Thread by @victorpan

Happy Friday - let's rant about keyword density, LSI keywords, internal/external links per N words, and word count - vernacular you want to get out of your processes sooner rather than later.

1/n

Keyword density, usually defined as a percentage of amount of times a keyword shows up divided by the total number of keywords on the page, is product of the keyword-stuffing years of SEO.

e.g. I heard SEO Expert Chad recommends pages should have a Keyword Density of 2% !

2/n

Search engines were still using the keywords meta tags for rankings, the amount of times a keyword was repeated on a page had a huge influence on whether a page would rank, and there were algorithms that would pick up keyword stuffing. Yo it's not the pre 2k or early 2000's.

3/n

The only reason why you'd have a "keyword density" check on your content in 2020 would be to make sure you're writing on-topic on a particular topic.

There's no perfect %. If someone tells X% keyword density is a requirement, they know just enough SEO to be dangerous.

4/n

What could be better? Queries are questions and SERP results are answers. Identify the underlying problem(s) a query signifies and assess pages by breadth and depth you should cover. People Also Asked is a good place to start.

Content comprehensiveness, not density.

5/n

Let's get to LSI, or latent-semantic indexing. Somehow the SEO industry appropriated latent semantic analysis to keywords in SEO via some weird branding. Actually HubSpot has done this with topic clusters (a data science term) with content strategy.

It's weird.

6/n

But the core premise of "LSI keywords" is adding synonyms and co-occuring terms to the topic you're writing about will improve your rankings. It's somehow the magic panacea to your ranking problems.

It's not. Latent-semantic analysis isn't used. Some still think it is. Why?

7/n

It goes back to confirmation bias and digging a deeper hole that public-facing Search Engineers from Google/Bing are lying to us. Those folks are doing their jobs and answering to the best of their ability on a huge codebase worked on by a bunch of engineers. Ya think?

8/n

So what can you replace "LSI keywords" with? Going back to queries are problems and search results are answers - think about the problem in different contexts - will weather, time of day, device type, age, nationality, etc. change what's a good answer?

9/n

Sounds fluffy and hard to grasp? One of my content crush sites (The Wirecutter) does a great job systematically creating three categories (and more) when they review the "best" products in their category. Budget, best for most people, upgrade picks.

10/n

Based on different needs of the searcher, the content talks about different things. There's no need to checkbox keywords from a list of keywords from your SEO tool. Those SEO tools are an aid, not a crutch to hold on. Empathy is the key. LSI keywords? 🤮 (views my own)

11/n

And now we get to "linking to things is important for SEO, so how often should I link?" I've been guilty of saying "one link per 300 words" as a comrpomise between UX and being visually spammy.

The problem with that answer is it's very keyword-density-like. I was wrong.

12/n

A white lie, ok? But that white lie perpetuates that SEO is some black internet magic (tbh it feels that way when you rank #1 on something). Folks start to assume you can do your "SEO juju" on their project ad-hoc.

Instead, have an internal editorial policy on linking out.

13/n

How about internally linking? Again, think about the users and the value they'd get - problems in life are a journey. If you've got a content strategy that addresses the stages of the searcher's problem, it's natural to link to things before and after.

14/n

So an example could be...

X
How to see if you have X
How to fix X
How to upgrade X
How to modify X
How to do X from scratch
X Reviews
The Best X in 20XX
X vs Y

etc. - you'd have placeholder internal links and directions to add them when your editorial calendar allows.

15/n

Last but not least, we have the beast that is word count. Full disclaimer - HubSpot used to have a strategy that was to create 5 blog posts per day. It worked very well until it didn't. Then there was a phase of "Ultimate Guides" and then we got "skyscrapered" like heck.

16/n

There's two confounding variables (possibly more) that happen when you write long-form content:

1. In some industries people are drawn to sharing long-form content. It's the digital equivalent of "I've read this timeless book so I'm brilliant." Links influence rankings

17/n

2. You force writers to come up with unique perspectives when more words are needed to be met. That was me in school in literature class. You capture more of the long tail

18/n

“I didn't have time to write a short letter, so I wrote a long one instead.”

You don't need a word count requirement for SEO. Frankly you just need writers to identify X potential problems the searcher is running into and Y variations that impact different results.

19/n

By giving writers the tools to empathize with searcher and research the complexities of their needs (identify 3 problems, identify 3 environmental factors for each problem) you've suddenly a much more actionable brief than keyword density, LSI keywords, and word count.

/end

Hope ya'll have a great weekend. For my older rants, enjoy the good ol' internet rabbit-hole:

https://twitter.com/victorpan/status/1304518059353427970?s=20