I need to find what is actually working in my market.

Ch. 01 What it is

Most people study their market by scrolling it — they see what's loud, mistake it for what's working, and copy the wrong thing. The fix is to stop reading the feed and start ranking it: pull the actual pieces, score each against what's normal for that account, and keep only the ones beating their own baseline. What's left is the real signal — and a swipe file you can mine instead of a memory you'll lose.

Ch. 02 The three ways to build it

Simplest path first. Every tier carries its real setup time and its honest trade-off — the cost is the part most write-ups leave out.

1
Tier 1 · simplest path
A swipe file you keep by hand
Setup~30 min
- a Notion / Airtable board
- or a plain doc
Pick six to ten accounts that sell what you sell, and a board to hold what you find. Every time a post or offer obviously over-performs theirs — far more comments than usual, a hook that stops you, a guarantee you haven't seen — you save it: the link, a screenshot, and one line on *why it worked*. That last line is the whole point. A swipe file of links you never annotated is a graveyard; a swipe file with a 'why' on every entry is a pattern library. After thirty or forty entries you start seeing the same moves repeat across accounts — and the repeats are the things actually working, not the one-off that happened to go viral.
The trade-off
It only catches what you happen to see. You're sampling your own feed, which the platform already skews toward what it thinks you'll engage with — so you'll over-index on a few loud accounts and miss the quieter operator who's converting better than all of them. And 'over-performs' is a gut call at this tier; you have no baseline yet, just instinct. Good enough to start, and on a small, focused market it often beats any tool. But know that you're seeing a slice, not the field.
2
Tier 2
A scrape that ranks against the account's own norm
Setup~half day
- a public-data scraper (ScrapeCreators / Apify)
- a sheet
Now you stop relying on what crosses your feed. Pull the recent public posts for your seed accounts on a schedule, and rank each one *against that account's own baseline* — a 200k-view reel on a million-follower account is a miss; the same on a 5k account is a breakout. Raw likes lie because they track follower count; engagement-versus-norm tells you which piece actually outran expectations. Keep the breakouts, drop the rest, and log each one once so a re-run never re-shows you the same post. You go from 'what I noticed' to a ranked, de-duplicated list of the pieces that genuinely beat their own room.
The trade-off
Engagement is not conversion — this is the trap the whole tier rests on. A reel can win the algorithm and sell nothing; a quiet carousel can drive every booking. Ranking by reach tells you what got *attention*, which is a real signal for hooks and formats, but it cannot tell you what made money. Read it as 'what's pulling eyeballs,' never 'what's pulling buyers,' and keep the offer-and-funnel reads (price, guarantee, where the link sends) separate from the reach reads. Public data only — no logged-in or gated sources, no scraping where the terms forbid it.
3
Tier 3
A standing intel engine that surfaces the winners for you
Setup~1–2 days
- the scraper + a scheduler
- a tagging step (LLM-assisted)
- a feedback signal from your own posts
The scrape becomes a loop that runs without you. Two lanes feed it: a *proven* lane of accounts you already trust, and an *explorer* lane that scans the broader niche by topic and surfaces creators you don't follow yet — and when one of them clears the bar enough times, it promotes itself into the proven lane. Each kept piece gets tagged by what it teaches — offer move, hook pattern, funnel mechanic, the exact words the audience uses in the comments — and routed to where you'd actually use it. Once your own posts are live, their performance feeds back in and re-weights what the engine looks for. It stops being research you do and becomes intel that arrives, sorted, on a cadence.
The trade-off
Two things break this if you let them. First, trends need a *time-series* to see — a single run is a snapshot, and 'what's rising vs. fading' only appears once you've been capturing for weeks, so don't expect the trend read on day one. Second, an automated tagger will confidently mislabel and surface the wrong piece at the wrong moment; keep a human glance on what it routes until the tags prove out. The explorer lane also drifts off-niche without a tight topic definition. Worth building only once Tiers 1–2 are visibly straining under volume you can't keep up with by hand — most markets never need this.

