Use Data Or Be Used By Data!

The March 16 issue of Seotistics is here for you!

If you ever wondered whether you could get more proficient in Analytics with LLMs... this is for you!

You absolutely can start with it.

Not all that glitters is gold though... because many professionals and agencies are still doing it wrong.

P.S. I've published a new article on the Minimal Web Analytics Stack For Small Businesses, go check it out 👀

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 📣📣

I am still busy with making some promo content for the course but this week I am gonna put the preorder out.

This will be for Analysts or everyone wanting to explore Web Data in general, not only SEO.

The topics I am planning to cover are:

Coding in general (Python, SQL, R - all aided by LLMs)
LLMs and how to use them properly
Auditing websites and deeper business considerations (so not only SEO)
The topics covered in this newsletter

If you forgot to vote and want to be included in future communications, please contact me.

I will speak at MeasureCamp Milan this Saturday, let me know if you are going too!

A Recap Of The Recommended Stack

Seotistics officially recommends Claude and Gemini.

If you ask me, pick Claude and be happy. Really.

It's the modern standard for tech and just works wonders.

Many people fail at learning not because they can't do it BUT because they procrastinate.

The best remedy is action, start doing something, get the hype and keep going.

Now let's move right into the gist, how LLMs can help you with stuff:

Most of it is Python and Javascript but don't worry, you can start with 0 knowledge.

Then, you slowly upskill.

Oh and also check this previous issue about a similar topic ;)

Scraping

Ok but why should I even bother with this stuff if I can use a tool like Screaming Frog?

Well, different use cases.

If you need to audit a website, Screaming Frog and Sitebulb are top tier choices, no discussion.

On the other end, if you need to actually retrieve the content for each page... that's a different story!

You need a dedicated tool that is fast enough and can be customized for your use cases.

Needless to say, LLMs help MASSIVELY with it.

I mentioned Crawl4AI before and even in some of my past issues.

Crawl4AI gets the results as markdown files, ready for LLM consumption.

Scraping is still one of the best ways to retrieve content across the web.

If you are working on your portfolio, this is what I often recommend to get some nice data.

Back then, it was a mess because you'd need to know how to scrape yourself:

where to look on the page
what to extract specifically
how to make the data tidy
rendering

It's not as easy as you imagine uh.

Now Claude can write you a functioning scraper in seconds, you just need to test it and find which websites push it.

Web Apps

SaaS has always been a predatory industry.

How many of us have subscribed to some tool with very specific features?

Guess what, most of them are bloat now.

The real differentiator for an actual SaaS is:

security/compliance
scale/architecture
performance

But let's be real, how often do you need an actual software?

Chances are you just need the basics and to use it internally, that's it.

A very simple example of a web app running locally

Quality SaaS have nothing to fear from AI!

If you are a technical user like me, you don't even need a user interface.

You can ask Claude to build a CLI app which in most cases is simple.

They are run from your terminal with a simple command:

and as shown above, you can even decide which arguments to pass to it.

In this case you have one for the number of clusters and another one to filter by clicks and select a CSV.

Dashboards

While it's not true at all that LLMs will replace dashboards, you can get some help to start with them.

And the best use case for me is mockups.

For years the biggest gap was between thinking and execution...

I have an idea in mind but I can't express it.

So we worked on a mockup in software like Miro or Excalidraw.

Now, you don't even need that, you can ask Claude to build a simple web app or even fake data for your Viz tool.

Or if it's something trivial just sketch it yourself inside Claude and then do it in Looker Studio...

The most controversial point is that dashboards are often spammed and used without any regard.

Maybe the business doesn't even need a dashboard...

yet, professionals still sell them because they make them money.

Trust me when I say this, an agency can ask you like $40K for a single dashboard at an enterprise level.

The amount behind a proper dashboard is huge though:

data modeling
user roles (who gets to see what)
filtering and all the data behind
calculations/custom logic

Still, you need at least one to monitor your website(s) and avoid checking 7 billion tools.

Automation

One of the cases where AI actually excels is Automation.

You don't know how many companies are still running on Excel and reek of outdated.

Microsoft has done irreparable damage to a lot of organizations. 💀

In most cases, what I recommend is quite simple yet effective.

Running your content calendar in a spreadsheet? Just use Airtable.

Keeping all of your logic in scattered spreadsheets? Get a proper data stack.

LLMs won't help you for your data architecture but you can get quite far with simple automation.

Some of the most successful examples of automation are:

automating topic research (e.g. scraping competitors, checking SERPs, social listening, etc.)
improving existing processes like auditing or even updating your product catalog
saving hours or weeks of data work and reporting (please don't do it manually)

But... keep *some* things manual too.

Everyone is automating and lowering the bar.

Be different and automate what does NOT add value and wastes your time.

Most of the grunt work falls under this category.

Quality work that requires a human should NOT be outsourced, never!

Actual content advice and strategy can NOT be automated.

Speaking of which, most n8n automations are terribad.

n8n and Make can be great tools but 99% of the content out there is just technical debt.

Please, don't start here, it will take you more time and you will produce headaches.

What's Next?

Analytics is much more than coding or making cool and nice apps.

This is the biggest difference with Engineering, if you ask me.

As an Analytics professional you should aim to optimize decisions, not tools.

In practice this means asking the why behind some actions and grasping the business model.

I am always surprised by the subpar advice on social media, including:

Just publish more content (on an Ecommerce lol), chances are you are getting robbed
Tech SEO for normal content websites, sorry folks, here the answer is publishing more and properly
Just spend all in Ads, it's measurable and fast (wait until you see your CPC skyrocket)

Be different and learn what actually matters: business and common sense.