Workflow 26: How to Analyze Your Social Media Analytics With AI

Seven years of making content and I still find the numbers confusing.

The individual numbers are fine. The question underneath them is the hard part: across four platforms, with different metrics on each, what should I make more of?

So there's a workflow that pulls the analytics into one place and answers exactly two questions. What to double down on, and what to stop doing.

What this replaces

Posting into the void and hoping.

The trap is having too much data in too many places. YouTube has its own dashboard, Instagram has another, LinkedIn has a third, and none of them talk to each other. So you look at four dashboards, feel vaguely informed, and then post whatever you were going to post anyway.

The other failure is quieter. You keep making a type of content nobody engages with, for months, because nothing ever told you plainly to stop.

Blueprint

Numbers in from everywhere, two answers out.

Trigger: manual, or on a regular review cycle. Inputs: analytics from every platform you publish on, however you can get them. Output: what to make more of, what to make less of.

Two ways the data arrives:

• Connected. Some tools have an MCP connection, so the workflow pulls the numbers itself. I use one that covers YouTube and Instagram.
• Manual. Some platforms still don't. LinkedIn is one. My assistant downloads the export and drops it in a folder, and the workflow reads the folder.

The manual route is unglamorous and completely fine. A folder is a connection.

Organize

What has to exist first:

• A list of the platforms you actually publish on and care about
• A connector for the ones that have one
• One folder where manual exports land, with a habit of putting them there
• Enough history that results mean something
• Some notion of what you're optimizing for, since the platforms all measure differently

That last point is worth sitting with. YouTube cares about watch time, Instagram about saves and shares, LinkedIn about something it won't explain. Comparing them straight across is meaningless. What you can compare is which topics and formats did well relative to your own average on each platform.

Systematize

You are my analytics reader. Look at my content performance and tell me
what's working.

Where the numbers come from:
- [CONNECTED TOOL]: [WHAT IT COVERS]
- [FOLDER PATH]: manual exports from [PLATFORMS]

What I publish: [FORMATS AND TOPICS]
What I'm optimizing for: [e.g. reach, saves, watch time, leads]

Do this:

1. Pull everything available. Say plainly if a platform's data is missing
   or out of date rather than working around it silently.

2. Compare each piece against my own average on that platform, not
   against the other platforms. They measure different things.

3. Tell me what to DOUBLE DOWN ON:
   - Which topics performed above my average, and by how much
   - Which formats performed above my average
   - Any pattern across platforms worth noticing

4. Tell me what to DO LESS OF:
   - Topics that consistently underperform
   - Formats that don't land
   Be direct about this. I'd rather hear it than keep making it.

5. Flag anything you're unsure about. A single post doing well is not a
   pattern. Say when the data is too thin to conclude anything.

Do not tell me what to post. Tell me what happened.

How to set it up

1. List your platforms and check which have a connector.
2. Set up one folder for the rest, and decide who downloads the exports and how often.
3. Run it on the last few months of content you already have.
4. Check its conclusions against what you already suspected. Where they disagree, that's the useful part.
5. Fold the output into whatever you use for planning content.

What if my numbers are small?

Then the "do less of" answer matters more than the "double down" one. With a small sample, a hit could be luck, but a format that never lands over twenty attempts is telling you something real.

Assign

The workflow gathers the numbers, compares them, and reports.

I decide what to make.

The distinction is sharper here than it sounds, because analytics are persuasive. A number saying a topic performed well is not the same as that topic being worth your next three months. Sometimes the best-performing thing is something you don't want to be known for, and following the data off a cliff is a real way to end up with an audience you don't want.

There's also a limit to what numbers can see. They tell you what got watched. They don't tell you what people wanted and didn't get.

This series is the example. It didn't come out of analytics. It came from reading comments, where the same questions kept appearing. How do you do this workflow? How does that process work in your business? Nothing in a dashboard would have surfaced that. People had to say it.

Rule of thumb from this build: analytics tell you what happened. You decide what it means.

Integrate

The analytics reader closes a loop that starts with planning:

• The content planner researches what people are asking and schedules the month
• I record, and the repurposing engine turns one video into everything else
• Those posts go out and generate numbers
• This workflow reads the numbers and says what to make more of
• That answer feeds the next month's planning

Three workflows, one cycle, and each one makes the next better informed.

The part I'd protect if I had to cut everything else is the comments. The numbers refine what's already working. The comments tell you what to build next.

Next: how to actually use AI to get clients, rather than just talking about it.