YouTube is the only ad that
keeps paying you after you stop paying.
A Meta ad stops the day your budget stops. A programmatic banner is a one-shot impression. A YouTube integration runs once — and then keeps running. It stays on the channel's page, in the uploads list, in search, in recommendations. People still find it, click it, watch it. The ad you paid $10k for in January is still collecting views in June, September, and deep into next year. You paid once. You get returns forever.
The problem nobody solved.
Every creator tool can tell you view counts. None of them calculate evergreenness at the video level across millions of videos. Why? Because you need three things that are individually hard and together almost impossible:
(1) Historical view snapshots going back years. (2) Time-series data at the article level, not just channel level. (3) The compute to run daily aggregations over billions of rows.
We built all three. Firebolt holds 7.4 billion video snapshots. A batch job recomputes evergreenness daily. Every tracked article in our database has an evergreenness field. Channels have aggregated scores for longform, Shorts, and live streams separately.
Now it's a CLI call.
evergreenness =
(views_180d − views_30d)
────────────────────────
views_30d
interpretation
0.0 – 0.5 weak (plateaus fast)
0.5 – 1.0 ok (some tail)
1.0 – 2.0 good (views doubling)
2.0+ strong
12.0+ exceptional Minimum 5,000 views per video. Minimum publish date Jan 2022. Batch-calculated daily. Honest methodology is part of the product.
The benchmarks.
Pulled from an internal TL research report, longform-vs-shorts-evergreen.md, across the full Firebolt dataset.
"A media buyer using VidIQ knows that longform probably lasts longer. A media buyer using ThoughtLeaders knows that General Knowledge longform lasts 9.5x longer than General Knowledge Shorts."
How to query it yourself.
Method 1: Read the score on each video
The fastest path. tl uploads returns the evergreenness field alongside view count and metadata.
$ tl uploads channel:12465 \ --fields title,view_count,evergreenness \ --limit 5 --md title views evergreen ───────────────────────────── ───── ───────── Every MCU Movie Ranked 8.2M 3.14 Why Avatar 2 Underperformed 4.7M 2.81 The Batman — 1 Year Later 2.9M 1.92 Dune Part Two Review 6.1M 4.02 ← elite Top 10 Tarantino Movies 11.4M 5.71 ← elite
Method 2: Pull the raw view curve
For when you want to compute the CPA drop yourself, or plot the curve, or catch anomalies. tl snapshots video gives you the Firebolt time-series directly.
$ tl snapshots video Jx0gZLS3B6U --channel 12465 --json [ {"age":1, "view_count":108868}, {"age":31, "view_count":1012340}, {"age":91, "view_count":1640289}, {"age":180,"view_count":2303412}, {"age":238,"view_count":2780996} ] # evergreenness = (2.3M − 1.0M) / 1.0M = 1.28 # still earning views 8 months after publish
Why your CFO will care.
Most marketing math gets worse over time. CAC rises as you scale. Targeting decays as audiences saturate. First-click attribution flatters everyone briefly, then breaks. Marketing managers spend their careers explaining to CFOs why the numbers aren't quite what the pitch deck promised.
YouTube sponsorship math is the exception. The effective CPA you see in week one drops — not just a little, but by half, sometimes more — over the next 6 to 12 months. And nobody was measuring it, so nobody was giving you credit for it.
Now you can measure it. You can go back to your CFO with a real chart showing the CPA you booked at day 7 and the CPA that actually played out over day 180. You can recompute ROI on last quarter's sponsorships using today's view counts. You can put a real number on "evergreen tail" for the first time.
That's not a CLI feature. That's a category upgrade for an entire advertising medium.