What are the biggest explainer video mistakes?

The biggest explainer video mistakes happen in planning, before a single frame moves: starting without a falsifiable one-sentence thesis, writing the narration first and illustrating it after, and going to full render without ever reviewing a still frame. The visual failures everyone can point at — dead frames, fake interface, text walls — are mostly downstream symptoms of these process mistakes, which is why fixing the video in the edit never quite works.

We keep a registry of the visual failure modes — this guide is its upstream companion. Across sixty-odd produced videos, nearly every rejected build traced back to one of the six planning mistakes below. Each one comes with its symptom (how it looks from outside), its cost (what it actually burned), and its fix.

1. No falsifiable thesis

The symptom: the video has a topic instead of a claim. Asked what it's about, the team says "our integrations" or "the new dashboard" — a noun, never a sentence. The finished video is a tour: competent, complete, and impossible to remember.

The cost: without a thesis there is no deletion criterion. Every scene is equally justified, so nothing gets cut, the video runs long, and reviewers give contradictory notes because nobody agreed on what "done" argues. The revision rounds that follow aren't polish; they're the team discovering the thesis late, at animation prices.

The fix: before any script, write one declarative sentence the whole video exists to make feel obvious. The test is falsifiability. "Our integrations are powerful" fails — no video could prove it wrong. "You connect a source once and your agent's knowledge updates itself instead of going stale" passes — a viewer can watch for exactly that. A real one from our curriculum: "An agent is just a workflow that can reason." Note it's a claim someone could dispute; that's what makes it a spine. Any scene that doesn't push the sentence toward obvious gets cut, however good it looks. One sentence for the video, then one per scene — the one idea per scene discipline is this rule applied recursively.

2. Describing instead of choosing

The symptom: the video gets specified in adjectives. The brief says "clean, modern, dynamic, not too corporate," the kickoff call produces a mood board, and everyone believes alignment happened. It didn't — five people hold five different videos in their heads, and the words fit all five.

The cost: the misalignment surfaces at the most expensive possible moment, first render, as "this isn't what I imagined." Which is literally true and nobody's fault: adjectives don't constrain pictures. The project then spends its revision budget converging on taste instead of refining a shared target.

The cost has a quieter form too: a beat you can't picture. Our planning rule is blunt about it — if a beat has no visual, the beat isn't understood yet. A scene list full of "we show how easy it is" is description standing in for a decision no one has made.

The fix: decide by picking between concrete artifacts, never by describing a hypothetical one. At every stage there's something cheap and real to choose between: thesis sentences, beat lists, rendered stills, candidate directions. Our production runs on this — quality came from making several complete attempts and selecting, never from polishing a single draft toward a verbal spec. A choice between two frames settles in seconds what three calls about "vibe" cannot. It's the reason our whole intake is a board of twenty rendered directions rather than a discovery meeting: picking is a decision; describing is a deferral.

3. Narration first, visuals bolted on

The symptom: the process starts with a voiceover script. It reads beautifully in the doc. Then the visuals are commissioned to "illustrate" it, and the result is a narrated slideshow — pictures that gesture at the words, timing that never quite lands, text drifting onto the screen to cover for visuals that can't carry the sentence.

The cost: sync, permanently. Narration written to a finished picture can land its key word on the key visual moment; visuals bolted onto fixed narration float beside it. Worse, the cost structure ends up inverted: words are the cheapest layer to revise and pictures the most expensive, so a narration-first pipeline welds the cheap layer underneath the expensive one — every wording change now threatens animation timing.

The fix: author in this order: concept, beats, visuals, narration, audio. The picture is locked first; the voice is written to what's actually on screen and says only what the picture can't show. One deliberate exception — the closing line gets written before anything, because the whole video aims at it (how to end the video covers why). The full authoring order, with the register rules for the prose itself, is in the script guide.

4. Straight to motion — no still review

The symptom: the first thing the reviewer sees is a finished 90-second render. All the money — animation, voiceover, rendering, the reviewer's watch-time — has been spent before anyone checked whether the layout was right, the content was real, or the framing cropped the set piece.

The cost: we measured this one. A production batch of ten videos once went to full render with no intermediate check: nine of ten were rejected. The failures were all visible in a single static frame — wrong surfaces, invented interface, broken framing. Catching a bad frame as a still costs minutes, roughly 5% of the cost of catching it as a finished video.

The fix: install a still gate. Before any motion or narration, render two frames — the money shot and the final frame — and review those. Yes or no takes one glance. After we installed the gate, the next cold batch passed first review five for five. The general principle transfers to any production, in-house or agency: put an approval gate at the cheapest artifact that exposes the most expensive failure class. For explainer videos that artifact is a still frame, and any vendor who resists showing you one before animating is asking you to buy the 9-in-10 batch.

