Start 7-day trial
9 min read · ·

How to find outlier posts on Instagram, TikTok and YouTube

An outlier is a post that did far better than its own account normally does. It is the most useful thing you can study, because it shows a decision that worked, not an audience that was already big.

Short answer. To find outlier posts, collect an account's last 30 or so posts, calculate the median views, and divide each post's views by that median. Any post at two times or more is worth studying, and five times or more is rare enough to treat as a strong signal. Use the median and not the mean, because one viral post drags the mean up. Exclude very recent posts that are still gathering views.

Why raw views mislead you

Sort any niche by views and you get a list of big accounts. That list tells you who has an audience. It does not tell you what works. A creator put the real question this way: "why some Reels hit 500k views and others get 2k."

A post with 300,000 views from an account with two million followers can be a weak post for that account. The same 300,000 views from an account with 5,000 followers is extraordinary. Ranked by views, the first one appears on top and the second may never appear. Yet only the second contains a lesson you can use.

Scoring a post against its own account controls for audience size. The creator, the audience and the camera stayed the same. The thing that changed is the post. So the difference in result points at a decision the creator made, such as a topic, an opening or a format.

The maths, in four steps

  1. Collect the last 30 posts from one account, on one format. Keep Reels with Reels and Shorts with Shorts. Long videos and short ones have different baselines.
  2. Drop anything newer than about a week. Recent posts are still growing, so they will look worse than they are.
  3. Find the median. Sort the view counts and take the middle value.
  4. Divide each post's views by the median. That is the outlier score.

A worked example

These numbers are made up to show the calculation. An account's ten most recent settled reels got 2,100, 1,800, 2,400, 1,600, 2,000, 19,000, 2,200, 1,700, 6,300 and 1,900 views.

Post viewsMedianScoreWhat it means
19,0002,0509.3×Strong outlier. Study first.
6,3002,0503.1×Solid outlier. Study second.
2,4002,0501.2×Normal variation. Ignore.
1,6002,0500.8×Normal variation. Ignore.

The mean of those ten numbers is 4,100, more than twice what the account normally gets. Against that number, eight of the ten posts look like failures, which is plainly wrong. This is why the median is the right tool. It is not moved by a single runaway post.

Thresholds that make sense

ScoreHow to read it
Under 1.5×Normal variation. Nothing to learn.
1.5× to 2×Interesting if it repeats. Not enough alone.
2× to 5×A real outlier. Worth a close look.
5× to 10×Strong. Note topic, opening and format.
Over 10×Very strong, or a fluke. Check for outside causes.

These bands are our working judgment and not an industry standard. Set your own once you have seen a few accounts. Some accounts are naturally spiky, so a 3× score means less there than on an account where every post lands close to the median.

Mistakes that give you false outliers

Doing it by hand on each platform

Instagram

Open the profile and go to the Reels tab, which shows play counts on each tile. Copy the numbers into a sheet from newest to oldest, skipping the newest week. Pinned reels sit at the top and break the date order, so watch for them.

TikTok

The profile grid shows view counts. Sort is limited, so take the latest 30 in order. Pinned videos appear first here too. A very spiky account is common on TikTok, so read scores with some care.

YouTube

Use the channel's Shorts and Videos tabs separately. Long videos keep gaining views for longer, so give them a wider settling period than a week. YouTube says its recommendations rely heavily on what each viewer has watched, so a video can take off late as it finds its audience.

Your own account

Your own insights give you more than views. Add saves, sends and follows from each post. An outlier on follows can matter more to you than an outlier on views. See views but no followers.

A full worked example

These numbers are made up to show the method. Here are the views on the last 11 reels of an imaginary fitness coach, in posting order.

2,100 · 1,800 · 2,400 · 9,600 · 1,500 · 2,000 · 2,300 · 1,700 · 31,000 · 2,200 · 1,900

  1. Sort them from low to high: 1,500, 1,700, 1,800, 1,900, 2,000, 2,100, 2,200, 2,300, 2,400, 9,600, 31,000.
  2. Take the middle value. With 11 posts it is the sixth one, so the median is 2,100.
  3. Divide each post by 2,100. The 9,600 post scores about 4.6×. The 31,000 post scores about 14.8×.
  4. Everything else scores between about 0.7× and 1.1×. That is normal variation for this account.

