I had forty followers and I was about to give up on the channel.

Pop Rocks Radio

Power pop, punk, indie and garage. Podcast, Substack, and a weekly show on Space 101.1 KMGP LP FM, Seattle.

The channel I was ready to abandon

Pop Rocks Radio is an independent music brand. I run all of it — the show, the writing, the editing, the posting.

Facebook was the weakest channel by a distance. Around forty followers and no traction at all. My working assumption was that the platform was wrong for the material.

That assumption was incorrect. The platform was fine. The format was wrong.

What I tested

The first post of the month was forty-seven seconds on The Rezillos: archival Top of the Pops footage, one band, factual text cards over the picture, no voiceover and no commentary. It did 306,869 views against a page with forty followers.

One post proves nothing. The question worth answering was whether the result came from the format or from that particular footage being charming — and the way to find out was to repeat it on bands with far less recognition and see whether the floor held.

So I wrote it down as a specification, ran it five more times across the month, and kept publishing my existing formats unchanged as a control.

The specification

• Length22–61 seconds, vertical
• SourceArchival broadcast footage, 1976–1980, mostly Top of the Pops
• SubjectOne band per video. Never a roundup
• TextText on the picture does all the work — no voiceover, no host, nothing explaining how to feel
• The job of the textReward the viewer for knowing who these people are
• TreatmentWhite type throughout, one consistent grade
• Published10:00 AM, every time
• SequencingOrdered deliberately, each left to plateau before the next

Six for six

1. 1
   Ramones22 sec · Aug 16

   590,413
2. 2
   The Tourists60 sec · Aug 22

   474,693
3. 3
   The Rezillos47 sec · Aug 1

   306,869
4. 4
   The Undertones58 sec · Aug 8

   296,064
5. 5
   Stiff Little Fingers61 sec · Aug 29

   164,244
6. 6
   Buzzcocks45 sec · Aug 15

   144,348

The floor was 144,000. Engagement held at 2.97% across the set — reactions, comments and shares against views — steady as reach scaled rather than thinning out.

The one that broke my own rules

Five of those six used escalating factual cards. The sixth: the Ramones video, and the best-performing post of the month, used comic thought bubbles instead, as a promotion for the channel.

It shouldn't have worked, by my own specification. But the jokes weren't generic: they were specific to each member. Johnny's politics. Joey and health food. Dee Dee and the pizza. If you know the band, the bubbles land. If you don't, they're still funny enough to hold you.

Which is the same job the fact cards do. The cards hand you the knowledge; the bubbles assume you already have it. Different register, identical mechanism.

The finding isn't that fact cards work. It's that archival footage plus text rewarding band knowledge works — and the register can flex from straight to comic without the floor moving.

That matters more than the tidier version would have. A format with only one voice in it is fragile and gets stale. A format with room in it can run for a year.

The control group

This is why I'm confident it was the format and not the algorithm having a good month. Same page, same weeks, same person, different formats, drawn to the same scale as the chart above.

1. Blind Box videosbest of three

   8,300
2. Podcast promotion video

   1,331
3. Carousel postsbest of three

   1,025
4. Text post

   183

The gap isn't incremental. The weakest archival short beat the strongest of everything else by seventeen times.

Where the followers came from

Growth from roughly forty to 6,845, against only 175 unfollows. People weren't bouncing on arrival.

The follow curve has two clear spikes, on 16 August and 21–22 August. Those are the publication dates of the Ramones video and The Tourists. Growth tracked publication, not drift.

98.9% of views came from people who didn't follow the page. This was a discovery engine, not a loop back through an audience I already had.

Who showed up

Top cities: Mexico City, Lima, New York, London, Melbourne, Los Angeles, Santiago, Sydney, São Paulo, Bogotá. In mid-August this audience was 63.5% British. By the end of the month Latin American cities held four of the top ten slots for content about British and Irish bands from 1978. The material traveled considerably further than its subject matter suggested it would.

What I haven't solved

Two million views. 6,845 new followers. Nineteen people who came back.

Average view time ran between 8.8 and 16.9 seconds against videos of 22 to 61 seconds. One-minute views totalled 11,527.

So the honest reading: I built an effective discovery engine attached to almost no retention. The format is exceptional at reaching strangers and unproven at converting them into an audience that returns on purpose.

That's the problem I'd want to work on next, and I have a hypothesis about it. The Ramones video was an episode trailer. My other podcast promotions did 1,331 and 287 views. Same objective, three orders of magnitude apart. The difference is that the trailer stayed inside the format that works instead of stepping outside it to ask for something.

If that generalizes, retention doesn't need a different format. It needs the return path built into the one that already earns attention.

What I'd test next

• LengthThe 22-second video beat everything at 45–61 seconds. That inverts my assumption, and it's the cheapest thing to test.
• RegisterOne comic entry, five straight. Not enough to conclude anything — worth running deliberately rather than by accident.
• PromotionRebuild show and episode promotion as archival shorts rather than as promotional posts.
• Return pathSeries structure, recurring cards, a reason to expect the next one — measured against returning viewers, not views.
• AttributionNothing currently measures what happens after the click. Adding GA4 to the site closes that gap.

What this demonstrates

I had a channel I was ready to abandon. Rather than concluding the platform was wrong, I found the one thing on it that worked, specified it as a repeatable format, held it constant, ran it against a control, and can tell you precisely what it did and did not accomplish, including the post that broke my own rules and why I think it still fits.

The niche is unusual. The method isn't. It works the same way when the subject is a product, a service or a brand.

A dead channel is usually a format problem wearing a platform problem's clothes.