How to Use Photos and Video as a Teaching Tool

Why this matters

Every shop already has thousands of field photos and a crew with cameras in their pockets, and almost none of it teaches anybody anything. The reason is that the shots were taken to prove what was there, and proving and teaching need different frames. Worse, most shops treat visual media as a handoff: send the video, assume the lesson landed. Watching is not learning. The techniques below are about the instruction, not the production, and the highest-yield one costs half an hour and uses footage the tech shoots of himself.

First, the two rules that keep this safe

Nobody films while performing a hazardous step. A phone in one hand during work on energized equipment, a pressurized system, stored energy, or anything at height is a hazard you introduced for the sake of a video. Have a second person hold the camera, mount the phone and step back, or narrate afterward over a still. This gets violated constantly because the tech is alone and wants the shot.

Never publish a capture where a safety step was skipped. If the footage opens with the panel already off and no lockout shown or narrated, throw it out and re-shoot. It does not matter how good the rest is. That clip teaches the shortcut with the authority of whoever is in it, and a newer tech will copy the whole thing, including the part you did not mean to teach.

Step 1: Shoot to teach, which is not how you shoot to document

The same job produces two different photo sets, and using one for the other is the most common reason a shop's photo archive is useless as curriculum.

A documentation photo of an installation shows the equipment, the surroundings, and enough context to establish where and when. A teaching photo of the same install shows one connection, filling the frame, with a hand in shot for scale. A learner given the first one does not know where to look, and looking in the wrong place is exactly what they are being trained out of.

If you skip this, you build a large archive that fails the one time you try to teach from it, and you conclude that photos do not work.

Step 2: Use pairs, because a correct example alone teaches almost nothing

A photo of the correct end state gives a learner nothing to measure it against. They see a thing. They have no idea which of the forty visible attributes is the one that matters.

Put correct and incorrect side by side and the difference does the teaching for you, in seconds, with no text. This is the single most efficient teaching image a shop can make and it is rare, because the incorrect example only exists if somebody thought to shoot the bad one before fixing it.

Make that a standing habit: when you find something wrong, photograph it before you correct it, then photograph the correction from the same angle. Same framing, same distance, same lighting where you can manage it. The angle matching matters more than the image quality; two shots from different positions force the viewer to do mental rotation instead of comparison, and the lesson gets lost in the work.

A pair carries meaning without words, which makes it the strongest asset you have when the crew does not all share a first language.

Step 3: Match the medium to what is being taught

• Video teaches motion, sequence, and sound. Hand position, the order operations have to happen in, what a bearing sounds like as it goes. Anything where the change over time is the content.
• Photos teach state and comparison. What correct looks like, what a failure pattern looks like, what changed. Photos hold still, can be annotated, and can be put next to each other.
• Neither teaches feel. Torque, resistance, the moment a fitting seats, the give in a fastener. That is hands-only and no amount of footage substitutes. Do not try, and tell the learner plainly that this part has to come from doing it, so they do not think they are missing something obvious.
• Sound is underused. Many diagnoses are partly acoustic, and a short library of clips labeled with what each one turned out to be is cheap to build and hard to get any other way. Record close, with equipment noise around you rather than wind.

Step 4: Watch it with them, and stop the tape

Handing someone a video is the weakest possible use of it. Watching passively produces recognition, which feels like learning and is not. The learner will recognize the fault next time they see it in a video and still not diagnose it in a crawlspace.

The technique that converts watching into thinking is to stop before the outcome and ask. In a ninety-second clip, plan three or four stops, roughly one every twenty to thirty seconds. At each stop, two questions:

1. What do you see right now? Forces them to look rather than wait.
2. What do you think happens next, and why? Forces a prediction they will then find out was right or wrong.

A wrong prediction is worth more than a right one, because the correction lands on a belief the learner just committed to out loud. That is the whole mechanism. A learner who watched the same clip alone would have absorbed the answer without ever forming the belief the answer corrects.

