By the time the seventh platform had the clip, the story was over.
A social desk cutting breaking video for seven surfaces, sequentially, by hand — and losing the news window somewhere around the fourth.
Snapshot
- Segment
- MediaDigital video publisher
- Team size
- 12Social desk, 6 verticals, 4 shows
- Channels
- 7YouTube, TikTok, IG, FB, X, Threads, Snap
- Time to value
- 5 weeksAPI integration with the existing CMS
One cut, seven surfaces, published in parallel
01 — The situation
Before
A newsroom's distribution problem is a clock problem, and this desk was losing on time.
Something breaks at 4:12pm. The video team has a 90-second cut ready by 4:26. From that moment the story is a race, and the desk's job is to get that cut onto seven surfaces before the rest of the internet has said the same thing.
Each surface wants something different. Vertical here, square there, 16:9 for one. Burned captions in one style for the platform where 85% of playback is muted, a different caption style elsewhere. A headline written for a platform where the audience skews twenty years younger than the platform next to it. A thumbnail for two of them and none for the others.
So a producer opened seven tabs and worked down the list. The first platform had the clip by 4:31. The seventh had it around 5:04.
02 — The friction
What it was costing
The desk's throughput was capped by hand-formatting, and the last platform on the list always lost.
Median time from finished cut to live on every surface was 38 minutes. That number hides the real damage: the spread. The first platform got the clip in five minutes and the last one in thirty-eight, so on any fast-moving story the tail platforms were publishing into a conversation that had already formed around someone else's version. Those channels' numbers were consistently weaker, and the desk had begun to read that as 'those platforms don't work for news' rather than 'we arrive there last'.
Across the week, the desk published around 190 items against a video library producing several times that. Archive material — evergreen explainers, show segments with a shelf life measured in years — was almost never republished, because every republish cost the same manual formatting as a breaking clip and breaking always won the queue.
Adding it up, roughly two full-time equivalents out of twelve were spent on re-cropping, re-captioning and re-titling material that had already been produced. On a desk where the constraint is journalists' time during a news cycle, two FTE of mechanical conversion is the entire difference between covering a story and covering it well.
What this was costing
- Median cut → live on all 7 surfaces38 min
- Spread, first platform to last≈33 min
- Items published per week≈190
- Desk capacity on reformatting≈2 FTE
03 — What changed
The reworked workflow
The desk stopped producing seven versions and started reviewing seven versions.
- API + webhooks
The newsroom's own systems trigger publishing
The existing video CMS calls the API when a cut is marked ready. Nobody exports a file and re-uploads it into a separate tool — the clip enters the publishing pipeline from the system the desk already works in. Webhooks send publication and performance events back, so the newsroom dashboard shows distribution status without anyone reporting it.
- Upload once → platform variants
Seven surfaces are produced in parallel, not in sequence
One master produces every aspect ratio, caption style, per-platform headline and thumbnail at the same time. The 33-minute spread between the first platform and the last collapsed, because the seventh is no longer waiting for a person to finish the sixth.
- Approval workflows
An editor reviews the set, once
For a newsroom this is the load-bearing part. Generated headlines and captions are drafts, and a desk editor sees all seven variants on one screen and approves or rewrites before anything publishes. The desk gained speed without giving up the check that stops a bad headline going out under a masthead.
- Cross-platform analytics
Performance is compared across surfaces for the same story
One story's performance on all seven platforms sits in a single view. That is what corrected the desk's assumption about the weaker channels: once the clip arrived everywhere at the same time, two of the surfaces they had written off performed within range of the leaders.
The number that matters to a newsroom is not posts per week. It is minutes between the cut being ready and the audience seeing it, and that is now a review cycle rather than a production cycle.
04 — Results
What moved, and by when
Median cut → live on all surfaces
38 min6 minby week 6
Items published per week
≈190≈415by week 10, no headcount added
Archive clips returned to distribution
≈04,100over 3 months via API
Desk capacity on reformatting
≈2 FTE≈0.4 FTEby week 10
Honest note on what drove this
Two caveats worth stating plainly. First, the six-minute figure is cut-to-live and excludes editorial review time, which the desk deliberately kept and which adds two to four minutes on a contested story — a newsroom that removed the review stage would post faster and should not. Second, the volume increase came substantially from the archive backfill; the sustainable weekly rate for new material settled nearer 260 than 415 once the back catalogue was through.
We had written off two platforms as bad for news. They weren't bad for news. We were arriving there half an hour late, every time.
05 — What to take from this
Transferable lessons
Useful whether or not you ever open NOWScale — these are the parts that generalise.
Measure the spread, not just the median
If your first and last destinations are half an hour apart, your average publishing time is describing an experience nobody has. The tail is where the loss is.
Check whether a weak channel is actually a late channel
Sequential manual publishing systematically penalises whatever sits at the bottom of the list. Before concluding a platform does not suit your content, equalise arrival time and look again.
Automate production, keep the editorial check
Generating seven headlines is a formatting task. Deciding whether a headline is defensible is not. The speed comes from removing the first, and the risk comes from also removing the second.
An archive is inventory you have already paid for
Evergreen material sits unpublished because each republish costs the same manual effort as breaking news, and breaking news always wins. Make republishing near-free and the back catalogue becomes a distribution channel.