Use Data Or Be Used By Data!

The November 25 issue of Seotistics is here for you!

I've been talking quite a few times about LLMs and the latest developments are exciting!

I spit some blunt facts that you will hardly read elsewhere, you can bet on it. 👀

Oh and all the details and procedures I will include in my products, soon to be updated 😎

Please move this email to your Primary inbox or reply to it. This is to prevent Seotistics goes into spam by accident. Gmail users can read this tutorial to do it.

Read this in your browser

📣📣 Important Announcements 📣📣

The Black Friday Discount is here!

40% off on all of my products from now until December 3 (CET), be quick! 👀

Use the code "BF2025" at checkout.

And... new updates coming this week.

Grab The Special Offer Now!

I had the pleasure to attend MeasureCamp Zurich last Saturday, my presentation is here.

(If you want a recap of the event, go check my forum on Discord).

Now it's time to look forward and think about 2026...

if you want to see me at some conference or have proposals, I am all ears.

My View On LLMs

Anyone can do Analytics, as long as they know what they are doing.

LLMs help, there is 0 doubts about that.

There is a lot of negativity surrounding this topic, coming in particular from "veterans".

Gatekeeping is one of the evils of any industry and what hinders progress and growth.

Ignore the noise and use AI to get better and faster at what you need.

Some countries and/or communities do it better than others.

For example, I have found that the international (digital) Analytics community is quite inclusive and welcoming...

but if we take other disciplines like SEO or Marketing, there is crazy gatekeeping in some countries.

Being defensive or reactive about innovation is the best way to lose your credibility or edge.

Even if you know that is BS, you need to show it with facts and propose a valid alternative.

For Coding

Since we aren't developing production software, LLMs are top tier for coding.

Spending hours on complex SQL is just a memory and the same goes for Javascript and Python.

Not using LLMs creates a massive disadvantage for you in terms of speed.

I've trained many people and you can see the difference is massive with LLMs... they empower users.

Analytics IS about questions and framing problems, not flexing the best code.

Data did NOT even create Analytics or Decision Making.

One of my friends, Matthew Brandt, who also spoke at MeasureCamp ZH last week, talked about decision making.

As he also said, Decision Science originated before data, by a wide margin.

Back to coding now, it's the most efficient way to handle most of the web data problems.

Automation and a proper data infrastructure go a long way...

can you even imagine using Excel to store your most sensitive data? 👀

One of the most simple yet effective wev stacks available.

For Prototyping

Now, this is an underrated use case but probably one of the best ones!

If you are like me and hate spending time on beta versions and stuff, LLMs can massively help you.

Yes, Claude can even write Excalidraw code which you can copy and paste straight into the tool to generate a visual:

Once again, my recommendation is to use Claude because it supports artifacts, these blocks with specific output (as above).

The key about prototyping is SPEED, not accuracy.

I often have drafts or beta versions just to point at the direction I wanna take.

For Documentation

No one wants to actually write documentation... the problem is that it's insanely useful.

Guess what, LLMs do it better than us.

If you are like me and need help with structure, LLMs can do the dirty work for you.

This is even more true when you code 😎

README.md files can also be huge, uh. I'd rather have a detailed file than a synthetic but incomplete one.

Claude Code usually creates some .md file before starting so that it can be referenced later.

This is a very good practice because trust me, you will forget what your code does.

An alternative is to provide some source files to act as references for future documentation.

For Brainstorming

You know, I store my content and other data in Obsidian:

Not even updated, my base is now much much bigger 😎

Then, I can use the Obsidian MCP in Claude to interact with my data.

An MCP (Model Context Protocol) is a way to connect LLMs to other software, that's it.

I can chat with Claude and ask questions or get insights about my Obsidian vault.

You need to read what Obsidian MCP can do as well.

With the right set of prompts, you can get insights about what content to write next or how to group the topics.

And yes, you guessed it, how you can distribute and repurpose your content on different channels:

It's incredible how I always end up talking about this topic... maybe because it's crucial? 👀

The "Stack"

I shared it a few times already and this is what I rely on:

Not all AI but it integrates with AI tools, which is extremely useful.

Maybe you want to use ChatGPT, maybe you like Google AI Studio...

all that matters is that you are effective and efficient.

I can only be happy if my students or clients get results.

The biggest issue I am noticing and what I plan on tackling next are:

communication/understanding
technical work that wastes time

LLMs provide great assistance but they don't suffice.

This is where analysts can shine... acting as a bridge between data and non-technical people.

The reason why I invest a lot of time in content is exactly for this reason.

Most projects are like this:

frame problems
communicate effectively why the problem matters
set up
solve said problem
show that you solved it and what is the outcome

You can easily survive with common sense and the usual boring skills.

What You Can Avoid AI For

There a few areas where AI/LLMs are underperforming.

Or rather, not exactly solving any real issue.

This would be:

writing content. Mainstream quality and zero effort. Use it for copywriting and briefing instead. Not for Content Marketing.
analyzing data. You, the human, are paid to analyze data. Chatting with data won't solve fundamental issues.
replacing what is already solved. Why use AI to cluster keywords when we have super effective methods for that?
calculations. LLMs excel in language and generate words. They don't actually get math :p

Seriously, don't replace what works for the unknown.

We need to use AI.

Probably your manager I guess.

You should be comfortable with being tool agnostic.

The tool doesn't matter as long as you solve the problem (effectiveness).

However, you also want to win at the lowest cost possible (efficiency).

LLMs in most cases are NOT the cheapest and fastest way to do something.