Quality Control for Content Teams: Our 4-Gate Publish System

Every film we publish passes four gates before it goes live: fact-check, script table-read, visual review, final watch. No gate, no upload. That rule holds for a flagship Blackfiles episode and for week-one content on our newest channel. Quality control for content teams isn't a vibe or a final skim — it's a sequence of named checkpoints, each with an owner and the authority to stop a release.

We run four channels shipping weekly, with 200+ films and 60M+ views behind the system. At that cadence, quality doesn't come from heroics or from one obsessive founder watching everything at 2 a.m. It comes from gates. Here's each of ours, with the actual checklists.

Why Quality Control for Content Teams Needs Gates

Most teams "review" content. Few gate it. The difference: a review is an opinion pass that happens when someone has time; a gate is a defined checkpoint with an owner, a checklist, and a binary outcome. Pass or fail. There is no "mostly fine, ship it."

In our genre the stakes are concrete. Our films cover real people — FBI agents, fraudsters, escapees, survivors. A factual error in an investigative documentary isn't a typo; it's a credibility hit, an angry comment section, and occasionally legal exposure. On that last point, talk to a lawyer, not a blog — this isn't legal advice.

Every gate in our pipeline has three properties. An owner who isn't the person who made the thing, wherever our ~25-person team allows it. A written checklist that evolves. And a status that Cortex, our production orchestration tool, tracks — so a film physically can't move to the next stage with an open gate.

Gate 1: The Fact-Check

We put 16–20 hours of research into every film before scripting starts. The fact-check gate isn't that research — it's the adversarial pass that happens after the script draft exists. The checker's job is to attack the script: every name, date, number, and claim gets traced back to a source.

• Every name, date, place, and figure verified against at least two independent sources
• Direct quotes traced to a primary source; paraphrases clearly marked as paraphrases
• Any claim about a living person double-sourced and reviewed separately
• Numbers sanity-checked: do the amounts, distances, and timelines actually add up?
• Disputed facts framed as disputed in the script, never asserted flat
• Source list archived with the project so we can answer challenges months later

A failed fact-check sends the script back to research, not to the editor with a note. We learned that the hard way: "we'll soften that line in the edit" is how errors survive. The fix happens at the layer where the error lives.

Gate 2: The Script Table-Read

A script that reads well on the page can die out loud. Our narration uses a directed AI voice, which makes this gate more important, not less — the voice will perform exactly what's written, including the clunky sentence you skimmed past. So every script gets read aloud before it goes anywhere near voice.

The table-read is a retention pass as much as a language pass. Our episodes run 20–37 minutes, and nobody finishes a 30-minute film out of politeness. We're listening for the exact spots where a viewer's thumb starts drifting toward the next video.

• First 30 seconds: does the cold open state the stakes, or warm up to them?
• Read aloud at full speed — flag every sentence the reader stumbles on
• An open question or unresolved tension every 2–3 minutes
• No paragraph explains something the audience already inferred
• Tone check: investigative, not breathless — we don't write like a true-crime trailer
• The ending earns the runtime: a payoff, not a fade-out summary

Gate 3: Visual Review

Everything on screen is original 3D animation — zero stock footage — with shots produced through Vertex, our generative image and video pipeline. Generative tooling raises throughput, which means visual QC matters more than it used to, not less. One mangled hand or anachronistic prop in frame 4,000 undoes the trust the other 3,999 built.

• Character consistency: same face, build, and wardrobe in every scene they appear in
• Era and place accuracy: vehicles, uniforms, technology, and signage match the period
• Artifact sweep: hands, on-screen text, reflections, background figures
• Shot-to-script match: the visual shows what the narration claims, scene by scene
• Readability at phone size — key details must survive a 6-inch screen
• Sensitive moments handled with restraint; we depict, we don't exploit

Visual review happens scene by scene during production, not as one giant pass at the end. A continuity break in scene 12 is cheap to fix on Tuesday and expensive on publish day.

Gate 4: The Final Watch

The last gate is brutally simple: one person who didn't touch the edit watches the finished film start to finish, headphones on, no timeline open, no skipping. They watch it the way a viewer will — full screen, real time. If the editor reviews their own final, this gate is decorative.

• Full watch-through, no scrubbing, in one sitting
• Audio mix: narration clear over music at phone-speaker volume
• Captions present, synced, and spell-checked
• The title and thumbnail promise matches what the film actually delivers
• Metadata, end screen, and pinned comment ready before upload, not after
• Gut check: would we show this film to the people who are in it?

Anyone running the final watch can stop a publish, and we've slipped upload slots for less than you'd think — a muddy mix, a thumbnail that oversold the story. A late film costs one slot in the weekly schedule. A wrong film costs trust we spent 200 films building.

Making Quality Control for Content Teams Scale

Gates only work if they survive growth. Three things keep ours intact across four channels and weekly uploads. First, separation: the person who made an asset never gates their own work. Second, instrumentation: Cortex tracks every film's gate status, so "did this get fact-checked?" is a lookup, not a Slack archaeology dig.

Third — and this is the part most teams skip — the checklists are living documents. Every error that reaches the public, and every near-miss caught at a gate, becomes a new checklist line within the week. The checklist is the studio's scar tissue: every line exists because something went wrong once, and we decided it would never go wrong silently again.

Total overhead: the four gates add roughly a production day per film. That sounds expensive until you price the alternative — a public correction, a re-edit, or a quiet delisting. We teach the full gate system with templates inside Sentris Academy, but the structure above is everything you need to start this week.

FAQ: Quality Control for Content Teams

How much time should QC add to a production cycle? Budget roughly 10–15% of total production time across all four gates. Most of it runs in parallel with other work — fact-check during early visual production, table-read before voice. Only the final watch is a hard serial step at the end.

Who should own QC on a small team? Whoever didn't make the thing. On a two-person team, swap: you check their script, they check yours. Solo, separate the hats with time — write today, table-read tomorrow morning, and never publish the same day you finish the edit.

What happens when a gate fails near a deadline? The deadline moves. We publish weekly on every channel and we've still slipped slots, because the schedule is a promise to the algorithm but accuracy is a promise to the audience. Only one of those forgives you.

Does AI tooling change the QC equation? It raises the stakes. Generative pipelines like our Vertex multiply output, which multiplies the surface area for errors. The gates are what let us keep the speed without inheriting the sloppiness.