The hook teardown

Somebody else already paid for the experiment. Go read the results.

Every post in your feed is a test that already ran. Most people swipe the ones with the biggest numbers, which is exactly how you end up copying the wrong thing... so here's the system I use instead, and the two tools that do the math for you.

Why swiping fails

The biggest number is usually the worst hook to steal.

A post with 3,000 reactions can be a flop and a post with 47 can be the best thing that account wrote all month. It depends entirely on what normal looks like for them.

Measure against their own median, not against each other.

Take one account's last 14 days, find the middle number, and anything that doubles it gets pulled. Sounds fussy - but it's the difference between reading a result and reading a follower count.

Throw out anything under 24 hours old.

It's still collecting. A post from this morning sitting at 3 reactions isn't a failure yet (it's just early), and leaving it in drags the median down for everybody else.

Throw out reposts too.

Somebody else wrote that hook. You're reading their writing, not the account's, and it'll skew what you think this person is good at.

Winning and being stealable are two different things.

The best performing post I found in my last teardown beat its own baseline by 23x. It was a photo of a restaurant sign about locking your phone in a box for free cream puffs... there's literally nothing in it you could use.

Tool one: find the bar

Paste the engagement numbers from one account's last two weeks, one per line. Reactions plus comments, added together. I'll hand you the median and the number anything has to beat.

Those nine are real, pulled from one home service founder's last 14 days.

0
Posts counted
0
Their median
0
The 2x bar

Tool two: can you actually use it?

Clearing the bar only proves it worked for them. Whether it'll work for you is a completely different question, and it's the one everybody skips. Five checks, tick the ones the hook passes.

0 / 10
Tick what the hook does
Nothing scored yet.

Copy and paste

The four prompts, in order

Run them start to finish and you'll come out the other side with one hook, scored, rewritten in your own voice. Paste them into whatever AI you already use.

The receipts

What it turned up last time

Six accounts, 14 days, 54 posts read by hand (LinkedIn strips posts out of the page as you scroll, so there was no shortcut). Eleven cleared their own 2x bar. Here's every one of them, scored.

MultipleHookAdaptability
23.5xA restaurant sign about locking your phone up for free cream puffs4 / 10
7.1xTen years since a PTO shaft caught his hoodie. Lying there he realised the business depended entirely on him and there were no real systems7 / 10
5.0xEvery year the event sells out, and last year we turned people away at the door3 / 10
4.2xA summer team photo2 / 10
3.6xAfter almost four years, we finally become profitable next month2 / 10
2.4xMe in 2010, telling everyone to answer every question their customers ask3 / 10
2.3xMeet this year's grant recipients1 / 10
2.3xWe raised the franchise fee for new single-unit owners by 40%3 / 10
2.3xI asked a legendary conversion pro why AI traffic converts so well2 / 10
2.1xHanging with my bestie2 / 10
2.1xThe most important, least appreciated metric. Skipped over in meetings. Brushed aside by specialists.Winner8 / 10
The trap
23.5x

The best performing post of all 54 scored 4 out of 10. Chasing the biggest multiple would have handed me a napkin.

The surprise
3

Three of the eight accounts I started with had nothing usable. One hadn't posted in FIVE years while releasing a podcast episode that same week.

One more thing

Hooks are the easy part.

The expensive problem is further down. You write something good, somebody actually clicks, and then the lead lands in a form nobody checks... or it never gets recorded at all, and the week just looks quiet.

I posed as a customer at a gym once and found three of those in about five minutes. The owner had no idea... nobody does, and that's the whole problem.

Find the leak, it takes 5 minutes

Anything on this page that's wrong, tell me and I'll fix it. I'd rather be corrected than confident.