Ch. 03 The detail

Most people study their market by scrolling it — they see what's loud, mistake it for what's working, and copy the wrong thing. The fix is to stop reading the feed and start ranking it: pull the actual pieces, score each against what's normal for that account, and keep only the ones beating their own baseline. What's left is the real signal — and a swipe file you can mine instead of a memory you'll lose.

Category: RevOps · Research & intel
Format: Engine
Level: intermediate
Provenance: Own-packaged

researchmarket-intelcompetitiveswipe-filerevops

Why the feed lies to you

The default way to learn a market is to live in its feed and absorb it. The problem is that a feed is not a sample of what works — it’s a sample of what’s loud, sorted by an algorithm tuned to your own attention. You end up studying a handful of accounts the platform decided you’d engage with, mistaking volume for validity, and copying the move that went viral once instead of the move that quietly converts every week. Worse, none of it is written down, so the pattern you half-noticed last month is gone by the time you sit down to use it.

Finding what’s actually working is a different act. It’s not consumption — it’s measurement. You pull the real pieces, you rank each against what’s normal for the account that posted it, and you keep only what beat its own baseline. Then you write down why each survivor worked, so a stack of observations becomes a library of patterns you can mine. Everything below is a way to do that measurement, and the tiers differ only in how much of the pulling, ranking, and sorting you do by hand versus hand off.

Rank against the room, not the raw number

The single mechanic that separates real intel from feed-scrolling is this: judge every piece against the account’s own norm. Likes and views track follower count, so a big account’s average post will always out-number a small account’s breakout — and if you rank by raw reach, you’ll conclude the big accounts are “what’s working” when all you’ve measured is who’s bigger. The breakout you want — the piece that beat its own expected performance — is the one carrying a transferable lesson. Tier 1 estimates that norm by gut; Tier 2 computes it; Tier 3 keeps computing it on a schedule. The rest is bookkeeping.

The honest line: attention is not money

The trade-off that runs through all three tiers, and the one most “competitor research” tools won’t say out loud: engagement measures attention, not conversion. A piece can dominate the algorithm and sell nothing; a flat-looking carousel can drive the whole funnel. So read the reach signals for what they are — hooks, formats, and angles that pull eyeballs — and read offers, prices, guarantees, and funnel paths on a separate track, because those are where the buying actually lives. The engine surfaces what’s getting noticed. Whether it’s getting bought is a question you answer by watching your own numbers once you act on what you found.

Start at Tier 1. A hand-kept swipe file with a real “why” on every entry will teach you more about your market in two weeks than any dashboard, and on a tight niche it often beats the tooling outright. Climb only when the volume of what’s worth tracking outruns your hands — not before.

The honest version. Each tier buys you something and costs you something — both are stated plainly, never buried.

Tier 1 · A swipe file you keep by hand

It only catches what you happen to see. You're sampling your own feed, which the platform already skews toward what it thinks you'll engage with — so you'll over-index on a few loud accounts and miss the quieter operator who's converting better than all of them. And 'over-performs' is a gut call at this tier; you have no baseline yet, just instinct. Good enough to start, and on a small, focused market it often beats any tool. But know that you're seeing a slice, not the field.
Tier 2 · A scrape that ranks against the account's own norm

Engagement is not conversion — this is the trap the whole tier rests on. A reel can win the algorithm and sell nothing; a quiet carousel can drive every booking. Ranking by reach tells you what got *attention*, which is a real signal for hooks and formats, but it cannot tell you what made money. Read it as 'what's pulling eyeballs,' never 'what's pulling buyers,' and keep the offer-and-funnel reads (price, guarantee, where the link sends) separate from the reach reads. Public data only — no logged-in or gated sources, no scraping where the terms forbid it.
Tier 3 · A standing intel engine that surfaces the winners for you

Two things break this if you let them. First, trends need a *time-series* to see — a single run is a snapshot, and 'what's rising vs. fading' only appears once you've been capturing for weeks, so don't expect the trend read on day one. Second, an automated tagger will confidently mislabel and surface the wrong piece at the wrong moment; keep a human glance on what it routes until the tags prove out. The explorer lane also drifts off-niche without a tight topic definition. Worth building only once Tiers 1–2 are visibly straining under volume you can't keep up with by hand — most markets never need this.

Edition June 2026 · Updated June 20, 2026