Course 09 · Lesson 10 of 13
Analysing What Is Working
Ten minutes at the end of each week closes the feedback loop and tells you exactly what to change
Output without feedback is just guessing on repeat.
You post, you move on, you post again, you move on. After a month you feel like marketing is not working but you cannot say why because you have never looked at what happened and changed anything because of it.
This lesson closes the loop from the first lesson. Ten minutes at the end of each week, Claude reads your results and tells you exactly what to do differently next week. That is it.
What Numbers to Look At
You do not need to become a data analyst. You need three things once a week.
For social posts you need impressions or views showing how many people saw it, engagement covering likes, comments, shares, and saves, and for question posts specifically how many people replied.
For your newsletter you need open rate showing how many subscribers opened it, click rate showing how many clicked a link if you included one, and replies which is the most underrated metric because a reply means someone cared enough to respond.
If you do not have exact numbers, rough descriptions are fine. "This got a lot of comments compared to usual" is usable data. "Nobody engaged with this one" is usable data. You are looking for patterns and not precision.
Where to Find the Numbers
On LinkedIn, click on a post and then "View analytics" below it. This shows impressions, reactions, comments, reposts, and profile visits from the post.
On Twitter and X, each post shows impressions. Click the bar chart icon under a tweet to see full analytics.
On Instagram, switch to a Professional or Creator account which is free. Each post then shows reach and interactions.
On Facebook, page posts show reach and engagement in the Page Insights section.
On Beehiiv, the dashboard shows open rate, click rate, and subscriber growth after each send.
Save these numbers somewhere simple each week, whether a notes file, a spreadsheet row, or a text file. You only need two to four weeks of data before Claude can read the pattern.
The Pattern Worth Finding
Individual post performance is noisy. One post doing better or worse than another tells you very little.
What matters is the pattern across several weeks. Which pillar consistently gets more engagement than the others? Which format including short post, carousel, question, and long post consistently gets the most replies? Which hook style from the hooks lesson keeps showing up in the posts that do well? Which day tends to get the most reach on your platform?
One great week might be luck. Two great weeks in the same pillar with the same format is a signal.
The Weekly Analysis Prompt
At the end of each batch session spend ten minutes running this prompt. Paste your recent posts with rough performance notes:
I want you to analyse what is working in my content. Here are my recent posts
and how they performed:
PILLAR | FORMAT | TOPIC OR IDEA | RESULT (rough or exact)
1. [PILLAR] | [FORMAT] | [TOPIC] | [e.g. "22 likes, 8 comments, 3 DMs" or
"high engagement" or "nobody engaged"]
2. [PILLAR] | [FORMAT] | [TOPIC] | [RESULT] 3. [PILLAR] | [FORMAT] | [TOPIC] | [RESULT] 4. [PILLAR] | [FORMAT] | [TOPIC] | [RESULT] 5. [PILLAR] | [FORMAT] | [TOPIC] | [RESULT]
[continue for all posts from the last 2 to 4 weeks]
My four content pillars: [FROM MARKETING.MD]
Look for patterns and answer these directly:
- Which pillar is consistently getting more engagement?
- Which format (short, long, carousel, question) gets the most interaction?
- Is there a pattern in the hooks that worked versus the ones that did not?
- What should I do more of next week?
- What should I do less of or stop doing?
Be specific. Tell me what to change — not just what happened.
How to Use the Output
Write down the one change Claude recommends most clearly. Then apply it to next week's batch session before you forget.
If it says your stance posts consistently outperform your teach posts, pull two stance ideas from your backlog next week instead of one.
If it says short punchy posts get more replies than long posts on your platform, write fewer long posts and more short ones.
If it says your curiosity-gap hooks outperform direct-address hooks, ask Claude for more curiosity-gap hooks in the next hooks session.
What to Never Change Based on Analytics
Do not change your audience. If your show-proof posts are underperforming that does not mean you have the wrong audience. It means proof posts might need a stronger hook, a more specific example, or a different format. Your audience stays fixed for months.
Do not change your one message. One bad week does not mean your message is wrong. You are looking for patterns and not reacting to individual results.
Do not chase platform trends. If a format you hate is trending, that is not a reason to start using it. Stay on the formats you can produce consistently and well.
The weekly analysis changes tactics covering which pillar, which format, and which hook style. It never changes strategy. Strategy stays fixed for quarters and not weeks.
Your Action Right Now
At the end of this week's batch session, fill in the analysis prompt with your last set of posts and results. Write down the one thing Claude tells you to change. Apply it in next week's session.
Do this every week. After four to six weeks of data, Claude's pattern reads become genuinely useful and your content performance will improve in a predictable systematic way rather than by accident.
Do it with other builders
Post your homework. Get unstuck.
Every course ends with homework. Post yours in Claude Codex Mastery, the community this course comes from, and get feedback from Yar and 400+ people building the same way.
Join the community · $29/mo