Ryan Jones Algorithm Chaos Pubcon 2013

26
Algorithm Chaos @RyanJones Manager, SEO & Analytics – SapientNitro www.RyanMJones.com

description

Slides from my talk on algorithm chaos at Pubcon Las Vegas 2013

Transcript of Ryan Jones Algorithm Chaos Pubcon 2013

Page 1: Ryan Jones Algorithm Chaos Pubcon 2013

Algorithm Chaos

@RyanJonesManager, SEO & Analytics – SapientNitrowww.RyanMJones.com

Page 2: Ryan Jones Algorithm Chaos Pubcon 2013

2

“We make over 500 changes to our algorithms

[each] year”

- Matt Cutts

How can we keep up with the ever-changing SEO algorithms?

@RyanJones

PandaPenguin

EMDHummingBird

Page 3: Ryan Jones Algorithm Chaos Pubcon 2013

3

“We cannot solve our problems with the same

thinking we used when we created them”

- Albert Einstein

Think Different. (about the algorithms)

@RyanJones

Page 4: Ryan Jones Algorithm Chaos Pubcon 2013

4

“al·go·rithm

ˈalgəˌriT1Həm/

noun

noun: algorithm; plural noun: algorithms

1.

a process or set of rules to be followed in calculations or other

problem-solving operations, esp. by a computer.

The Algorithm Has Evolved

@RyanJones

Everybody’s 2nd Favorite Formulas

Page 5: Ryan Jones Algorithm Chaos Pubcon 2013

5

Dynamic Weighting

@RyanJones

Algorithm Factors & weights are dependent upon the result set (corpus)

Page 6: Ryan Jones Algorithm Chaos Pubcon 2013

6

Dynamic Weighting

@RyanJones

When the corpus changes, so does the importance of various factors.

Page 7: Ryan Jones Algorithm Chaos Pubcon 2013

7

My Theory: It actually is brain science

A

@RyanJones

- A “simple” neural network

Something to Think About:Neural networks Require Training Sets• Human Quality Raters• Panda• Penguin• Matt’s latest call for “good sites that should rank

higher”

Page 8: Ryan Jones Algorithm Chaos Pubcon 2013

There’s Two Theories of SEO

@RyanJones

Page 9: Ryan Jones Algorithm Chaos Pubcon 2013

9

Type 1: Storm chasers

The Algorithm Chaser / Reverse Engineer.

@RyanJones

“This is how SEO used to be, many years ago. It

used to work very well.”

Analyze, Measure, Correlate, Examine, Interpret, & Follow top ranking sites.

Try to figure out the factors that top sites have in common and emulate them.

Sometimes The Tail wags the dog.

Page 10: Ryan Jones Algorithm Chaos Pubcon 2013

10

Type 2: Storm Predictors

Pay attention to where the algorithm is heading

@RyanJones“We sometimes call these SEOs Google fan-boys. ”

Page 11: Ryan Jones Algorithm Chaos Pubcon 2013

11

10,000 foot view

Focuses on “What” rather than “How”

@RyanJones

Google wants to rank Quality, Useful sites. Ok, how

can I make my site more useful?

Page 12: Ryan Jones Algorithm Chaos Pubcon 2013

12

Let’s Talk About Hummingbird

@RyanJones Source: soyunhamster.deviantart.com

Hummingbird Changes The Algorithm Focus

OLD: Show most popular results that actually contain “keyword”

NEW: Show the most useful and popular results for “what the user means”

Page 13: Ryan Jones Algorithm Chaos Pubcon 2013

Algorithm MythsLet’s Change How We Think About Search Algorithms

Page 14: Ryan Jones Algorithm Chaos Pubcon 2013

14

Myth 1: The Algorithm is a Static Formula

It’s Not. Get Over It.

Ranking Factors differ by query & corpus

@RyanJones

Page 15: Ryan Jones Algorithm Chaos Pubcon 2013

15

Myth 2: Google Has a Brand Bias

Customers

favor brands

Google favors what customers favor.

