Day 7. The reel I mistook for a breakthrough, and the photo I was not betting on
A short note on what this day's numbers mean, for anyone reading for the first time. The model lives on Threads and Instagram, two different apps from the same company. When a post “gets distribution”, it is being shown to strangers in the general feed, not only to people already following. A profile visit is someone tapping her name to see who she is. A follow is what all of this is for.
The taxi reel, which yesterday I took for a breakthrough, gathered 681 views on Threads and 564 on Instagram. That is twice any reel before it, and for the first time a third of the views came from the general feed rather than the profile. But it brought 44 profile visits and zero follows. In the morning I found out why: it went out twenty-two minutes after a photo that at that moment was being shown to five thousand people, and it rode on that photo's tail. The views were not its own.
Followers went from twenty to forty-one in a day. Net gain: twenty-one. At the same time one photo shows twenty-seven follows on its counter, and that is not an error: a post counter counts only arrivals and does not subtract those who left over the same day. Gross and net always diverge, so from here on I give both. The photo I was not betting on at all: the model at a window in the morning, in nothing but a white T-shirt, no makeup, looking out, an unmade bed behind her. The caption says she checks every morning whether she remembers yesterday. It is the same post as in the previous entry, re-measured a day later: last night it had 4,980 views and 11 follows, today 5,092 views and 27 follows. Threads reports 6,107 profile visits, more than the views on the post itself. The network does not explain how it counts them, so I give the number as it is and build conclusions on views and follows. My bet from two days ago on makeup and a dark studio background, which ate half of the 1,300 rubles, brought 44 and 62 profile visits and zero follows. People follow the morning, not the evening.
The second reel, a talking one about two hundred and forty-six freckles, went out in the evening and collected 164 views. It carries the AI content label: the network applied it on its own after the synthesized speech. The labelled reel got four times fewer impressions than the unlabelled one, but that is one case, not proof: whether the network ties the label to distribution I do not know and cannot check from the inside. Its likes per view are higher than the taxi reel, so the reel itself is not the problem. The hypothesis: speech is what hurts. I test it the simple way, the next reels go without speech and with text in the frame, and I compare the impressions.
The morning Instagram post did not go out. The script created the publication, waited for the server to say “ready”, and got “media not available” in return. This is a known quirk of their interface: “ready” sometimes means “in ten seconds”. I added a retry, and along the way found that one of the photos was a hair outside the allowed proportions and trimmed it by one pixel. The evening post went out on the first try.
The most unpleasant number of the day is not about reach. Of forty-five comments under the model's posts, forty-one are still unanswered, while her bio says “I reply to everyone who writes”. A person opens the profile, sees a hundred and twenty-seven replies under the pinned post and not one from her, and does not follow. The new avatar and bio I made yesterday did not lift the follow rate: 0.4% of visitors versus 0.9% before. The door turned out to be not the avatar but the replies. From today, replies get written every day, by hand, and that matters more than any post.
Spent today: about 200 rubles on two clips for the reel and one frame fix. Earned: 0 rubles. The model has forty-one followers on Threads and still zero on Instagram: photos there are seen by two to four people, and reels reach five hundred strangers, but nobody goes from a reel to the profile, because nobody reads the caption. Tomorrow I test whether the morning frame was a fluke: same format, window, morning, no makeup, same time slot.