Some interesting considerations on data for #SEO and how it's very easy to lie with them.
And ok, some curiosities and strange facts as well, keep reading. A thread to open your eyes π§΅
And ok, some curiosities and strange facts as well, keep reading. A thread to open your eyes π§΅
Let's start with the very basics. For big brands, you should always filter out branded keywords in Google Search Console, no option here.
You want to attract those people who aren't directly searching for you.
You want to attract those people who aren't directly searching for you.
Filter by query and select Custom (regex), then just select "doesn't match" and insert all the branded terms. To select more use the pipe operator |, it means OR. 
Another common error in GSC is about the position. It is calculated as the average of the topmost positions.
If for query X your property appears at positions 1, 3, and 7, the value will be 1. For query Y, you may get position 5. The avg. position is 3 then >>
If for query X your property appears at positions 1, 3, and 7, the value will be 1. For query Y, you may get position 5. The avg. position is 3 then >>
Sometimes you may indeed notice a very high CTR with a low position, check the SERP and you will find why.
As explained by Google being position 11 is not necessarily bad!
As explained by Google being position 11 is not necessarily bad!
Naturally, positions and data may differ from desktop to mobile. Sometimes I see pages ranking much lower on mobile or vice versa.
The Coverage section is quite important and now we also got the new Inspection API. It's one of the most important sections of the tool and necessary if you are selling tons of products.
Google Analytics is a tricky tool as well, plenty of strange metrics for beginners, and some of them are similar.
My suggestion is to avoid focusing on vanity metrics like page views or sessions unless you have a certain business model (so this doesn't apply).
My suggestion is to avoid focusing on vanity metrics like page views or sessions unless you have a certain business model (so this doesn't apply).
Sampling is another factor to take into account. Check if you have a yellow or green badge. The former means your data is sampled!
For reporting avoid it and use Data Studio, you cannot just export CSVs and pretend you have done your job.
I prefer to use GA for tracking behavior or how people move along our website.
I prefer to use GA for tracking behavior or how people move along our website.
Data Studio teaches you how metrics work as well, much clearer and you have a lot of room to practice.
GA4 is the new challenge in the industry right now and I have to understand if it is worth learning it or just changing software.
GA4 is the new challenge in the industry right now and I have to understand if it is worth learning it or just changing software.
I mean, I will have to learn it for cheaper projects and because it will be the new Google standard.
I am pretty sure that GA4 is packed with a lot of misleading metrics that people will certainly use wrong.
I am pretty sure that GA4 is packed with a lot of misleading metrics that people will certainly use wrong.
Speaking of third-party tools I do not rely on any of their metrics in general. Share of voice and visibility are nice concepts, yet I think that other metrics or estimates can give you much more value.
Adapt your reporting according to business needs.
Adapt your reporting according to business needs.
Some knowledge of #DataScience or Data Engineering is good to understand data quality. I wouldn't say SEOs work with good data, as we often deal with very inaccurate data.
A common pitfall and danger of using tools are falling into the search volume trap. I addressed this point yesterday when talking about Keyword Research:
https://twitter.com/GiordMarco96/status/1488149539076853773
Another common mistake is to think that SEO audits are based on meaningless metrics like DA/DR or any other acronym. You cannot judge a website with an integer number, it's pointless.
Correlation and causation are super common topics nowadays. It's pointless to draw conclusions based on circumstantial evidence. Plus, some tactics work in some industries, they are not generalizable.
"Garbage in, garbage out" If you have bad data your model is going to suck too. That is why you should understand some of your data at least. Be extremely careful when doing analyses.
So, you have understood that third-party data are not that good, still, they are useful. Sometimes you have to think directionally and not focus on individual numbers.
This is (partially) the philosophy of topical maps.
This is (partially) the philosophy of topical maps.
For keyword research, you don't care about the exact volume, as you may be rather interested in its range. How popular is this topic compared to another one? >>
>> a topic may seem less popular but when you dig deeper there are infinite subtopics. This happens to me all the time and you can use competitor analysis to see what your rivals are doing.
Speaking of volume, it's better to rank one page for any queries than doing a 1:1 mapping. An important page would never rank for just one keyword alone.
#Python3, R, and SQL are powerful allies but they don't replace your knowledge. For sure, you get to know data better if you know some coding.
Another thread about Data Science and SEO:
https://twitter.com/GiordMarco96/status/1503098590725476355
If you want total control over what you get, learn some scripting and build your own tools. It's very helpful for basic usage, as for advanced use cases I'd always recommend purchasing something already tested.
