Angelica.
← all resources

Find every post that worked

threads breakdown

Scrolling someone's profile to study what works is a waste of an afternoon. Threads shows you the recent, not the best, and you end up remembering whatever you happened to read last. Twenty minutes of the method below gives you a spreadsheet of every post that landed, sorted, with the text attached.

Full disclosure before you read further: the tool in this guide is Sortaic, and I built it. I built it because I was doing this by hand and it was taking hours. Use it or don't — the method in the middle of this page is the part that matters, and it works with any export.
start here, not with the tool

Build the watchlist first

Five to eight accounts. Not fifty. You are going to read these properly, more than once, and fifty means you will read none of them.

  1. Same audience, different niche. The most useful accounts are the ones talking to your people about something else. Copy someone in your own niche and you produce a slightly worse version of their post. Take a structure from a different field and it arrives as new.
  2. Somewhere near your size, plus one or two above. Accounts with hundreds of thousands of followers run on different physics. Their throwaway post outperforms your best one and tells you nothing you can use. Include one as a stretch, not five.
  3. Replies that look like conversation. Open three of their posts and read the comments. If the replies are emoji and "this!", the account is getting agreement. If people are telling their own stories underneath, that account knows how to open a door. Those are the ones worth studying.
  4. At least one you find slightly irritating. Someone who clearly works, whose style you would not choose. They are the ones you learn mechanics from, because you are not tempted to copy the voice.
Write the list down somewhere. You will re-run this every few months and the value compounds: the second pull shows you what is new, what died, and what they quietly stopped doing.
the mechanical part

Pulling the data

Sortaic is a Chrome extension. Desktop only, no account, no login, nothing to set up. It reads the page you are already looking at.

  1. Install it. Go to sortaic.com and add it to Chrome. [[serve: link diretto Chrome Web Store, se vuoi saltare il passaggio dal sito]]
  2. Open the first profile on your watchlist. On desktop, in Chrome, at threads.net. Not the app.
  3. Run Sortaic on the page. [[serve: testo esatto del bottone/pannello che vedono, es. "clicca l'icona nella toolbar" oppure "il pannello compare in alto a destra"]]
  4. Sort by likes. You will change this in a minute, but likes first gives you the widest view of what travelled.
  5. Export the CSV. The export carries the post text, likes, replies, shares, timestamp and a direct link to each post. The text and the link are the two columns you will actually live in.
  6. Repeat for every account on the list. Keep them as separate files. Merging them hides the thing you most want to see, which is what one person does repeatedly.
Free covers the last 25 posts or a 7-day window, which is enough to run this whole method once on one account and decide whether it is worth it. Pro is a one-time $27 and lifts it to a much deeper history, which is what you want when you are pulling eight accounts and looking for patterns across months rather than weeks. Start free. Genuinely.
where most people stop too early

Reading the spreadsheet

The temptation is to open the CSV, look at the top row, and go write something like it. The top post is the least useful row in the file. It is usually an outlier: a moment, a reshare, an argument. You cannot reproduce it and you should not try.

What you want is the top ten as a group. Add three columns by hand and fill them in. It takes fifteen minutes per account and it is the entire exercise.

Column A — what kind of post is it?Story, opinion, list, question, confession, result, correction. Use a short fixed vocabulary and reuse it, or you cannot count anything at the end.
Column B — what does the first line do?Names a feeling, states a number, makes a claim, admits something, sets a scene. This is the one that transfers best to your own writing.
Column C — what does it ask of the reader?Nothing, agree, disagree, tell me yours, go do this. Most high-performing posts ask for exactly one thing. Most low-performing ones ask for nothing or three things.

Now sort your top ten by each column and count. You are not looking for the best post. You are looking for the thing that shows up six times out of ten while its author probably could not tell you they do it.

the number nobody looks at

Replies divided by likes

Add one more column: replies ÷ likes. Then sort the whole export by it.

Likes measure agreement. Someone read a sentence, nodded, moved on. Replies measure that the post left a hole the reader wanted to fill. A post with 800 likes and 12 replies is a poster. A post with 300 likes and 90 replies is a conversation, and it is worth more to you in almost every way that matters: more reach from the replies themselves, more people who now recognise your name, more material for the next post.

Sorting by likes shows you what travels. Sorting by this ratio shows you what connects. If you are building something you eventually want to sell, the second list is the one to study.

Run the same calculation on your own export. Most people find their highest-ratio posts are ones they nearly did not publish, and their lowest are the ones they worked on longest. That comparison alone is worth doing once.
the transfer

Turning a shape into your post

Here is the rule that keeps this from being theft, and also the reason it works better than copying: keep the structure, replace every fact.

What you take is the skeleton: a post opens by admitting something the author got wrong, spends three lines on what it cost, and ends with what they do differently now. That structure belongs to nobody. What you must not take is the admission, the cost or the change — those are theirs, and borrowed, they read as hollow, because they are.

Their post, reworded. Same mistake, same numbers, same turn at the end, different adjectives.
Their structure, your material. The mistake you actually made, what it actually cost, what you actually changed.
  1. Pick one post from your top ten. Write its structure out in five lines, as instructions, with no content in them at all. If you cannot do that, you have not understood why it worked.
  2. Put the original away. Do not have it open. Looking at it while you write is how you end up with the same sentences.
  3. Fill the structure from your own history. Your mistake, your cost, your change. If nothing fits, the structure is wrong for you — pick another one instead of inventing a story.
  4. Read your version and the original side by side, once, at the end. If a phrase survived from theirs, cut it. If the rhythm is identical, break it somewhere on purpose.
If you find yourself struggling to fill the structure because nothing in your life matches it, that is useful information and not a failure. It means you are studying someone whose material is genuinely unlike yours. Note it and move down the list. Forcing it produces the specific kind of post that reads as borrowed, and readers are far better at detecting that than anyone expects.
Do this on your own account too. My highest replies-to-likes post took four minutes to write and I nearly deleted it. The one I spent a whole morning on is somewhere near the bottom.
what to do with the rest

The posts that did not work

Sort ascending. Read the bottom twenty from one account.

This is the half of the export nobody opens, and it is where the fastest lesson is. Almost always, a good creator's worst posts share one property: they are about the creator's own interests rather than the reader's problem. Announcements. Meta-commentary about posting. Photos of their desk.

You will recognise your own drafts in there. That recognition is the cheapest editing improvement available — it costs one afternoon and it stops you publishing the same category of post for a year.

how often

Do this quarterly, not weekly

Every three months is right. Often enough to catch a shift, rare enough that you spend the time between actually writing instead of analysing. The second pull is more valuable than the first, because you can diff them: what is new, what they dropped, what has quietly stopped working.

And run it on yourself every time. Your own export, same three columns, same ratio. Studying eight other people and never your own numbers is the most common way this exercise turns into procrastination that feels like strategy.

Sortaic is at sortaic.com — free tier, no login, and I read every feature request that comes in, because there is nobody else to read them.
More free resources Work with me