more bookmarks on their long posts than on everything else they write
From another creator's report, 171 of their 624 postsStop guessing why your posts work.
X shows you what happened. Own Edge tells you what to do about it. Our labelling engine reads every post you have written, scores it on 24 dimensions and hands you the patterns behind your best posts, plus the ones quietly costing you reach.
Free · two minutes · your last 50 posts · a card to share
fewer replies on his quote posts than on everything else he writes
From his own report, 31 posts against 150Sound familiar?
- Which of my posts actually work?
- 600 impressions, then 60. Why?
- Why do some posts get no replies at all?
AI auditAdvice from the posts you paste in
Own EdgeYour posts pulled from the official X API and measured
AI auditTips that fit any account
Own EdgeGroups that beat your own median, with the posts behind each
AI auditA pattern for everything
Own EdgeOnly patterns that survived 1,000 shuffles of your labels
What we have already found in 5,488 posts
One analysis over every labelled post of 15 creator accounts, run through the same pipeline your report runs. Your own report tests these same groups on your posts, where the answers are often different.
Posts that ask nothing get saved
Bookmarks. What people keep is not what they answer, and most creators optimise for the wrong one.
Plain beats loud for replies
Replies. Hyped posts in the same set land 26% below average on reach, the opposite of what the advice says.
No ask, no conversation
Replies. The single most common mistake in the set: more than half the posts ask the reader for nothing.
Short first lines travel further
Reach. Sixty characters or less before the first line break, measured across two thirds of the set.
We read 24 things in every post you write
Each one becomes a group we can measure against the rest of your account.
ok, serious question: if you cannot read the code the model wrote, how do you know it works?
- Post typequestion
- Opening lineclaim
- Who the post talks tothe reader
- Toneplain
- Call to actionopen question
- Time of day12:00 to 16:00
- 18 more
Start free. Pay once if you want the full picture.
Quick Check
- Your latest 50 posts
- Three findings with numbers
- 9 tables
- Shareable result card
Deep Check
- Up to 500 posts plus your replies
- Every feature: type, hook, addressee, tone, timing
- Where your replies land
- A step to take on every finding
- Web report and PDF
Weekly Report
- Your new posts every week
- What changed against last week
- Deep Check included
This is what lands in your account
A real Deep Check of @lucas_builds, 256 posts. Here: where his replies actually land, and what to do with that. From login to report takes about two minutes.

We ran 339 comparisons on those posts and threw away 323
Every faded dot looked like a pattern and was not. 16 survived 1,000 shuffles of the labels, 4 of them pointing up. Your own report goes through exactly this, on your posts. Nobody pads it to look busy.
- +204%posts a reader is not meant to answer
- +177%posts that never say I or my
- +85%plain-toned posts
- +42%posts whose first line is 60 characters or shorter
- −90%posts about building your product
- −56%posts without a call to action
- −61%posts that talk to the reader
- −81%quote posts
- −69%posts of 201 to 280 characters
- −71%posts published on Thursdays
Each dot is one group measured on one metric. Blue sits above the median of the set, red below. The lower a dot sits, the less often chance produced a gap that size.
Questions
Why not just use X analytics?
X gives you 7 numbers for each post and never groups your posts, so it can never tell you what actually drives them. Own Edge turns every post into 24 values and compares each group with the rest of your posts.
Do you post on my behalf?
No. Read access only (tweet.read, users.read, offline.access). Own Edge cannot post, like, follow or send messages.
What exactly do you read?
Your handle and follower count, and your own posts with their metrics. The Deep Check also reads your replies and the follower count of the accounts you replied to.
How are posts classified?
An AI model (Claude by Anthropic) reads the text of each post and fills in 13 fields: post type, opening line, addressee, tone, what the post asks of readers and more. Another 13 are counted from the text alone. The model never sees your numbers.
How do you know a pattern is not chance?
We shuffle the labels across your posts 1,000 times and count how often a gap this large appears by accident, then correct for the number of groups tested. Groups under 8 posts are never compared, and a gap that does not survive is not shown.
Can I delete my data?
Yes, with one click. Account, tokens, posts, labels and reports go, and we revoke our access at X.
Who sees my results?
Only you, unless you create a share link yourself. You can revoke it any time.