5. Unbounded revision

The symptom: either extreme. "Unlimited revisions" in the contract, or no revision plan at all — just an assumption that the first render will be close. Feedback arrives as a rolling stream of notes, each round patches the previous one, and version three is version one with scar tissue.

The cost: revision isn't the exception you're pricing away; it's half the work. In our production records, about 27% of all mainline commits are revision-shaped, and counting the parallel rejected attempts, roughly half of all effort is post-review response. Our flagship videos took 16 and 24 revisions across three staged versions each. A plan that budgets zero for this doesn't eliminate the cost — it just converts it into scope fights. And unlimited-revision pricing is worse, not better: it prices the chaos in and removes everyone's incentive to make any single round decisive.

The fix: structure revision instead of forbidding or unbounding it. Three rules from our pipeline that transfer whole. Version takes, never overwrite — every script and cut is v1, v2, v3, kept side by side, so rejecting a version costs nothing and comparing them stays possible. Decide restage versus patch honestly — a structurally wrong video needs a diagnosis and a rebuild, and polishing it is the expensive way to keep it wrong; our best videos were re-made, not touched up. And write every rejection down the same day — the note explaining why a take failed is what stops round four from re-shipping round two's mistake. Fixed rounds with a stated change-order policy (wording, pacing, and swaps included; new scenes or a new thesis quoted separately) keep both sides honest about which kind of change is being requested.

6. Wrong altitude for the audience

The symptom: a video for buyers that explains the algorithm, or a video for engineers that explains why efficiency matters. Both are "accurate." Both lose the room — one audience is drowning in mechanism it didn't ask for, the other is being sold to when it came to learn.

The cost: length, mostly, and then attention. Wrong-altitude scenes are the hardest cut to make because each one is individually true and well-made; nobody wants to delete the correct explanation of the retry logic. So the video runs two minutes longer than its job requires, and the viewer it was actually for checks out somewhere in the middle.

The fix: set an altitude ceiling before scripting and name it. Our curriculum rule: a video should carry only enough mechanism to make the product's behavior predictable — never an internals deep-dive. If a scene is explaining how the algorithm works under the hood, it's too low; reframe to what it does and when you'd reach for it. Equally load-bearing is the other list: write down what the video deliberately does not teach, and where each cut topic lives instead (the docs, another video, a screen recording). Scope isn't what you ran out of time for. It's a list of named exclusions, each with a destination.

The pattern underneath all six

Every mistake on this page survives because it feels productive while it's happening. Writing a lush voiceover script feels like progress. Skipping the stills review feels like speed. Unlimited revisions feel like generosity. A tour of every feature feels thorough. The work isn't visibly wrong until the render, which is exactly why the render is the wrong place to find out.

The shared fix is one habit: make the decision at the cheapest artifact that can carry it. A thesis is testable as a sentence. A beat is testable as a described picture. A layout is testable as a still. None of them need a finished video to be judged — and a finished video is the worst place to judge any of them. The teams that ship good explainers aren't the ones that animate better on the first try; they're the ones that never let an undecided question travel downstream to where deciding it costs hours instead of minutes.

FAQ

Which mistake is the most common? Narration-first authoring, by a wide margin — it's the intuitive order, since scripts feel like the natural starting point. The most expensive per incident is skipping the still review, because it lets every other mistake reach the render before anyone sees it.

We already have a finished video with these problems. Patch or redo? Diagnose first. If the thesis and structure are sound and the problems are surface-level (wording, pacing, a bad scene), patch. If the video never had a falsifiable thesis or was built narration-first, a restage from a proper scene list is cheaper than it sounds — polishing a structurally wrong video is the slowest and most expensive way to keep it wrong.

How do I audit a vendor's process for these mistakes before hiring them? Three questions. Ask what the video's one-sentence thesis is — a topic answer instead of a claim is mistake one. Ask what you'll approve before animation starts — if the first reviewable artifact is a finished cut, that's mistake four. Ask what their revision policy covers — "unlimited" is mistake five wearing a bow.

Aren't the visual mistakes — dead holds, fake UI — the real problem? They're the visible problem. When we rebuilt identical topics twice, rejected and accepted takes had similar effort and similar code volume; what separated them was upstream discipline. Fix the process and most of the visual registry stops appearing on its own.

If you'd rather start from a decision than a brief, send us your product's URL — twenty rendered directions to pick from is mistake number two, solved.