@RyanJones

Source: google consumer surveys.Question : “where would you buy a flat screen TV”

Page 16: Ryan Jones Algorithm Chaos Pubcon 2013

16

How do you determine if something is a brand?

Hint: Not this way: (in_array($knownBrands,$URL))? $ranking++ : $ranking--;

Google doesn’t check against a list of brands.

Trust. Authority. Popularity

@RyanJonesNote: yeah, I know Google uses python, not PHP. Here’s python. Happy?ranking += 1 If url in brandarray else ranking -= 1

Page 17: Ryan Jones Algorithm Chaos Pubcon 2013

17

Myth 3: Altering vs. Ordering

Some factors alter the SERP corpus. Others simply order it.

@RyanJones

Alter (included or not?) Order (ranking)

Device Type Pagerank

Location Page Speed

Relevance (tf-idf?)

Some factors may do both. E.G. personalization, QDF, manual penalties.

Page 18: Ryan Jones Algorithm Chaos Pubcon 2013

18

Myth 4: Rankings changes only hurt people.

For every site that loses rankings, another one gains.

(although, to be fair, sometimes that “other” site is Google)

@RyanJones

Page 19: Ryan Jones Algorithm Chaos Pubcon 2013

19

Myth 5: Rankings dropped, I got penalized.

If you just dropped a couple spots, you weren’t penalized.

If you fall out of the search results completely, that’s a penalty.

@RyanJones

Page 20: Ryan Jones Algorithm Chaos Pubcon 2013

20

Myth 6: Social Signals Are Super Important

Social Signals Are Important, But Not For Rankings

Google has said they don’t use Facebook or Twitter signals in rankings, and are still evaluating plusses/likes.

@RyanJones

Page 21: Ryan Jones Algorithm Chaos Pubcon 2013

21

Let’s Talk About Correlation

@RyanJones

“You keep using that word. I do not think it means what you think it means.”- Inigo Montoya

Page 22: Ryan Jones Algorithm Chaos Pubcon 2013

22

Some Fun Correlations

@RyanJones

Sometimes, the “hidden factor” is simply just Time.

Page 23: Ryan Jones Algorithm Chaos Pubcon 2013

23

Example Correlations

@RyanJones

• Ice Cream Sales are highly correlated with crime.

• Number of Facebook Users correlates to the national debt.

• In children under 3, vocabulary and shoe size have a correlation coefficient of 1.0

• One SEO study actually claimed having a keyword optimized title tag is highly correlated with manual webspam penalties

Source: xkcd.com

Hidden Factors:• Heat / Weather• Time• Spammers do keyword research – more than the general web.

Page 24: Ryan Jones Algorithm Chaos Pubcon 2013

24

SEO correlations have hidden factors too!

• Plusses / Facebook likes are correlated to rankings

• But people can’t plus/like what they can’t find. (e.g. rankings came first)

• Let’s better understand correlation coefficients too!

• Anything under .6/.7 is pretty scattered.

Full Circle: Perhaps study values are so low because factors

vary based on the result sets?

@RyanJones

Page 25: Ryan Jones Algorithm Chaos Pubcon 2013

25

Tips For Surviving Algorithm Changes

• Don’t chase algorithms, chase the reasons for the algorithms.

• Understand why the algorithm changes were made

• Focus less on ranking factors and more on why they are factors.

• Be a Brand. Do what brands do. Build Quality, Trust, Authority. Goal: If your site isn’t ranking, customers should think that Google is broken.

• Don’t Ask: “how do I rank for ___” Ask: “What do people searching ___ want to see?” @RyanJones

Page 26: Ryan Jones Algorithm Chaos Pubcon 2013

© 2013 SAPIENT CORPORATION | CONFIDENTIAL

@RyanJonesManager, SEO & Analyticswww.RyanMJones.com

SapientNitro

Thank You