web project. The most important thing to me is that the reform bill prompted by our work passed the assembly last week, leaving only the governor’s signature for it to become law. Not one person has voted against it through both houses and four committees. Barring a change of heart, it will likely be the catalyst that completely rebuilds a medical examiner system that has withered on the vine for more than four decades.

I still think our original concept was a good one, just not good enough to delay or ruin 18 months of reporting and investigation.

• Share on Twitter
• Share on Facebook