A creator wrote to me last year with a spreadsheet of 340 videos, every one of them colour coded, and the question at the top was simply "how many of these should I redo?" His plan was to repackage all 340 over six weeks. That plan dies in week two. It always does.
So let me give the number I actually give people, and then the reasoning behind it, because the number changes depending on how big your library is and how much time you personally can spend reviewing work.
For most channels, the answer is 6 to 10 old videos per month. Not more. That is roughly two per week, which is slow enough that you can look at every result properly and fast enough that you feel the effect inside a quarter.
Why not repackage everything at once?
Because you learn nothing and you buy a lot of risk. When you swap 40 thumbnails on the same Tuesday, you have no idea which change moved the needle. Search and suggested traffic shift together, seasonality moves, one video gets picked up somewhere, and your analytics turn into soup.
There is a practical cost too. Every repackaged video needs you in the loop. You have to approve the concept, react to the draft, sometimes say "the face is wrong, it looks smug". That is real attention. Ten of those a month is a comfortable amount of decision making. Forty is a second job, and the quality of your feedback collapses somewhere around video twelve.
And there is a third reason that people underrate. A thumbnail change resets how a video is being served. YouTube's systems re-test the video against fresh impressions, which means the first week after a swap is noisy. If you change everything simultaneously, you have no stable baseline left to compare against. Google's own documentation on the impressions and click through rate metrics in YouTube's help resources for creators is worth reading once, specifically the part about how CTR varies enormously by traffic source. That is the reason a single blended number lies to you.
What is the actual math behind 6 to 10 per month?
Start with impressions, not with your feelings about which video deserved better. Pull the last 90 days, sort by impressions, and look at the top 30 videos. That is your candidate pool, and it is almost never the pool people expect.
Now apply three filters.
- Still getting served. If a video earned under about 2,000 impressions in 90 days, a new thumbnail has almost nothing to work with. You are polishing a video nobody is being offered.
- Underperforming its own neighbours. Compare each video's CTR against the median CTR of your channel for the same traffic source. A video sitting three points below its neighbours is a packaging problem, not a content problem.
- Content still stands up. If the information is stale or the audio is rough, no thumbnail saves it. Repackaging amplifies what is there. It does not repair it.
What usually survives all three filters is somewhere between 8 and 25 videos. If you work through that at 8 per month, you finish the meaningful part of your catalogue in one to three months, then you move to a maintenance rhythm. I wrote a longer breakdown of the ordering logic in the post on choosing which old videos to repackage first, because sequence matters more than volume.
The point of a monthly cadence is not throughput. It is that you can still tell which change caused which result.
How does channel size change the number?
It changes it a lot, and mostly through how fast you get readable data.
Under 20k subscribers. Do 4 to 6 per month. Your videos take longer to accumulate impressions, so results arrive slowly. Going faster just means you are guessing with fewer data points. At this size I would rather see you fix the four videos that carry your search traffic and leave the rest alone.
20k to 200k subscribers. This is the sweet spot for 8 per month. You have enough impression volume that a repackaged video shows a direction within 10 to 14 days, and you likely have 15 to 30 videos genuinely worth redoing. A gaming channel around 60k subs and a personal finance channel around 80k subs will both fit this band comfortably, even though their packaging conventions are nothing alike.
Above 200k subscribers, or over 300 videos. Now the constraint flips. You have more worthy videos than you can review, so the bottleneck is your approval time, not the design. Channels at this size usually run 8 per month as a permanent standing lane rather than a project, and some run two lanes. I covered the operational side of that in the piece on running a repackaging service across a large library, including how to avoid the situation where nobody remembers which videos were already done.
What does a sane monthly cycle look like?
Here is the rhythm I recommend, and it fits neatly into a calendar month.
- Days 1 to 3. Pull analytics, pick the batch, write one sentence per video explaining why it is underperforming. That sentence is the brief.
- Days 3 to 14. Design and revision. Each custom thumbnail takes 48 to 72 hours, so a batch of 8 moves in parallel and lands well inside two weeks. Titles get revisited in the same pass, because a thumbnail written against the wrong title is wasted work.
- Day 15. Swap all 8 on the same day. One swap date per month keeps your before and after windows clean.
- Days 15 to 30. Leave them alone. Resist the urge to tweak on day three. Early numbers after a swap are unreliable.
- End of month. Compare each video's CTR and views per day against its own prior 30 days. Keep, revert, or queue a variant.
That last step is the one everyone skips. Repackaging without a review loop is just redecorating.
Should you repackage new videos too, or only old ones?
Repackage old ones first, almost always. A new video gets one shot at a launch window, and you should package it properly on day one instead of fixing it later. Old videos are different. They already have watch history, they already sit in suggested feeds, and they have proven that people finish them. That is why the back catalogue is usually the cheapest growth a channel has available.
The economics are blunt. A new video costs you a script, a shoot, an edit and a week. A repackaged old video costs you one thumbnail and a title decision. At $250 for 8 custom thumbnails, that is about $31 per video, with unlimited revisions and title consultation included. Compared against the cost of producing another upload, it is not close.
Podcast libraries are the extreme version of this. Long episodes with weak packaging tend to hold enormous unrealised traffic, which is why I treat repackaging old podcast episodes as its own workflow with different rules about faces and guest names. If the underlying edit also needs work, my sister studio edits video podcasts separately.
How many should you repackage before deciding it works?
Eight. That is the honest minimum before you draw a conclusion, and it is not an accident that our batch size matches it.
With one repackaged video you have an anecdote. With three you have noise. With eight you start to see a pattern: which style of framing lifted CTR, whether your audience responds to faces or to text, whether the title carried the click or the image did. That pattern is the actual deliverable. The thumbnails are just how you acquire it.
This is also why niche conventions matter so much. A science channel and an anime channel need opposite treatments, and testing eight in the wrong visual language teaches you the wrong lesson. If you want to see how the reasoning differs by category, the notes on thumbnail design for science channels show how much the frame changes when the subject is an idea rather than a character.
What if you do not know which videos to start with?
Then do not guess, audit. Ranking a catalogue takes a couple of hours of analytics work and it removes the biggest failure mode in repackaging, which is spending money on videos that were never going to be served again. A packaging audit gives you an ordered list, and then the monthly cadence becomes mechanical. You just take the top 8 off the list each month.
One more caution. Do not repackage a video twice in the same quarter. It is tempting when the first swap underperforms, but you need at least 30 days of clean data before you touch it again. If a video genuinely needs a second angle, buy the A/B variant at $15 and run the two properly instead of overwriting your own experiment.
The number, one more time
Eight old videos per month, sequenced by impressions, swapped on a single day, reviewed 30 days later. Fewer if you are under 20k subs. A standing lane rather than a project if you are over 200k or past 300 uploads.
Everything we make is custom per video, never a template, and the styles we use are tested on our own channels before they go near a client. If you want a look at the range or the current terms, the pricing and services page has both, and you are welcome to send me your top 30 by impressions and I will tell you which 8 I would start with.