How Followlly measures trends

For each of 8 niches we collect public videos found through a set of niche hashtags on Instagram and TikTok. These pulls run about every 7 days. Each pull is merged into a rolling 21-day window every 6 hours: videos are matched by ID, the latest view counts win, and anything posted more than 21 days ago drops out. Videos without a cover image or with zero views are skipped.

Current window sizes per niche and platform are below. We open niche pages to search engines in stages: a page is indexed only when its niche and platform are on our launch list (right now TikTok fitness, food and beauty; Instagram fitness, food, travel and beauty), and it has at least 30 videos posted in the last 21 days. Only fresh videos count, so a sample that stops refreshing drops below the bar as its videos age out of the window. Every other niche page still renders and states its sample size, but asks search engines not to index it. Linked numbers are the 5 niche pages indexed right now.

All of it is public posts, collected by a third-party data provider. We never log in to a creator’s account or ask for passwords, and we keep only what a public post shows: caption, handle, cover, sound name, hashtags, post time, length and public counts like views and likes. Views are the public counts the platforms show (Instagram reports plays for Reels).

Total views favor old videos. To compare a two-day-old video with a two-week-old one, every video on the trend pages is ranked by views per day:

Age is measured from the post time to the moment the page is built. The 0.5-day floor stops a video that is a few hours old from looking absurdly fast: 30,000 views after 6 hours counts as 60,000 views a day, not 120,000.

Worked example: 120,000 views on a video posted 3 days ago = 40,000 views per day. In the current samples the top video, by @406bbq on TikTok, has 7.5M views at 9.6 days old — about 779K views per day.

Views per day is an average since posting, not the current speed: a video that took off on day one and stopped keeps a high number for a while. For current speed, see “rising now” below.

An outlier is a video that did far better than its creator usually does. For each creator we track, we take their most recent public videos (up to 30; on Instagram, from the past 60 days) and compute the median views of the ones older than 3 days (younger videos are still growing; if there are none, we use the median of all of them).

Example: a creator’s median is 20,000 views; a new video with 75,000 views has a ratio of 3.75 — an outlier. The same 75,000 views from a creator whose median is 60,000 is ordinary (1.25×). That’s why a small account’s breakout can rank above a big account’s average week.

Outliers need a creator’s history, which a hashtag sample doesn’t include — so niche trend pages rank by views per day, and outliers are computed for creators you track in the app (competitors, re-checked every 2–3 days depending on the plan). The 3× threshold is the app’s default and can be adjusted per workspace.

The free outlier finder uses the same median but a looser cut — it lists every video at 2× the median or more — so you can check any public profile and still see smaller breakouts on steady accounts.

“Rising now” looks at speed, not totals. Each time a tracked creator is refreshed we store the view count with a timestamp; the speed of a video is the views it gained between the last two checks (at least an hour apart), per hour. The first time we see a video that is 2–72 hours old, we use its average speed since posting instead.

The creator’s usual pace is the median speed of their other videos from the past 7 days when there are at least three of them (never less than a quarter of the estimate below). Otherwise it’s estimated as their median views spread over 48 hours, since a typical short video collects most of its views in about two days.

Sounds. Within one niche sample on one platform, we group videos by the name of the sound they use and count how many videos use each one, plus the total views of those videos. “Original sound” and “original audio” are excluded — they are unique to each creator, not a shared trend. Sounds are ranked by number of videos, then by views.

In small samples, repeated sounds are rare: right now 69% of sampled niche videos have no named sound, and 5 sounds appear in two or more videos of the same sample. So we don’t label a sound “trending” from one or two uses. Sound lists are framed as the sounds behind the niche’s top videos, show how many videos each one appears in and the reach of those videos, and state the sample size.

Hashtags. We count each hashtag once per video and add up the views of the videos that use it. Generic tags (#fyp, #foryou, #viral, #trending, #reels, #explore and similar), tags shorter than 3 characters and pure numbers are removed. Seed hashtags naturally rank high, because they are how the sample was pulled — the useful signal is the tags that show up next to them.

TikTok’s US trending feed and hashtags. Separately, a third-party TikTok trends collector returns up to 40 videos from TikTok’s US trending feed and up to 30 tags from its US popular-hashtag scan, refreshed every 1–3 days. Videos from it are ranked by views per day like everything else. For hashtags, “videos” is how many of the scanned videos carried the tag and “views” is the total on those videos — counts within the collector’s US popular-hashtag scan, not platform-wide totals. Generic tags like #fyp do appear there — that’s what the scan contains. Latest scan: Oct 3, 2026.

Inside the app, your own videos are judged against your own baseline — the median views of your videos that are 7–120 days old — adjusted for age, because short videos collect most of their views in the first days:

That curve puts a typical video at about 33% of its final views after one day, 70% after three days and 94% after a week — so this morning’s post is compared with your other videos at the same age, not called a flop while it’s still growing. Videos younger than 12 hours are marked “early” and not judged. The hit and flop cut-offs are defaults that can be changed per workspace.

• Samples, not the whole platform. A niche page shows the videos found through a set of niche hashtags, not everything posted in the niche. Videos that don’t use those hashtags aren’t in it.
• The platform decides what we see. Hashtag pulls return what Instagram and TikTok surface for that tag at the time, which leans toward popular and recent posts.
• Language and geography. Seed hashtags are English, and the trending feed is the US one, so the pages lean toward English-speaking, mostly US audiences.
• Counts lag. View counts are as of the last pull (about every 7 days for niche samples, every 1–3 days for the US trending feed). Age keeps growing between pulls, so views per day drifts slightly down until the next refresh.
• Small samples. Some niche × platform samples are small. Pages with fewer than 30 videos still render and say so; they are kept out of search results, as are niche pages not yet on our launch list.
• Correlation, not causation. A sound or hashtag on a top video doesn’t prove it caused the views. Treat lists as places to look, then judge the mechanic yourself.
• Approximate dates. When the trends collector reports a video’s age rather than its post time, the post time is estimated from that age.

The method above only needs public numbers: views, post time, sound name and hashtags of public posts, plus the stats exports app users upload for their own accounts. Nothing on the public pages comes from a private account or a login. Every video shown links to the original post under the creator’s handle; the videos stay on Instagram and TikTok. Creators can have a video, cover or handle taken off Followlly — see removal requests (handled within 24 hours) and the privacy policy.