Outlier score for Reels and TikTok: how to calculate it

An outlier score is how many times more views a post got than its creator usually gets. Divide the post's views by the creator's baseline (ideally the median views of their recent earlier posts). A score of 1× is normal, 2× to 3× is notable, and 5× or more is a strong outlier worth studying.

The idea became popular on YouTube, where tools score videos against a channel's typical performance (OutlierKit explains the YouTube version). It works just as well for Instagram Reels, TikTok and Facebook Reels, with a few adjustments covered below.

The formula

Outlier score = post views ÷ baseline views

The baseline is what this creator normally gets. Everything interesting about outlier scoring is in how you define that baseline.

Why median beats average

Most guides tell you to divide by the account's average views. That breaks as soon as an account has one big hit.

Imagine a creator whose last nine posts got around 20,000 views each, plus one post that reached 2,000,000:

• Average of those ten posts: about 218,000 views. Against that, a new post at 60,000 looks like a flop (0.3×).
• Median of those ten posts: about 20,000 views. Against that, the same post at 60,000 is a clear 3× outlier.

The median is the middle value, so one runaway post cannot distort it. For short-form, where single posts regularly explode, the median is the safer choice.

Only compare with older posts

The baseline should include only posts published before the one you are scoring. If you include newer posts, or the post itself, you mix the result into its own benchmark. You also accidentally score posts against a future the creator had not reached yet.

Two more rules keep the baseline honest:

• Skip posts with unknown views. Missing data is not the same as zero views.
• Treat a zero baseline as "no score", not infinity. A brand-new account has no normal yet.

How many posts should the baseline use?

Too few posts and one oddity skews everything. Too many and you compare against a version of the account that no longer exists.

A good range is 10 to 30 recent earlier posts. ViralWatch uses the median of up to 20 earlier posts from the same account on the same platform, and labels scores built from fewer than 10 posts as low confidence.

What counts as a breakout?

There is no universal threshold, but these bands are a sensible starting point:

Add a minimum view floor too. A 3× post on an account that usually gets 800 views reached 2,400 people. That might be interesting, but it is weak evidence.

ViralWatch combines both: a post is flagged as a breakout when it reaches at least 2× the median of up to 20 earlier posts and at least 5,000 views. The feed then sorts by how strong the breakout is, so small accounts and huge ones appear on the same honest scale.

Worked example

Here is the sample breakout we use across the ViralWatch site (an illustrative account, not a real creator):

1. @mia.moves has a median of 24,000 views across her previous 20 Reels.
2. A new Reel reaches 284,000 views.
3. 284,000 ÷ 24,000 = 11.8×.

That clears the 2× threshold easily, and 284,000 is far above the 5,000 view floor. It is a strong breakout and worth studying.

Factor in post age

A post's views keep growing for days. That creates two traps:

• Comparing a 6-hour-old post with posts that have had weeks to grow understates new posts. If a new post is already at 2× after a few hours, it is probably heading higher.
• Comparing a month-old post with brand-new ones can overstate it.

Simple fixes: note each post's age next to its score, and give extra attention to young posts that are already outliers. ViralWatch's breakout score mixes the multiplier with recency for exactly this reason, so a fresh 5× post ranks above an old 5× post.

Look beyond views

The outlier score tells you which posts to open. A few extra ratios tell you what kind of outlier it is:

• Engagement rate: (likes + comments + shares) ÷ views. High means the audience reacted, not just watched.
• Share ratio: shares ÷ views. High usually means the idea travels from person to person, which lines up with what Instagram says helps reach non-followers (Social Media Today).
• Views per hour since posting: a rough sense of momentum.

We go further into which signals matter in what Reels and TikTok reward in 2026.

Does it work the same on every platform?

The formula is identical, but each platform has quirks worth knowing:

• Instagram Reels: view counts include replays and are shown on the Reel itself, which makes Reels easy to score. Only score Reels against other Reels from the same account, not against photos or carousels.
• TikTok: views are public on every video and accounts often post more frequently, so baselines fill up quickly. Pinned videos can be much older than the rest of the grid, so check publication dates rather than grid position.
• Facebook Reels: view counts are available for public Pages, but many Pages post Reels less often, so baselines can be thin. Treat scores built on fewer than 10 earlier Reels as low confidence.

Never compare across platforms. A creator's TikTok baseline and Instagram baseline can differ by an order of magnitude.

A simple spreadsheet setup

If you want to score posts by hand, one tab per tracked account works well:

1. Column A: publication date. Column B: views. Column C: post link.
2. Sort by date, oldest first.
3. In column D, starting from the 11th post, calculate the median of the up to 20 posts above it. In Google Sheets or Excel that is a MEDIAN over the previous rows.
4. In column E, divide views by that median. That is the outlier score.
5. Use conditional formatting to highlight scores of 2 or more where views are also above 5,000.

Update views once a week, because recent posts keep growing.

Common outlier score mistakes

• Using follower count as the baseline. Followers and views are only loosely related on short-form platforms. TikTok has said follower count is not a direct factor in For You recommendations (TikTok newsroom).
• Using the average. One viral hit makes everything else look like a failure.
• Including newer posts in the baseline.
• Treating the score as the reason. A high score says "look here". The reason is in the content.

Calculate it by hand or automatically

By hand, a spreadsheet works: list an account's last 20 posts with views, take the median, and divide each new post by it. It is accurate but slow once you track more than a few accounts.

ViralWatch does the same calculation automatically every time it scans the accounts you track, then ranks breakouts first and explains the likely pattern behind each one. For the wider workflow, see our competitor content research guide. For what to do once you find an outlier, read how to turn a viral reel into an original script.