Another problem is with models. There are SEOs convinced that you can optimize for BERT/MUM.
That is false. You can improve your syntax and include proper entities, ok that's fine.
There are no tricks, you can improve the quality of your text though.
That is false. You can improve your syntax and include proper entities, ok that's fine.
There are no tricks, you can improve the quality of your text though.
A well-written text with proper entities and a clear syntax is easier to understand for a machine.
If you follow basic copywriting advice and you are informative, then you have no problems.
Don't fall for the LSI keywords scam, it's just people without studies in CS.
If you follow basic copywriting advice and you are informative, then you have no problems.
Don't fall for the LSI keywords scam, it's just people without studies in CS.
LSI keywords or any other technical term from different subjects must not be used as a synonym of entities or whatever.
It's like teaching a statistician the definition of "mean" because in the SEO world it's used incorrectly.
Be careful of terminology then.
It's like teaching a statistician the definition of "mean" because in the SEO world it's used incorrectly.
Be careful of terminology then.
While Machine Learning, Deep Learning and Data Science are all different subjects, we can argue that they're playing an essential role in today's society.
The same goes for SEO and we have to ensure that misinformation is not spread across social media.
The same goes for SEO and we have to ensure that misinformation is not spread across social media.
https://twitter.com/GiordMarco96/status/1504178194894635009
The future will be more and more data-centric so that's why learning the basics of some other subjects is essential.
An SEO doesn't need to be a web dev and the same is true for data. You can just learn the basic terms correctly and focus on the consequences on SEO.
An SEO doesn't need to be a web dev and the same is true for data. You can just learn the basic terms correctly and focus on the consequences on SEO.
I am not a big fan of putting everything into the broad domain of Technical SEO.
UX and Data Science may be considered part of it but they are different subjects. Let's treat them correctly for once!
This was also true years ago when you had the rise of "Excel for SEO" articles
UX and Data Science may be considered part of it but they are different subjects. Let's treat them correctly for once!
This was also true years ago when you had the rise of "Excel for SEO" articles
There are Python/R libraries specifically built for SEO. In this case, we are restricting the focus on our domain,
However, 99% of the code we produce is not exclusive to SEO, then why call it Technical SEO?
However, 99% of the code we produce is not exclusive to SEO, then why call it Technical SEO?
Technical SEO is a un umbrella and I love it. SEOs not proficient with data may think that data should necessarily fit into this bracket.
Imho there is no need for such an exemplification. It's just borrowing from another subject (actually many).
Imho there is no need for such an exemplification. It's just borrowing from another subject (actually many).
Another common problem with data is given by those pesky case studies.
It's very hard to prove causation and gathering data from 100K websites is not a robust statistical methodology to do that.
I've seen some good case studies in the past but I don't even recall where!
It's very hard to prove causation and gathering data from 100K websites is not a robust statistical methodology to do that.
I've seen some good case studies in the past but I don't even recall where!
Correlation is a useful indicator and can raise the suspicion that something is fishy. This doesn't mean that noticing some correlation in your data is automatically enough to justify publishing a case study.
Another factor involves the markets/niches and the sampling.
Another factor involves the markets/niches and the sampling.
Most case studies are biased towards the American/English market.
They are the most competitive out there and there is objectively more competition on average.
Some SEOs just don't work in those markets or maybe an International SEO wants to know more about something else.
They are the most competitive out there and there is objectively more competition on average.
Some SEOs just don't work in those markets or maybe an International SEO wants to know more about something else.
All these considerations about data are true for other industries as well. The only difference is that SEO is not regulated and we cannot really prove certain theories.
When data are not enough, we have to rely on our innate tools. Hunch your best friend for some decisions.
When data are not enough, we have to rely on our innate tools. Hunch your best friend for some decisions.
Having experience in a niche is a huge advantage, especially if you don't have data.
Don't work with bad data, use your intelligence. If you feel like your data are nonsense, ditch them.
This statement assumes you have domain knowledge on a topic.
Don't work with bad data, use your intelligence. If you feel like your data are nonsense, ditch them.
This statement assumes you have domain knowledge on a topic.
Don't rely too much on your instinct or you can get destroyed by the harsh reality.
You have to be cold and factual when you have the opportunity to! Use your instinct for those rare scenarios where you have no time or data to make a decision.
You have to be cold and factual when you have the opportunity to! Use your instinct for those rare scenarios where you have no time or data to make a decision.
β’ β’ β’
Missing some Tweet in this thread? You can try to
force a refresh
γ
