I was in for an MRI recently, sitting in my wheelchair out in the hallway, when I noticed the techs huddled over a battered three-ring binder. They couldn't find the diagram they needed to run my procedure. The confusion was obvious enough that I offered to help from where I sat. One of them handed me the binder, and inside were screen captures and hand-drawn diagrams of terminals and plugs, laid out in tables, some full page, almost all of it covered in handwritten corrections to what the printed page said. A few pages had version control dates going back to 2015.
When I asked why so much of the binder was handwritten notes, nobody had a good answer. But one thing did stick with me: the only formal training these techs ever got was a short video. After that, they were on their own, quietly stitching together the reference material that should have existed from day one, in an environment where getting a terminal or connection wrong is not a small mistake.
I sat there in that hallway and thought about what I would have built instead. And I made myself a promise: I was going to write about why manuals, and job aids more broadly, still matter, even with video sitting right there and easier to produce than ever.
The pendulum has swung toward video, hard
I'm not imagining this. Video is faster to produce than it used to be, screen recording tools are everywhere, and who doesn't love how polished a talking head or a software demo looks next to a page of bullet points (which, to be fair, isn't always the gold standard either). There's real research behind some of that appeal. Software walkthroughs, the kind you see out of Adobe or Microsoft, are considered the standard for teaching people to use an application, mostly because a learner gets to watch a task happen instead of reading about it in the abstract.
But "video is engaging" and "video is sufficient" are two different claims, and L&D has a habit of quietly treating the first as proof of the second.
What actually happens when video is the only thing you've got
Nielsen Norman Group ran a study on exactly this question, watching people try to learn tasks from instructional video versus written steps. What they found lines up almost exactly with what I saw in that hallway: people wanted to read the instructions first and only reached for the video once they got stuck on a specific step. When a company made video the only way in, like a support page where a repair step existed solely as a clip, people got frustrated. Not grateful. Frustrated.
That's the binder, in a nutshell. Those techs weren't rejecting their training video. They were reaching for exactly the kind of reference material Nielsen Norman's participants asked for, and building it themselves because nobody else had.
There's clinical evidence pointing in the same direction. A randomized trial published in the National Journal of Medicine looked at whether a stand-alone instructional video could teach practical clinical skills to medical students, and the researchers called the results "controversial" in their own words, meaning mixed at best across studies on whether video alone builds skill that actually sticks. Video is genuinely good at building a mental model of how something works. It's a much shakier bet as the only tool for a skill you need to execute correctly under pressure, months later, with nobody there to rewind the clip for you.
Job aids exist because memorization was never the plan
There's a name for what that binder was trying to be: a job aid, sometimes called a performance support tool. The whole premise is that you don't expect someone to memorize every step. You give them something to consult in the moment instead, which is exactly why solid instructional design programs build job aids into training rather than swapping them out for it. That matters most when a task is complex, high-stakes, or rare enough that memory alone can't be trusted. The University of San Diego has a good rundown on building these well, if you want more on that.
Bob Mosher and Conrad Gottfredson's Five Moments of Need model makes this case cleanly: formal training covers the moments when someone is learning something new or building on what they already know. But three other moments get handled far better by support that's available right when you need it than by another course or another video:
- applying what was learned
- solving a problem when something goes wrong
- adapting to a change
A short onboarding video covers "learning new" well enough. It does almost nothing for the tech six months later, trying to remember which of four nearly identical connectors goes where, under time pressure, with a patient waiting. The eLearning Coach has a great piece on the Five Moments of Need and how to retrain your instincts around what a learner actually needs, worth a look if this framework is new to you.
Why the binder, specifically, was a documentation failure
None of this is an argument against video. It's an argument against treating video like a complete substitute for reference documentation, especially anywhere you've got:
- a task infrequent enough that nobody retains every step from one viewing
- real consequences for getting it wrong (clinical, safety, compliance - take your pick)
- people who need to look something up mid-task without stopping to scrub through a video
- information that changes often enough that someone needs a fast way to update it
A well-built job aid, whether that's a quick reference card, a laminated one-pager, or even a properly maintained binder with real version control, solves all four of those problems in a way video was never built to. And when the aid doesn't exist, the need for it doesn't go away either. People just build a worse version of it themselves, by hand, with a Sharpie, one outdated page at a time.
What I'd have done differently
If I were building that program, I'd have kept the video for what it's actually good at: showing the whole workflow once, end-to-end, so a new tech walks away with a mental model of the process. Then I'd pair it with a genuinely good job aid. Clean graphics of each terminal type, laid out for fast scanning, version controlled properly so nobody's cross-referencing a page from 2015, reviewed on an actual schedule instead of waiting for whoever notices something's wrong this time.
To be clear, this isn't a nostalgia pitch for paper over pixels. It's a performance support argument: match the format to the moment of need, not to whichever format happens to be trendiest this year. Sometimes that's a video. Sometimes it's a page you can flip to in four seconds with gloves on. Good instructional design has room for both. Honestly, it needs both.