Most repackaging campaigns get judged on a feeling. Someone swaps eight thumbnails, checks the channel two days later, sees a number that looks bigger than it did on Tuesday, and calls it a win. Then a month passes, the lift fades, and nobody can say whether the new thumbnails did anything at all.
I run a packaging studio. I have watched this exact loop happen enough times that we now set up the measurement before we touch a single file. Measuring a repackaging campaign is not hard, but it is fiddly, and almost every mistake comes from comparing the wrong two numbers.
Here is how we actually do it, including the parts that are annoying.
What are you actually measuring when you repackage a video?
You are measuring whether a video that was already published starts earning more impressions and converting more of them, without you making anything new. That is the whole bet. The video is fixed. The audience is roughly fixed. The only variable you changed is the packaging, which means the thumbnail, the title, and the promise the first frame makes.
Three numbers carry the campaign:
- Impressions. How often YouTube offered the video to somebody.
- Impressions click through rate. How often that offer got taken.
- Views from those impressions. The number that actually pays you.
People fixate on the middle one. CTR is the most sensitive to packaging, so it moves first and it moves most, which makes it a great early signal and a terrible final scorecard. A thumbnail can lift CTR from 4.1 percent to 5.6 percent and still leave the channel flat, because impressions collapsed at the same time. The reverse happens too. CTR dips slightly while impressions triple, and the video earns more views than it ever did.
So the honest headline metric is views per day from browse and suggested, tracked against the same window before the change. CTR is your diagnostic. Views are your verdict.
How long should you wait before reading the results?
Wait 28 days minimum, and take the first 7 with a large pinch of salt. YouTube treats a packaging change as new information. It will test the new thumbnail on a slightly different slice of people, and that test period is noisy in both directions. The first three days often look amazing. They are not real yet.
Our rule internally is simple:
- Days 1 to 7: watch for disasters only. If CTR drops more than 20 percent relative, we intervene. Otherwise we do nothing.
- Days 8 to 28: this is the read. Compare it to the 28 days immediately before the swap.
- Day 60: confirm. If the lift held for two months, it was packaging, not weather.
The 28 day comparison window matters more than people think. Traffic to older videos is lumpy. One mention somewhere, one seasonal spike, one algorithm mood swing, and a 7 day window tells you a story that is completely fictional. Twenty eight days smooths most of it out. YouTube's own analytics documentation is worth a read on how impressions are counted, because a surprising number of arguments about CTR are really arguments about definitions.
Which comparison should you trust, before and after or against the rest of the catalogue?
Both, and the second one is the tiebreaker. A before and after read on a single video tells you what happened. It does not tell you why. If your whole channel had a good month, every video looks like a winner.
So we always keep a control group. When a client sends eight videos for a thumbnail refresh rather than another new upload, we ask for a list of five or six similar videos from the same era that we are deliberately not touching. Same rough topic, same rough age, same rough original performance. Those become the baseline.
Then the read is a ratio, not a raw number:
- Repackaged videos: views per day up 61 percent over the previous 28 days.
- Untouched control videos: views per day up 14 percent over the same window.
- Attributable lift: roughly 47 points, which is the part the packaging earned.
That subtraction is the single most useful thing in this whole article. Without it you are taking credit for the season.
If you cannot name the videos you deliberately did not change, you do not have a measurement. You have a mood.
What counts as a good result?
For a back catalogue repackage, a relative CTR lift of 20 to 40 percent on the changed videos is a normal good outcome, and the views lift is usually larger than the CTR lift because impressions expand once the click rate improves. A doubling of views per day on a dormant video is common. A doubling of CTR is not.
Some honest context on the shape of results. In a batch of eight, we expect something like this:
- Two videos jump hard. These are usually the ones where the original thumbnail was fighting the title instead of supporting it.
- Four move modestly, somewhere between 10 and 40 percent on views per day.
- One is flat.
- One goes down, and we revise it.
That last one is not a failure of process, it is the process. Unlimited revisions exist because the first read on a thumbnail is a hypothesis. If a thumbnail underperforms its predecessor over a clean 28 day window, we change it. That is cheaper and faster than pretending it worked.
Why does traffic source break the numbers if you ignore it?
Because browse, suggested, and search reward completely different thumbnails, and a blended CTR number hides which one moved. A repackaged video can lose search CTR and gain browse CTR at the same time, netting to zero, and you would swear nothing happened.
