(For a quick summary of Search engine optimization, check out the Wikipedia entry)
SEO is a black art – with an emphasis on Art not Science
If you’ve ever wanted to increase website traffic, you’ve likely stumbled on a dozen or more guides on SEO. In fact, if you google "SEO tips" and look specifically for PDFs, you end up with this results page. Click through a couple and check them out.
In general, you’ll find that most of the SEO articles seem to consist of a long list of rules for what’s "good SEO" like:
- Put the right keywords into your URL
- Change your META keyword list to include important phrases
- H1 is so much better than H2
- Use sitemaps, because that helps your indexing
- … blah blah
OK, I’m sure that most of these work, at least to some degree. And at the very least, large groups of consultants and agencies benefit from this confusion, because it ends up looking like a tax code – if you don’t know the ins and outs of hundreds of possible rules, then it’s hard to think about tackling this in-house, or even evaluating the performance.
Like viral marketing, the vast majority of the knowledge out there consists of little lists of "rules of thumb" ideas to implement. And also like viral marketing, there’s a lot of tribal knowledge and hucksters who make a living writing about cute ideas, without putting it all together into a coherent theory.
If SEO were a science…
Let’s think about how the medical sciences has evolved over time:
- 15,000BC to 1700s AD:
- Hmm, if you’re sick and you eat god-knows-what, you seem to get better!
- Also, you might decide to make up some rules for WHY it is that something happened
- This could be, of course confusing correlation with causation
- 1746 AD:
- British guy does the first ever A/B test to figure out that citrus cures scurvy
- Of course, you still don’t know why – but it works
- 1853 AD:
- Aspirin is finally extracted from its herbal form, which is tree bark
- Yay for drugs ;-)
And of course, it was only recently that we started to understand HOW and WHY certain drugs work, through a deeper understanding of the human body and the chemistry involved.
So back to SEO, if the accumulated knowledge essentially consists as a bunch of "best practices," then what you have is a very nonscientific approach to the problem. It’s basically at the stage of eating various things, seeing if they kill you, and collecting a list of plants that may or may not help you. And as a result, you’d have a list of "Eat this, and not this"-type rules, rather than a unified way of approaching the problem.
What are the right SEO metrics?
First off, if anyone has found some interesting data or articles in this direction – please send them to me – I’d love to read more about the topic.
In the meantime, I’ll talk about what I’ve found so far: After searching around, I definitely found some interesting articles for how to incorporate these ideas into a broader framework – here are two:
Both are worth a quick read.
The idea is to separate out different parts of SEO. You have 3 components:
- Methods (do X not Y, do A not B)
- Performance dashboard (our # of pages crawled is X, let’s get it to Y)
- A/B testing
So generically, you basically have the actual experiment you are running – like whether or not a change you make to the site helps – and then an experimental framework that helps validate what is happening.
As a result, you end up breaking down a performance dashboard that has variables like:
- Number of pages crawled
- Number of pages indexed (3 Main Engines)
- Number of days X % of the pages survive in Google from first full crawl to deindex (What I call ‘Burn Rate’)
- Indexing methods used
- Uniques
- Page views
- Clicks
- Revenue
(quoted from SEOIdiot)
Or, metrics like:
- Brand-to-nonbrand ratio
- Unique pages
- Page yield
- Keyword yield
- Visitors per keyword
- Index-to-crawl ratio
- Engine yield
(quoted from Stephan Spencer)
But having all of this stuff together is pretty confusing, isn’t it? Especially when you can’t figure out what question you are really trying to answer?
Putting these metrics into a dashboard
Tracking a large collection of metrics is not interesting in itself. What you are essentially looking for is a series of steps – I’ll frame this a community site that generates SEO pages, since I’m sure that’s what a lot of people are working on:
- User comes to the site
- User writes an article
- Article is then published to a unique URL
- URL is then crawled
- URL is then indexed
- Keywords are then searched
- URL then shows up in the search result page
- URL is then clicked on
- Then a user ends up back on the site
What you basically want to do is to step through and track variables against each step of this process – the more granularity you can get, the better.
In fact, let me restate the really interesting question you’re trying to answer from these stats:
For each page created by a user, how many additional
users its likely to attract over its lifetime?
Integrating this into a
holistic dashboard, you’re then able to see how the changes you make
impact the variables throughout the site.
However, that’s not what most people do – they might track only steps #1,#9, through aggregate pageview numbers. Or if they are going to track steps #1,2,3,9, since those are the only things happening directly on their site.
But of course, in the case where you can’t measure directly, you can measure indirectly via using comScore or Compete data, or doing external searches yourself, or whatever. Or, you start using measures like in-bound links, or some other metrics, which lets you at least roughly estimate how a change you make alters the numbers for what Google will give you.
If someone has developed or used a dashboard like this, I’d love to talk about it ;-)
This still leaves A/B testing, which I’ll leave to a thought experiment for the reader, or a future blog.
UPDATE: My good friend and old college roommate, Eric Peters, gives his take here. Eric is a metrics guru and was previously at MSN working on monetization, so his opinion is absolutely worth reading.