Use Data Or Be Used By Data!

The March 4 issue of Seotistics is here for you!

Ahhh reporting!

An area where many SEOs struggle and that is still under-discussed.

I've made many mistakes before understanding how to report effectively.

At the end of this newsletter, you will be able to craft more effective reports!

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.

The Goals Of Reporting

Let's focus on the other side, our audience. What would they like to hear or learn?

Assess the current situation
Explain why there are problems/opportunities
Explain what can be done

Simple, right?

In practice, it requires some skills in finding the right insights and understandably summarizing them.

Remember, non-technical people have a hard time trusting you... so make sure to convince them!

Looker Studio

How many of you use Looker Studio for reporting?

I bet too many.

The tool is OK for most of SEO work but you need to be careful about:

Telling too much
Highlighting the wrong things
Bad filters

It's common to download your favorite dashboard online and ask yourself "OK, how do I use it?"

Well, start by asking yourself what's the minimum viable product you can show!

This also involves correctly filtering your data... to remove branded terms and random pages (or foreign queries).

Of course, you don't have to make it THAT simple!!! Change the labels here and there and be sure to use the right colors.

The example above is quite minimalistic but it's much easier to consume compared to the average dashboard with 5 pages.

Alternatives

Looker Studio isn't the only tool out there and I enjoy using other tools.

Why you may ask? Because LS is limited and doesn't grant you the same freedom as coding.

For quick work, it's more convenient to use it because you get some nice features out of the box.

In all the other cases, you can have a look at:

Shiny (R/Python)
Dash (Python)
Streamlit (Python)
Tableau/PowerBI (PBI)
Just R/Python/Javascript

Different companies have different stacks but you should at least be confident with PBI and Looker Studio.

Streamlit is nice for prototypes but I'd avoid it for finalized reports.

Dash and Shiny are great allies for building cool dashboards and web apps, so for more complex requests.

In the vast majority of cases, just stick to Looker Studio!

The example above illustrates a simple Shiny App that takes a CSV from the GSC API as input and gives me some basic summaries.

It didn't take much to build it and I got immense help from ChatGPT.

What I Recommend To Report

There are many answers depending on which type of website you are analyzing but at least:

Organic Performance
Financial Metrics

You may also measure Engagement metrics but you must be quite good at it.

The average time on page alone doesn't tell much!

For bigger websites, you want to show more and also give a deep look at internal linking.

Try to avoid the basic tables you get in Looker Studio, like this one:

This is quite hard to consume and the average stakeholder couldn't care less about this level of detail.

Aggregate by page or query instead and match it to some specific action.

More Examples

My specialty is content websites and this is what I also keep track of:

👉 Content Decay

👉 Evergreen Content

👉 % of clicks brought by your top 1-10% pages by clicks

👉 % of 0-clicks pages

👉 How long it takes to publish

👉 How long it takes to update content

As you notice, this isn't what you often read online...

Because the average Web Analytics article spends 99% of its content on GA4 and GTM.

Nothing wrong but as SEOs we are domain experts.

This means we know how websites can rank and what's important to know.

P.S. You can see how I score these metrics in the BigQuery Handbook; Content Decay is in my ebook; the last 2 don't need many explanations.

The Most Effective Plots

If you feel conflicted about what to show... worry not, the easier the better!

After many trials, I can tell you that these work:

Histograms
Barplots
Bubble plots
Area charts
Formatted tables (heatmaps)
Big numbers

You can easily tell a LOT with the above plots without sweating too much.

According to the book "Storytelling with Data", these are among the best choices:

As you can see, the recommended plots are simple to understand! Simplicity over Complexity!

I'd say there are other options too but you can't go wrong with them.

AVOID PIE & DONUT CHARTS!!!

I see them everywhere, especially in corporations.

Avoid them like the plague because they are misleading (and quite ugly).

Even worse are 3D plots because they tend to confuse your audience.

The heatmap below is one of my best examples because it's effective:

This is nothing but a table with conditional formatting on it.

The example below is a fancy line plot showcasing the performance of 3 clusters.

This isn't possible with Looker Studio or it would require a lot of tampering.

Once you have the code, you can save it and recycle it for new projects.

This is much better than doing everything via an interface, which is extremely error-prone.

Presenting To People

So far my best advice is to explain the main insights shortly and tell them right away what to do.

No one will check all of your work and they just want a quick summary.

As always, try these 3 techniques marked inside the red circle:

The names are quite fitting because:

Big Idea means having 1-2 sentences with the main insights.
A 3-minute story is a timespan of 3 minutes where you have to summarize your insights
Actionable lists... simple bullet lists like this one with the takeaways

The first 2 ideas are in the storytelling book I just linked, the last one is my personal take.

N.B. We'll see more about these concepts in the next issues!