I have watched the same thing happen on a dozen channels. Part 1 does 180k views. Part 2 does 34k. Part 3 does 11k, and by then the creator has quietly decided that series do not work on their channel.
Series work fine. What usually fails is the packaging. A multi-part series asks the viewer to do something no standalone video asks: commit twice. And most creators package part 2 as if it were a sequel that people are already waiting for, when in reality the vast majority of the people who will ever click part 2 have never seen part 1 at all.
That single fact should change almost every decision you make.
What is different about packaging for a multi-part series?
The difference is that you are packaging two audiences at once, and they want opposite things. Returning viewers want continuity, proof this is the same thing they liked, and a signal of where they left off. New viewers want a video that stands on its own and does not punish them for arriving late.
A standalone video only has to answer one question: is this worth my next twelve minutes. A series entry has to answer that, plus a second one: do I need to have watched something else first. If the answer to the second question looks like yes, a big slice of your potential audience leaves before the video even loads.
So the job is a set of packages that look like family without looking like homework. Same world, same visual grammar, but each one closes its own loop.
That is a design problem, not a labelling problem. Which is why the standard fix, sticking a big number in the corner, tends to make things worse.
Do episode numbers in thumbnails hurt click-through rate?
Mostly yes, and the reason is simple. A number is a barrier disguised as an organiser. "Part 3" tells a new viewer that two things exist which they have not seen. It is honest, it is tidy, and it costs you clicks from exactly the people you need most.
I am not absolutist about it. Numbers earn their place in three situations:
- The series is a genuine countdown or ranking where the number is the hook, not the index.
- Your channel is small enough that the series is the brand, and the audience arrives already inside it.
- The content is sequential in a way viewers actively want tracked, like a build or a challenge with a running total.
Outside those cases, put the number in the title, not the thumbnail. Titles are read after the image has already done its work, and by then the viewer has decided they want the thing. A number in the title reads as depth. The same number stamped across the image reads as a prerequisite.
Better still, replace the number with a stage. "The part nobody survives" beats "Part 4" every time, because it promises a specific moment rather than a position in a queue.
A series number tells people where they are. A stage tells people what they are about to get. Only one of those is a reason to click.
How do you make a series look like a set without looking repetitive?
You lock two elements and vary everything else. That is the whole method, and it is what separates a system from a template.
When we build a series set at ClickTheory, we usually lock:
- One structural constant. Subject placement, crop, or the shape the composition sits in. Something the eye recognises at thumbnail size before it reads anything.
- One color constant. A single anchor color that appears in every entry, ideally one that is not already dominant on the channel's homepage neighbours.
Then we deliberately vary the subject, the expression, the object in frame, and the text. If part 2 looks like part 1 with a different word on it, the browse page treats it as a duplicate and the viewer's eye skips it. Repetition is not consistency. Consistency is a repeated rule with different outputs.
The test I use is the shelf test. Line all the thumbnails up in a row at about 210 pixels wide. You should be able to tell instantly that they belong together, and just as instantly which one you have not seen. If both are not true, the set is not finished. You can see how that plays out across finished sets in our work.
This is also where the discipline of designing per video matters. We never run a series off a template with swapped text, because templates optimise for production speed and series need the opposite: the same rule expressed freshly each time.
Should every part be watchable on its own?
Yes, and the packaging has to say so out loud. The video can absolutely reward continuity. The package cannot depend on it.
Practically that means each entry gets its own complete promise. Not "the story continues", but a self-contained claim: a specific outcome, a specific question, a specific reveal. If your part 3 title only makes sense to someone who watched part 2, you have built a package that can only be clicked by people you already have.
The same logic runs through long-form storytelling generally. Much of what I covered about packaging a video essay applies here, because a multi-part series is essentially a video essay that had to be cut into pieces, and each piece still needs its own reason to exist.
One useful reframe: stop thinking of your series as episodes and start thinking of it as chapters of a book that people can open anywhere. Each chapter title should be interesting cold. YouTube's own guidance for creators over on the YouTube creators hub makes a similar point about browse behaviour, and it matches what I see in the numbers. Most series entries are discovered, not resumed.
How should you package part 2 differently from part 1?
Part 1 sells the premise. Part 2 sells the escalation. They are not the same sales job, and packaging them identically is why the drop happens.
Here is the pattern I use:
- Part 1 carries the widest, most accessible promise. It should work for someone who has never heard of the topic. This is your recruitment package.
- Part 2 raises stakes visibly. Something in the image should be worse, bigger, further along, or more surprising than part 1. The viewer who saw part 1 gets a payoff signal. The viewer who did not still sees an interesting image.
- Part 3 and beyond go specific. By now you have a narrower, more committed audience, so the promise can be sharper and stranger. This is where the best-performing entries in a series usually live.
The common mistake is doing this backwards: burning the most dramatic image on part 1 and then having nothing left. Plan the whole arc before you design the first one. When we take on a series, the title consultation covers the full run, not just the video in front of us, precisely because that ordering decision is worth more than any single design choice.
How do you know if the series packaging is actually working?
Compare each part against its own expected reach, not against part 1. Part 1 usually had a wider top of funnel by design, so measuring part 3 against it will tell you almost nothing useful.
The signals worth watching:
- Impressions on part 2 relative to subscriber count, which tells you whether YouTube is still offering it to browse.
- The share of part 2 traffic coming from browse and suggested rather than from the channel page. High channel page share means only existing fans are finding it.
- Whether average view duration holds across the set. If it drops sharply, the problem is content structure, not packaging.
And please be careful with thumbnail tests here. Series entries often run at lower volume than your flagship uploads, which means a test can look decisive long before it actually is. I wrote about that trap in detail in when a thumbnail test is statistically meaningless, and series videos are the single most common place I see people fooled by small numbers.
What about series you already published?
This is the cheapest win available to most channels, and almost nobody takes it. A finished series is a set of videos that already have watch time data, already have a proven premise, and are already sitting there being ignored because parts 2 through 5 were packaged as sequels.
Repackaging a back catalogue series often beats making a new one. The content already works. The doors are just badly labelled. I would rather rebuild the packaging on a six part series that underperformed than shoot a seventh part into the same wall.
When we do this, the order is usually: fix part 2 first, because it is the biggest leak, then part 1 to widen the top, then the rest as a set. Fixing the tail before the leak is wasted effort.
What does this cost to get done properly?
A series is the one case where buying a batch actually matches the shape of the work. Our standard package is $250 for 8 custom thumbnails, roughly $31 each, one per video, with unlimited revisions, a dedicated designer, and title consultation included. Turnaround is 48 to 72 hours per thumbnail.
Eight is a convenient number for a series because it covers most runs outright, or covers a run plus the two or three back catalogue entries you want rebuilt at the same time. Add-ons exist where they help: an A/B variant at $15 for the entry you want to test, a title pack at $25 when the naming across the run needs real work, and rush at $20 when something is already scheduled.
If you want the fuller breakdown of what studios charge and why, I have written about how much does a YouTube packaging agency cost and, more simply, what thumbnails should cost. If you are still deciding whether an agency is the right call at all, what a YouTube packaging agency actually does covers the scope honestly, including the parts that are not design.
Every style we use gets tested on our own channels before it goes near a client's series, and nothing is templated. Full details sit on our pricing and services page.
If you have a series that stalled after part 1, send me the links and I will tell you where the leak is.