Traditional ranking factors have been great proxy metrics for the past years and have made Google the best search engine in the world. But as Google advances to an AI first company, SEOs have to change their metrics and toolsets as well. Learn about how Machine Learning changes SEO and why User Signals and UX have to guide your work.
4. Search Engine Optimization
Artificial Intelligence
#SEO
#AI
1. Status | 2. Functionality | 3. Implications | 4. How to get started
Slide Deck: http://twitter.com/pip_net
16. The Future of Search Engine Optimisation
SEO
C P T User Experience
17. Caffeine
Update
June 2010
February 2011
Google Panda
Update
Hummingbird
August 2013
March 2015
RankBrain
Amit Singhal
steps down
February 2016
May 2017
“AI first”
Company
Timeline: How AI entered Google Search
Artificial Intelligence gradually changes how search works
23. This is not just some update. This is
Google finally getting much better at
„Determining Quality“.
24. In other words… Awesomeness.
If you find better ways of
measuring „Awesomeness“.
Why stick to Ranking
Factors?
25. Training Machine Learning on User Signals
Refined
Search
CTR, Short &
Long Clicks
Task
Completion
(Intent)
Quality
Raters (EAT,
Guidelines)
Browsing &
Supplemental
Data
More Background: https://www.cnbc.com/2018/09/17/google-tests-changes-to-its-search-algorithm-how-search-works.html
26. Search is Simple According to Gary Illyes
„Dwell time, CTR, whatever Fishkin‘s new theory is,
those are generally made up crap. Search is much
more simple than people think.“
Gary Illyes, Head of Fake News at Google
„…waiting for a plane and I‘m bored“
27. Training Machine Learning on User Signals
More Background: https://www.cnbc.com/2018/09/17/google-tests-changes-to-its-search-algorithm-how-search-works.html
"On the most basic level, Google could see how satisfied users were.
To paraphrase Tolstoy, happy users were all the same. The best sign of
their happiness was the "Long Click" — This occurred when someone
went to a search result, ideally the top one, and did not return. That
meant Google has successfully fulfilled the query.“
~ Steven Levy. In the Plex: How Google Thinks, Works, and Shapes our Lives (April 2011)
28. Google Patent: Quality Scores for Queries
A long click occurs when a user, after clicking on a search result, dwells on the landing page (i.e.,
the document to which the search result links) of the search result or clicks on additional links
that are present on the landing page. A long click can be interpreted as a signal that the query
identified information that the user deemed to be interesting, as the user either spent a certain
amount of time on the landing page or found additional items of interest on the landing page.
A click-through reversion (also known as a “short click”) occurs when a user, after clicking on a
search result and being provided the referenced document, quickly returns to the search results
page from the referenced document. A click-through reversion can be interpreted as a signal
that the query did not identify information that the user deemed to be interesting, as the user
quickly returned to the search results page.
Source: http://www.seobythesea.com/2018/07/quality-scores-for-queries/
32. Hidden Content Penalty
or User Metrics?
• Content has always been in
DOM
• CSS Layer
• “Black on black” style
• Manual interference?
• Automatic Hidden Text
trigger?
• Or fast UX safeguards?
• Would Google penalize
anti-Art. 13 activism?
42. The Future of Search Engine Optimisation
SEO
C P T User Experience
43. The Future of Search Engine Optimisation: Implications
SEO
C P T User Experience
Hygiene
Factors
Defining the Top Rankings
Ranking Factors:
Good Proxy
Signals for
Website Quality
• Direct Measurement of User Experience and Satisfaction (Long Clicks)
• Extensive Testing and Incremental Learning
• Training AI, based on Immediate User Feedback
• Data Aggregation, „Reverse Aging“, Even Stronger Network Effects and Moats
Relevance
(You still need this!)
„Understanding Quality for the User“
44. Implications of an AI-centric Search Environment
User Focus
More than ever: Relentless
Focus on the User
• No compromise on user experience
• Understanding Intent, Needs and
Objections
• No more “SEO text”, useful content
or no text at all
• Strong Focus on Mobile and Speed
• Paywalls/Signup Layer might incur
high indirect costs
• Embrace Voice and Conversational
Search
• “Panda Diet” – Trimming the weak
spots of your website
• Continuous UX/CRO Improvements
instead of SEO Tricks & Arbitrage
KPIs & Metrics
How to Measure a Great
Product and User Experience?
• Conversion? (sparse data)
• Bounce Rate? (unreliable)
• Session Time?
• Pages/Session?
• Retention?