Split every read by traffic source. In YouTube Studio the same video will often show 3 percent in search, 5 percent on browse, and 7 percent in suggested, and those are not comparable to each other. They are only comparable to their own past. We wrote about the mechanics of this in more detail in why browse, suggested, and search need different thumbnails, and it is the fastest way to explain a result that looks confusing on the surface.
The practical rule: when you report a repackaging campaign, report CTR per source, and report views total. Mixing those two levels is where most bad conclusions come from.
How do you measure a title change separately from a thumbnail change?
Honestly, you mostly cannot, and I would rather say that plainly than sell you a method that does not survive contact with real data. YouTube gives you one CTR figure per video per source. If you swapped both the thumbnail and the title at once, the number is a joint result.
There are two ways to handle that:
- Stagger them. Change the thumbnail, wait 28 days, read it, then change the title and read again. Slow but clean. Worth doing on your five biggest evergreen videos and nothing else.
- Change both and accept joint attribution. Faster, and usually correct anyway, because a thumbnail and a title work as one unit. This is why we treat packaging as a single object rather than a design deliverable with a copy deliverable stapled on.
For most channels option two is right. The point of repackaging a back catalogue is throughput. You want eight videos fixed this month, not one video studied for a quarter. Title consultation is bundled into our work for exactly this reason, because sending a beautiful thumbnail to sit next to a title that contradicts it is a waste of both.
A/B variants are the exception. If you buy the variant add on we produce a second thumbnail for the same video, and a genuine head to head over equal time windows is the cleanest evidence you will ever get about what a specific audience responds to. It costs 15 dollars. It usually teaches us something we reuse across the next four videos.
What should a repackaging report actually contain?
A one page table and three sentences. Anything longer and nobody reads it, including me.
The table, one row per video:
- Video title and age
- Old CTR and new CTR, split by browse and suggested
- Views per day, previous 28 days and following 28 days
- Control group movement over the same window
- Verdict: keep, revise, or revert
The three sentences: what worked, what did not, what we are changing next round. That last sentence is the one that compounds. Packaging gets better because you keep a record of which faces, which crops, and which contrast choices earned clicks on this specific channel, not on some general theory of thumbnails.
Different formats need different record keeping too. A commentary channel lives on face, reaction, and stakes, which is why thumbnail design for news and commentary reads differently from an education library. And when the catalogue is course led, the win is usually attention flowing to an evergreen playlist, which is the pattern behind our packaging work for course creators.
What tends to cause a bad measurement?
Small windows, missing controls, and reading CTR without impressions. Those three account for nearly every argument I have had with a client about whether a campaign worked.
A few more that catch people:
- Comparing a repackaged video to the channel average. Old videos are not average. They have their own decay curve.
- Ignoring the fact that YouTube reissues impressions unevenly. A video can sit at 400 impressions a day for months, then get 40,000 in a week because the thumbnail finally survived a scan. That is the mechanic behind how viewers scan a feed, and it is why patience beats panic.
- Counting shorts feed impressions in the same bucket. Different surface, different rules.
- Changing thumbnails on all your videos at once. No control group left. You have destroyed your own evidence.
Running the campaign itself
Eight videos is the right batch size, which is why our pricing is built around it. Eight custom thumbnails is 250 dollars, one per video, about 31 dollars each, with a dedicated designer, title consultation, and unlimited revisions. Each thumbnail takes 48 to 72 hours. Nothing is templated, and every style we use has been tested on our own channels before it goes near a client channel.
Eight is also the smallest batch that produces a readable result. Fewer than five and one lucky video dominates the average. More than twelve at once and you have no untouched catalogue left to compare against.
Pick the eight by dormant potential, not by nostalgia. Good candidates have decent watch time, weak CTR, and a topic that is still true. A video with 30 percent average view duration and 2 percent browse CTR is a packaging problem. A video with 8 percent view duration is a content problem and no thumbnail will save it.
If you want the same discipline applied to a long form show, my team also edits video podcasts under the Pixel Samy name, and the measurement habits transfer almost exactly.
We are Dubai based and work worldwide, async, so timezones rarely matter. If you want a look at pricing and services or you just want to send me a list of eight sleepy videos and ask which ones are worth reviving, write to samy at pixelsamy dot com and I will give you a straight answer.