Guides

How to find out what content works in your niche

Stop guessing what to post. A repeatable research loop: find outlier posts in your niche, name the transferable pattern, and make your version.

Dan Koe··7 min read

The fastest way to find what content works in your niche is to study outliers: posts that massively beat their own creator's normal performance. Outliers strip out audience-size luck and leave the thing that actually traveled, the idea and its packaging. This guide gives you a repeatable 20-minute research loop, manual or with Eden's outlier index doing the heavy lifting.

Full disclosure: we make Eden, and the fast path below uses it. The loop itself is tool-agnostic.

Why most content research fails

Most creators "research" by scrolling their home feed. Three problems:

  • The algorithm shows you what engages you, not what is growing accounts in your niche.
  • Big accounts distort everything. A 2M-follower account gets 5,000 likes on a grocery list. That teaches you nothing.
  • No memory. You see a great post, feel inspired, and lose it forever. Nothing compounds.

The fix for all three is the same: judge posts against their own creator's baseline, search deliberately instead of scrolling, and save what you find into one place.

The 20-minute research loop

  1. Define the niche in search terms, not vibes. "Productivity for developers," "strength training over 40," "solo travel on a budget." Two or three phrasings.
  2. Search for outliers, not top posts. You want posts that did 5x, 10x, 30x their creator's normal. Those are ideas that carried the account, not accounts that carried the idea.
  3. Save everything that clears the bar. Ten to twenty posts into one collection. Do not judge yet, just collect.
  4. Name what repeats. Look across the set: which hook shapes recur (contrarian claim, specific number, before-and-after, list promise)? Which formats (thread, carousel, talking-head short, long caption)? Which angles on the topic?
  5. Write one version. Take the strongest repeating pattern and make it yours: your topic, your voice, your example. Schedule it. Next week, repeat the loop and check how your post did against your own baseline.

Run this weekly and you build two compounding assets: a swipe file of proven patterns and a feedback loop on which patterns work for you.

The fast path in Eden

Eden's Discover feed exists to make steps 2 and 3 nearly instant:

  • Outlier scoring is built in. Every post in the index (millions across X, LinkedIn, Instagram, TikTok, YouTube, and Substack) carries an outlier multiplier against its creator's baseline. Filter your feed to "5x or better in my niche" and step 2 is done.
  • Search by topic, filter by reality. Type what your niche talks about and filter by platform, follower range, and minimum multiplier. You can also search creators by topic and get their matching posts as receipts.
  • One-click saving into boards. Your research lands in a swipe file board, each card keeping its author, metrics, and multiplier.
  • Pattern-naming with help. Ask Eden's chat to read the board: "what do these 15 hooks have in common?" Or go bigger and run Deep Social, an agent that sweeps your niche and returns a receipts-backed report.
  • Close the loop. Draft in Eden (a voice-trained Custom AI can draft in your cadence), schedule it, and check your own analytics next week.

If you work inside Claude or ChatGPT, the whole loop runs there too through the Eden MCP: "find the biggest outliers about home espresso this month, then draft three takes in my voice."

What "works" actually means

One caution that will save you months: an outlier tells you the idea and packaging traveled. It does not tell you it converted. Balance your research set across three jobs: growth posts (reach), trust posts (depth, stories, proof), and offer posts (asks). A niche feed of pure viral bait grows an audience that wants bait. Study outliers within each job.

Frequently asked questions

How do I find what works if my niche is tiny?

Widen one ring. If "hand tool woodworking" is thin, research "woodworking" and filter for the small accounts overperforming. Small-account outliers are the most transferable signal there is, because they won without distribution.

How often should I do content research?

Weekly, in one sitting, 20 to 30 minutes. Daily research becomes procrastination. The point is a steady intake of proven patterns feeding a steady output of your versions.

Is studying other creators' posts copying?

Studying patterns is not copying posts. Every craft works this way: you internalize a hundred examples and write your own with the rhythm in your head. Never republish someone's post, and never imitate one person wholesale. Study the pattern across many.

Can AI just tell me what to post?

Generic AI guesses from training data. Grounded AI is different: Eden's chat and Deep Social answer from actual indexed posts in your niche, with the sources shown. Ask "what should I post" to something that can cite receipts.

Try the loop once. Start free (100 saved items, no card), search your niche, and filter to 5x outliers. Twenty minutes from now you will have a better content plan than most creators guess their way to in a month.