• Time Spent per Month
Organisation
Challenging the classical “SEO
Team” Approach
• SEO has interfaces with
Dev, Product, PR, Design,
Content/Editorial, etc
• Ideal World: SEO would be
a frontend PM skill
• Install tech. SEO as a PM
• Create independent x-
functional SEO dev team
• Integrate Data Scientists &
Analysts
Tools & Insights
From Ranking-Tracking to Pattern
Recognition
• „Why have I lost rankings?“ -
„Who won and why?“
• Typical SEO Analysis: Data
Aggregation & Pattern
Recognition
• New Tools should identify
patterns and less obvious
correlations automatically
• AI can assist SEO (Indexation,
Thin Content, Classification &
Categorisation, Estimation,
etc)
46. The Only SEO Metric I Measure Daily…
What matters most
• Conversion (indexed)
• Impressions/Session
• Time On Site
• Bounce Rate
• Returning Visitors
• Ultimately: Monthly Usage Time
47. Is Dwell Time the
Perfect UX Metric?
1. Indicative for success
2. Measurable
3. Manageable
4. Universally available
5. Gets even better if multiplied with no.
of sessions (Time Commitment = )
What about...?
• Conversion Rate
• Bounce Rate
• Task Completion
• Happyness/NPS
48. A Few Perspectives on Dwell Time
1. Average Dwell Time is heavily
influenced by short clicks
2. Loyal users and brand queries
always create better signals
(unfair advantage)
3. Good Dwell Time Segment is:
„New Organic User w/o
Homepage“
4. Dwell Time IS NOT a ranking
factor, but includes lots of quality
signals
5. Dwell Time is still a proxy metric
for measurement of quality, not
an input factor
49. Eric Schmidt on Brands‘ Unfair Advantage
Brands‘ function is to reduce cost of research:
Same information sold by brand = more value.
50. Is Dwell Time a Great Second Best Metric?
"A simple statement is bound to be untrue.
One that is not simple cannot be utilized."
~ Paul Valery (1937)
(later “Bonini’s Paradox”)
"All models are wrong,
but some are useful."
~ George Box
51. The Outlook
• By collecting user data Google continuously gets
better at Popularity, Quality & User Experience
• Classic SEO can’t save a shitty product anymore
• More Affiliate and MFA pages will vanish for good
• SEO Darlings will have to surrender market share
to brands playing catch-up
56. Four Infamous User Experience Sins
Navi vs. Search
• Internal Search is existential for
big websites
• Users don‘t „navigate“
• …apart from „back to Google“
• Competition is one click away
• Internal Search heavily impacts
page per session and AOV
Pagination
• Never made sense in the first
place
• Unnecessary burden for the user
• Even worse when site is slow
• Modern learned navigation
patterns are swiping and scrolling
• Showing more items per page is
already good
„SEO Texte“
• Don’t call ‘em like that or they
will look like that
• Usually not essential for a
better shopping experience
• Unless you can help with
research (ready to buy)
• Promoting USP instead much
better use of space
Brand & Trust
• EAT for YMYL
• Borrow Authority
• Quality Rater Guidelines
• https://dejanseo.com.au/trust/
58. What if I slow down my site to increase Time on Site?
What if I slow down the ICE train to make you
spend more time on trains?
1. Faster websites have a HIGHER Time on Site
2. They also are visited more often
3. You can’t torture people to stay longer
(as long as the BACK button exists)
59. What if my website is serving short answers really fast?
Unless you‘re Google, you don‘t have a business.
Good luck.
1. Quick answers will not generate any website visits
in the future
2. Still, good websites serving information fast have
a higher dwell time
3. Google‘s Time on Site is 10 minutes by the way…
61. Theory: Typical Page Quality (Qp) over Number of Pages (np)
np
Qp
Homepage
Category
Category+Brand
Facetted Search
Thin Catalogue (low inventory)
Dupe Content page „no results“ page
highestlowestmediorceuseful
400.000200.000 300.000100.000
Page Quality (Qp) can
be defined as content
richness, engagement,
ultimateley how useful
the page is to the user.
But also its revenue
potential.
PROBLEM:
Since Panda (2011) this
structure has become toxic.
63. Theory: Typical Page Quality (Qp) over Number of Pages (np)
np
Qp
highestlowestmediorceuseful
400.000200.000 300.000100.000
Average Quality
😞
Quality Threshold 80% (mediocre and better)
NOINDEX
(320.000)
INDEX
(80.000)
New Average Quality
QTYINCREASE
80%
Quality Threshold
RANKINGS
Page Quality (Qp) can
be defined as content
richness, engagement,
ultimateley how useful
the page is to the user.
But also its revenue
potential.