Competitor content analysis: what to actually measure
Most competitor analysis produces a document nobody uses — follower counts, posting frequency, a note about trending audio. Useful analysis answers one question: what did their best posts have that their normal posts didn’t?
Most competitor analysis on social produces a document nobody uses: follower counts, posting frequency, a note that they use trending audio. All true, none of it actionable, because none of it explains why one of their videos did fifteen times what the others did.
Useful analysis answers one question: what did their best-performing posts have that their normal posts didn’t? Everything below is in service of that.
Pick the right competitors first
The instinct is to analyse the biggest accounts in the category. It is the wrong instinct — their results are inflated by audience size, so you cannot separate what worked from who posted it.
Analyse accounts three to ten times your size, with your audience. Ten to fifteen of them. Close enough that their viewers are viewers you could have, far enough ahead that they have already solved problems you have not. Add two or three from an adjacent niche, because formats cross over before they saturate.
One useful test: if you would be pleased to have their commenters, they belong in the set. If their audience is a different market with a different budget, they do not, however impressive the numbers are.
The baseline method
This is the whole technique and it takes two minutes per account.
- Look at their last twenty posts and estimate the median view count — the middle value, not the average, because one runaway hit distorts an average badly.
- Mark anything at two to three times the median or above. Those are outliers.
- Ignore everything else, including their most-viewed post of all time if it is a year old.
An outlier is a video that beat its own account’s ceiling, which means something inside it did the work rather than the audience. That is the only category of post worth your attention, and sorting for it is what separates this from scrolling with a spreadsheet open.
Two traps. A post with an unusual number of comments relative to views is often engagement-farming rather than a strong video. And a recent post is still moving — give anything under 48 hours old time to settle before you judge it.
What to record for each outlier
Six fields, no more. The discipline matters: an open-ended note turns into a review of the content, which is exactly the failure mode.
| Field | What you are capturing |
|---|---|
| First line | The opening spoken sentence, word for word |
| First frame | What is visible and what text is on screen at second zero |
| Time to first cut | Seconds. A proxy for pacing |
| Format | Correction, demonstration, story, list, myth, reaction, comparison |
| Payoff position | Whether the useful part lands early, middle or at the end |
| Send reason | One sentence: why would someone forward this to a specific person? |
That last field is the one people skip and the one that matters most, because sends are the strongest distribution signal Instagram has — see how the Reels algorithm decides who sees a video.
Reading the pattern
Fill the table for twenty or thirty outliers across your reference set and the patterns are usually blunt:
- A format cluster. Six of thirty are corrections. That is your next four videos.
- A pacing floor. Almost every outlier cuts inside two seconds while normal posts run five. That is a rule you can adopt tomorrow.
- A hook family. Most openings name a symptom the viewer already has. That is a template, not a topic.
- A length band. Outliers cluster at 18–25 seconds while the account’s usual output runs 45. Worth copying immediately.
None of those findings are about subject matter, and that is the point. Structure transfers between accounts; subject matter does not.
What never to copy
- The topic. You will make a worse version of a video that already exists.
- The script. Beyond being obvious to their audience and yours, it is their work.
- The personality. Delivery that works because of who someone is does not survive transplant.
- The audio, by default. By the time a sound is identifiable as theirs, joining is late.
The line is simple enough to hold in your head: shapes and structures are common property, specific expressions are not. “Open by naming a symptom, cut at 1.5 seconds, payoff at the halfway mark” is a shape. Their actual sentences are theirs.
How often, and when to stop
Weekly, forty-five minutes, as part of the wider content research loop. Rebuild the reference set every couple of months — accounts plateau, change direction, or get bought.
Stop when the same four findings keep reappearing. At that point you do not need more analysis, you need to film against what you already found. The most common failure in competitor research is not doing too little of it; it is doing it forever instead of posting.
Questions people actually ask
How do I do a competitor content analysis for social media?
Build a set of ten to fifteen accounts three to ten times your size with your audience. For each, estimate the median view count of the last twenty posts and mark anything at two to three times that or above. For each outlier record six fields: first spoken line, first frame, seconds to the first cut, format, where the payoff lands, and why someone would send it on.
Which competitors should I analyse?
Accounts three to ten times your size in your niche, not the biggest ones. Large accounts' numbers are inflated by audience size, so you cannot separate what made a video work from who posted it. A useful test: if you would be pleased to have their commenters, they belong in your set.
Why use the median instead of the average view count?
Because a single runaway hit distorts an average badly, which raises the bar so far that genuine outliers stop registering. The median — the middle value of the last twenty posts — describes the account's normal performance, which is what you are comparing against.
Is it legal to copy a competitor's content format?
Structures and formats are not anyone's property — opening by naming a symptom, cutting at 1.5 seconds, or placing the payoff halfway through are shapes anyone can use. Their specific script, footage and phrasing are their work and copying those is both obvious to viewers and wrong.
How often should I analyse competitors?
About forty-five minutes a week, and rebuild the reference set every couple of months as accounts plateau or change direction. Stop when the same findings keep reappearing — at that point you need to film against what you found rather than keep researching.
Skip the spreadsheet
Runwave computes each competitor’s baseline for you, flags every reel that beat it, and breaks down the hook, pacing and retention shape behind each one — then writes your version in your voice.
Start free for 7 days