More than four or five stops in a short clip turns the session into an interrogation and the learner starts giving safe non-answers. Keep it light and keep moving.

Step 5: Have the tech film themselves, and review it with them

This is the highest-yield technique in the article and the least used.

People correct things on their own footage that they will argue about all day in conversation. Telling a tech he rushes the verification step produces a defense. Watching himself rush the verification step produces a fix, immediately, with no argument, because he saw it.

The setup: pick a task where their rework or callbacks cluster. Have them film one normal occurrence, phone mounted, from a position that shows the work. No performance, no retakes, no explaining to the camera. Then sit down together and run the freeze-and-predict from step 4 on their own footage.

Give them the first pass. Ask what they notice before you name anything. Almost everyone finds most of it themselves, and what they find themselves is what sticks.

A worked review session

A tech has a rework cluster on one task: 4 reworks across the previous 12 occurrences, which is about a third.

He films one occurrence. Four minutes of footage. The review runs about 25 minutes with four stops. Total invested, including the filming, is a little over half an hour.

Across the session, five deviations from the shop standard show up in the footage. He identifies 3 of the 5 himself before the lead names anything, which is typical and is the reason you give the first pass to the tech. The lead names the remaining 2, one of which he disputes until the clip is replayed at the stop point.

Over the next 12 occurrences of that task, the 3 deviations he caught himself do not recur at all. Of the 2 he was told about, 1 comes back once. This is one shop's observation and not a law, but the direction is consistent with why the technique works: a correction you arrive at yourself is attached to a memory of seeing it, and a correction handed to you is attached to a memory of being told.

The rework count over those next 12 occurrences comes in at 1, against 4 in the previous 12. That is a drop from about a third to roughly one in twelve, for about half an hour of combined time and no money spent.

What would change this. If the tech's rework cluster came from a parts or scheduling cause rather than technique, the footage would show nothing and the session would waste an hour. Check the rework causes first. If the tech freezes on camera and performs unnaturally, self-filming backfires; switch to someone else's footage with the freeze-and-predict, and lose the self-identification advantage. And if the deviations turn out to be five variations of one underlying misunderstanding rather than five separate habits, treat it as one competency gap and route it into a pairing, because a review session fixes habits, not missing knowledge.

Step 6: Set the boundaries before this becomes routine

Filming people at work touches a few things that will bite you if you leave them implicit.

• Consent. Everyone in frame knows they are being filmed and agreed to it. Get this straight before self-filming becomes a normal expectation, and check whether any employment agreement or local law in your area constrains recording workers.
• Customer privacy. No identifiable customer information, no house numbers, no documents in frame. A teaching-tight frame usually solves this for free, which is a side benefit of step 1.
• Never use footage to embarrass someone. Showing a tech's mistake to the crew, even lightly, even as a joke, ends the program that day. Self-review footage stays between the tech and the reviewer unless the tech offers it up. If a clip is genuinely valuable as shop-wide curriculum, ask them, and accept no as an answer.

How to verify it is working

• Ask the learner to predict, not to recall. Show a clip they have not seen of the same fault class and stop it early. If they can say what is coming and why, the teaching landed. If they can only tell you what the last video showed, they have memorized a case.
• Watch the specific rework cause, not the total. You reviewed one cluster, so measure that cluster. Total callback counts move with job volume and will tell you nothing.
• Check whether pairs exist in your archive. If every photo in the library is a correct example, nobody is capturing the before, and step 2 is not really running.
• Count sessions, not files. A growing video folder with no review sessions attached to it means the shop went back to treating media as a handoff, which is where it started.

References

• OSHA general industry standards on hazard-specific training and safe work practices during recording
• U.S. Small Business Administration (SBA), guidance on employee training and workplace privacy considerations
• See related: How to Build Training Material With the Tools You Have
• See related: Teaching Through Demonstration, Not Just Telling
• See related: The Mentor Pairing SOP