Bad data isn't firefighter laziness. Everybody did their job and the record still lied.
Robert Grand · Battalion Chief who still runs calls
A neighbor called it in around dinner time. He could see flames through the front window of a house a few doors down and across the street, and he did exactly what you want a neighbor to do. He picked up the phone.
We got it as a house fire. Full assignment, and we moved like it. When we arrived the doors were locked and nobody was answering, so we started working on access, because that’s what you do when a neighbor watched fire in the living room and the house won’t open.
Then the front door opened. The owner stood there looking at us, reasonably confused, and asked what we were doing.
We told him we had a report of a house fire at his address. He said the only fire going was his gas fireplace. He and his wife were having dinner and watching a movie, and they’d turned it on to set the mood.
It was funny. It’s still funny. Nobody got hurt, the couple got their evening back, and the report went in accurately as what it was, which was date night with a gas fireplace burning.
Everybody did their job, and the first record still lied.
Everybody in That Chain Was Right
Walk it back and find the mistake. You won’t.
The neighbor was correct about every single thing he could see. There were flames. They were in the living room. It was that house. He was not exaggerating, he was not confused, and if the fireplace had been a couch he’d have saved a life by calling when he did. Dispatch coded what they were told, which is what dispatch is supposed to do, and the protocol they used is built to lean loud on purpose. We responded to the worst credible version, because that’s the job.
Four links in the chain and not one of them got it wrong. And the record that existed before we knocked on that door said structure fire, at that address, with visible flames.
That record has a life. It gets a timestamp, a unit assignment, a response time, and it goes into the pile that we later hold up in a budget meeting and call our data.
We caught that one because the guy opened the door and told us. Most calls don’t hand you the punchline that cleanly.
The Only One Who Actually Knew
Here’s the part I’ve been chewing on.
Everything upstream of our arrival is somebody’s best guess made over a phone under stress. That’s not a knock on callers or on dispatch, it’s just what the information is. Chest pain that turns out to be anxiety, a two acre brush fire that’s ten feet by ten, a smoke report that’s a barbecue. Every one of those started with a person doing their honest best to describe something they’d never seen before to a stranger who has to make a decision in twenty seconds.
Everything else we collect automatically is a timestamp. Enroute, arrival, unit ID, GPS breadcrumb. Perfectly accurate and it tells you nothing about what happened.
Which leaves exactly one person in the entire system who actually knows what happened, and he’s standing in a living room looking at a gas fireplace.
Not the caller, not the telecommunicator, not the software. Him. And he’s the one we’ve invested the least in, handed the worst tool, and asked to enter the data at the end of a long shift.
We built a whole profession’s dataset on top of one person’s memory and never once designed around him.
We Got Good at Measuring What Nobody Types
Look at what we’re most confident about and you’ll notice a pattern. Turnout times. Travel times. Unit hour utilization. Call volume by station by hour. All of it clean, all of it defensible, and every bit of it collected without a human being involved.
Now look at what we’d actually want to know. What was there when we walked in. What worked. What the preplan got wrong. Which door on the north side is chained from the inside. What the family told us that never made it anywhere. All of that requires a person to stop and type it.
So we optimized for the collectible and started treating it as the important. Not because anybody decided to, but because the clean numbers were sitting right there and the messy ones needed somebody to sit in the front seat at 0300 and write.
That’s not a data quality problem. That’s a design decision somebody made years ago and nobody wrote down.
What a Bad Form Teaches a Crew
I’ll say the honest part. I have written narratives I would not want read back to me. Nothing went wrong on the call. It was late, the box was small, the drop-down didn’t have the thing I was actually looking at, and I wrote what would clear the validation so I could go to bed.
That’s not laziness and I don’t accept the framing that it is. That’s a person responding correctly to a form that was never built for them.
Think about what that form is teaching, run after run. Half the required fields are there for a state submission nobody in the building will ever read. The drop-downs don’t contain the thing you found. The narrative box is a two inch window on a screen you’re holding at an angle in the dark. Every one of those is a small signal that this is an obstacle between you and the end of your shift, not a record anybody intends to read.
And once a crew learns that lesson, it applies it everywhere. The one field that would have mattered gets the same treatment as the twelve that didn’t, because nothing in the experience distinguished them.
Here’s the piece that bothers me most. Almost nobody who will ever enter a report is in the room when the reporting software gets chosen. We evaluate on state compliance, on billing integration, on what the dashboard looks like in a demo. The person who has to live in it for the next seven years, at two in the morning, on a screen, after a bad one, is not at the table and usually was not asked.
We pride ourselves on being data driven while treating the front end as somebody else’s problem. Those two things cannot both be true.
The Front End Is Finally Worth Building
Here’s why I think this is the right moment to say all of this out loud.
The capture layer is genuinely getting interesting for the first time in my career. Voice dictation and image capture are showing up inside ePCR products now, with tools that read a chart before submission and flag what’s missing rather than just rejecting it (EMS1, https://www.ems1.com/data-management/webinar-ai-assist-in-action-smarter-data-capture-and-confident-documentation-from-start-to-submit). That is a real change in kind. For most of my career the only way to get what an officer knew into the record was to make him stop and type it. That’s no longer the only way.
None of that fixes anything by itself, and I want to be careful here. A voice tool pointed at a bad taxonomy just fills a bad taxonomy faster. If the drop-down still doesn’t have what I found, letting me say it out loud doesn’t help.
What it does is change what’s worth asking for. The question in front of us stops being how do we get crews to document better and becomes what do we actually want to know, and what would it take to make telling us the easiest thing a crew does all shift.
That’s a question a department can answer without buying anything. Pick the five things you’d want a company officer twenty years from now to be able to find. Go look at what it currently takes to enter those five things. Then put a captain in the room the next time somebody’s picking software, and let that person talk first.
We keep buying better ways to look at our data. Nobody has bought a better way to give it to us. That gap isn’t a reporting problem and it won’t be solved by a cleaner dashboard, because the dashboard was never the part that was broken. The most reliable thing this service has is a person standing in a room who knows exactly what’s in front of him. We’ve spent twenty years asking him to describe it through a form built for somebody else. The tools are finally good enough to build it the other way around.
Robert Grand is a Battalion Chief at Eugene Springfield Fire with 24 years of service. He writes Frontline Intelligence, a newsletter on operational doctrine, technology, and leadership in Fire & EMS.
If this landed, share it with someone in Fire & EMS who needs to hear it.
#FireService #EMS #EMSDocumentation #FireDispatch #PublicSafetyLeadership #AI
Subscribe
From the floor, not the vendor booth. Two times a week.