How to analyse a competitor by figures rather than by feel
Scrolling somebody's account for twenty minutes is not analysis: you remembered what you personally liked. The real question is which subjects beat their own average, and eyes cannot answer it. Below is how that gets counted.
Scrolling somebody's account for twenty minutes is not analysis: you remembered what you personally liked. The real question is which subjects beat their own average, and eyes cannot answer it. Below is how that gets counted.
---
"It seems to work for them" is not analysis
The standard way of studying a competitor: open their account, scroll for twenty minutes, note that "theirs feels livelier" and "they post more often".
Not one decision can be made from that. You looked at what came up on the first screen and remembered what you personally liked.
The real question is different: which subjects and formats get a response above their own average. Not above yours, not above "the market" — above their usual level. Only that separates luck from a pattern.
Answering it by eye is impossible. You need figures for every post, and you need a body of them rather than the last dozen.
What such an analysis showed for me
I collected 6 502 posts from eleven channels over 19 months and counted the shares by subject.
The result surprised everybody in the market. Posts with price guidance: nine. Nought point one per cent of the feed. About a first commission: seven. About a service as paid work: twenty.
Meanwhile the "analysis of mistakes" format showed 14% hot posts against an 11.6% baseline and worked for eleven channels out of eleven. The strongest measured format in the niche — and almost nobody uses it systematically.
Not one of those conclusions is available from reading a feed. They can only be counted.
And my whole content strategy is built on those two figures: I write about what is massively needed and massively unspoken.
How this gets done without manual work
The AI Workflow section has a block that collects data from social media. It goes at the start of a chain: real figures get pulled in first and the remaining steps work from them.
Six sources
| Scraper | What it collects |
|---|---|
| Instagram Reels | recent reels from a public account: views, likes, comments, direct links |
| TikTok Videos | a creator's videos: views, likes, comments, shares, duration |
| TikTok Profile | the profile: statistics, followers, description, avatar |
| YouTube Shorts | a channel's Shorts: views, likes, comments, dates, metadata |
| Facebook Reels | a page's reels: descriptions, views, reactions, comments |
| Twitter / X Posts | a user's posts: text, views, likes, replies, reposts |
The key point: they return not only the text but the metrics. That is material suitable for analysis rather than for reading.
They work with public accounts.
The whole chain
A scraper on its own gives you a table. The value appears when there are steps behind it.
Step 1. Collection. The scraper pulls the recent posts of the account you want, with metrics.
Step 2. Analysis. A chat model gets the body of data and looks for a pattern: which hooks appear in the first seconds, which subjects get a response above that same account's average, which length works better.
The phrasing of the request matters here: the comparison has to be against their own median rather than against absolute numbers. A post with a million views on an account with a million followers is normal rather than lucky.
Step 3. Generation. The model makes your versions along the logic it found — not copies but material on the same principle.
Step 4. Publishing. The publishing block sends the finished pieces to Instagram, TikTok, YouTube, Facebook, X, LinkedIn, Threads or Pinterest.
The whole loop inside one chain: collected → analysed → made your own → published.
Three uses that pay back immediately
Analysing competitors continuously. Not once before a strategy session but regularly. A niche changes, and what worked six months ago is now the norm.
Content from your own statistics. Pull your own posts with metrics, find the most successful and ask the model to make new ones on the same logic. The most underrated use: you already have data on what works for you specifically and you are not using it.
Monitoring the niche on a schedule. The chain collects recent videos from several accounts itself and sends a summary. With no scrolling of feeds.
The last works on a schedule — that is the Full plan.
What scrapers do not do
Honestly, so expectations do not run high.
They do not read private accounts. They work with public ones.
They do not explain the cause. They show that this post landed and that one did not. Why is the work of the next step and of you.
They do not replace the choice. The data will show what works for a competitor. Doing the same thing is not always the right decision: sometimes the most valuable thing in an analysis is finding a subject they do not have, as happened to me with prices.
And they are not about copying. The analysis exists to understand the mechanics rather than to repeat somebody's post in your own words.
How this sells
Competitive analysis as a service in itself. It used to be expensive precisely because it ran into person-hours: somebody had to compile hundreds of posts into a table by hand. Now the collection takes minutes and the work stays in the interpretation — that is, in what people pay for.
What gets delivered to the client: a table of every post with its metrics, the shares by subject and format, what gets a response above their own average, the open niches — subjects that are needed and that nobody has — and format recommendations justified by figures.
How to approach a client. Not with an offer of analysis but with an analysis of their own account already done. Three precise observations gathered before the first conversation work better than any commercial proposal.
What is needed on the platform
AI Workflow — the section where chains get assembled. Scrapers, chat models, generation, publishing — all as blocks on one canvas. Thirteen ready templates, so you do not start from a blank page.
Saving in your profile. An assembled chain stays with you permanently: open it, substitute a new account, run it. Build it once and use it permanently.
Scheduled runs — the Full plan. The chain collects data daily by itself.
The quote before the run. You see what the whole run will cost and what will be left on the balance. Charging follows the steps actually completed, and money comes back automatically for anything not delivered.
The run continues on the server even if you close the tab.
Registration is free and opens three days of full Basic access — every assistant, the prompt texts, the skill files and six foundational express courses, including "Perplexity as an investigative tool" on the method of verifying data.
Where to start today
One action. Take one competitor's account and collect their last thirty posts with metrics. Do not read them — count them.
Work out the median views and find the ones one and a half times above it. Usually three to five out of thirty, and they have something in common.
That something is the answer to the question you spent twenty minutes trying to get by scrolling a feed.
---
A ready tool for finding open niches
Analysing a competitor answers the question "what are they doing". There is a tool for a different question — where is there any space at all.
The finder for hidden profitable niches on YouTube from the automations archive works like this: you enter a query, cap the maximum age of a channel in days and set the search depth.
The system finds videos from the last fortnight with maximum views, requests the channels' statistics, calculates their age and keeps only those younger than your limit. Duplicates get removed and the result lands in a spreadsheet with links.
The logic of the strategy. If a channel is twenty or thirty days old and its video has hundreds of thousands of views, the subject has enormous potential. You find not where you would be competing with giants but where even a newcomer with no history gets traffic.
What that gives you for competitor analysis. Instead of analysing one well-known player, you get a list of young channels that broke through in weeks — that is, fresh working approaches rather than established ones.
The archive — N8N and Full.
---
What to read next
[Build once, run always](/en/blog/build-once-run-always) — how to assemble a chain like this.
[Content gets made while you sleep](/en/blog/content-while-you-sleep) — how to put the analysis on a schedule.
[Finding trends](/en/blog/finding-trends) — what to do with what you found.
[Audience segmentation as a service](/en/blog/audience-segmentation-service) — how to sell an analysis like this.
[n8n and Make automation templates](/en/blog/n8n-automation-templates-guide) — the full guide to automations.