Now see why the average would have misled you. These 11 posts add up to 58,500, so the average is about 5,300. Against that number, nine of the 11 posts look like failures and the 9,600 post scores under 2×. The median shows what actually happened: the account is steady at about 2,000, and two posts broke away.

Eleven posts is below the 20 to 30 we recommend, so treat this as a demonstration. With a real account, you would want more posts before you trust the pattern.

A simple spreadsheet you can reuse

You do not need a tool to start. A sheet with four columns is enough: post link, date, format and views. Put views in column D, starting in row 2. Then add a fifth column for the score.

  1. In an empty cell, type =MEDIAN(D2:D31) to get the median of 30 posts.
  2. In E2, divide the views by that cell. If the median sits in H1, the formula is =D2/$H$1.
  3. Fill the formula down the column.
  4. Sort by column E, highest first.
  5. Add a notes column and write the topic and the opening line for anything above your threshold.

Keep one tab per account. After a few accounts, the notes column becomes the useful part, because you can see the same topics appearing in different places.

When views are the wrong number

Views measure reach. They do not always measure what you want. If your goal is followers or customers, score posts on that instead. The method stays the same and only the column changes.

Your goalNumber to scoreWho can see it
ReachViewsAnyone, for most public posts
FollowersFollows per 1,000 reachedOnly you, in your own insights
Saves and sharesSaves or sends per 1,000 reachedOnly you, in your own insights
LeadsEnquiries per post, tracked by handOnly you
DiscussionComments per 1,000 viewsAnyone, as a rough proxy

For competitors you only have public numbers, so views and comments are your limit. For your own account, use the private ones. We wrote a separate guide on views that do not turn into followers if that is your problem.

How often to run it

Run it on your own account once a month. A reel takes time to finish collecting views, so leave out posts from the last week or so, because their counts are not settled and they will drag your median down.

Run it on competitors whenever you plan a new batch of content. The point is to catch what has changed since last time. A topic that starts to appear across several accounts in the same month deserves more attention than an old hit that everyone has already copied.

What an outlier can and cannot tell you

An outlier shows that something happened. It does not prove why. The topic, the opening, the timing and plain luck all sit inside that one number. One creator made a fair point about this kind of data: "Consistency helps but it's not enough. I think testing different content and doing more of what actually works matters way more." Treat every outlier as a guess about what worked, then test the guess on your next few posts.

What to do with an outlier once you find it

Finding it is half the job. Write a short note for each outlier using the same headings, so you can compare them later.

After 10 to 15 notes across several accounts, patterns appear. That is the base for working out why a post worked and for turning it into your own script.

Doing it with a tool

By hand, one account takes roughly 15 minutes, which is fine for five accounts and painful for fifty. Several tools now automate this. They differ mostly in platform coverage and in what happens after the score. We compare them in the best tools to find viral reels.

When you assess a tool, check three things. Does it score against each account's median or just rank by views? Does it tell you when the sample is too small? And can you see the posts behind the score? A score you cannot inspect is hard to trust.

XCut AI scores every post against its own account's median across Instagram, TikTok, YouTube and Facebook, and shows the multiple on each one. You can run the free Instagram scan or the TikTok version on your own account without signing up.

Frequently asked questions

What is an outlier post?

A post that performed far better than its own account's usual result. It is measured as a multiple of that account's median, so a 5× outlier got five times the normal views.

Why use the median and not the average?

One viral post pulls the average up and makes normal posts look like failures. The median is the middle value, so a single big post barely moves it.

How many posts do I need?

Aim for 20 to 30 from the same format. Below about ten the median is unreliable, and the result should be treated as a hint only.

Can I use likes instead of views?

Yes, the same maths works on any metric. Views are the most comparable across accounts, while saves and sends can tell you more about value on your own account.

Are outliers on big accounts useful?

They can be, because the score already controls for size. But smaller accounts are closer to your situation, so weight them more when you pick what to try.

What if an account has no outliers at all?

Some accounts are very steady, so nothing crosses 2×. Lower your threshold to about 1.5× and look for a repeating theme, or study a different account. A steady account can still teach you what its normal post looks like.

Should I leave my newest posts out?

Yes. Recent posts are still collecting views, so they pull the median down and make older posts look better than they were. Leave out roughly the last week.

References

See which of your posts beat your average

Paste your handle at xcut.ai/instagram for a free scan with no account, or start a trial to research any niche.

Start your 7-day trial

$5 for 7 days · 1,000 credits · cancel anytime