Darren Shaw - User Behavior and Local Search - Dallas State of Search 2014

User Behavior and Local Search
Darren Shaw
State of Search 2014
@DarrenShaw_ +DarrenShaw

Do clicks affect rankings?

Let’s test

Everyone, get out your phone,
disconnect from wifi,
and Google: dallas wedding dresses

Scroll down and click on
mockingbirdbridal.com in the
organic results.

Now, let’s amplify this via
social media
@DarrenShaw_ AND +DarrenShaw_

How I got obsessed with user
behaviors as ranking signals

bit.ly/11mkCru

bit.ly/1ureMjr

bit.ly/1kxP9ar

slidesha.re/WdbkLK

Pull out your phones, search for “seattle
wedding dresses”, and click on Belltown Bride

Within 10 minutes they had moved up
from the bottom of the page to #2

They also moved into position B in the local pack,
and they still hold this spot today.

This demands further testing!

Test 1 – Hammer Time
Low competition market
Click website link 100 times per day
Repeat for 30 days

Starting position on
October 4th
E in the local pack
Search for “roanoke accountants”
Click on “Stanley G Boatwright”

Checked on November 15th
No change…

Checked from my
hotel room last
night. Again, no
change.

Test 2 – A Softer Approach
Repeat for 7 days

October 6th
G in the local pack
Search for “wilmington nc accountants ”
Click on “Joseph Guo CPA,PLLC ”

Checked on
November 15th
and look at this!

Checked from my
hotel room last
night…

Test 3 – Click Burst
High competition market
repeat for 3 days

Search for “los angeles personal injury lawyers”
Click on “The Torkzadeh Law Firm”
October 3rd
C in the local pack

Watch what happens
over the 3 days

Why did test 3 seem to work, but
the first test didn’t?

REMINDER
Test 1 – Hammer Time
Repeat for 30 days

Theory
The first test was way overboard. An
unnatural amount of clicks for that
city and keyword.

Test 4 – Slow and Steady
Repeat for 30 days

Search for “los angeles moving companies”
Click on “West LA Movers”
October 3rd
G in the local pack

Watch what happens
over the thirty days

I checked from my
room in Dallas last
night, and still
at position F

Hmm. Inconclusive.

Test 5 – Clicks on Google+ Link
Click “Google+ Page” 50 times per day
Repeat for 3 days

Search for “miami personal injury lawyer”
Click on the Google+ link for “The Altman Law Firm”
October 3rd
G in the local pack

Click this link, not the business name.

Over the course of the
3 days, not much changed.
This is the last screenshot
taken by our system

But I checked in my
browser from Denver
yesterday, and looky
here!

Theory
Like with the test on Linda’s forum,
perhaps it just needed some time to
“simmer” before rankings were
affected.

Burning Question
Can you do this with Mechanical
Turk?

Search for “plumbers san antonio tx”
Click on “Bryco Plumbing Co”

November 4th
G in the local pack

November 7th
No change

November 13th
Moved up one
position…

November 16th
They’ve now moved
up 2 spots

Search for “plumbers philadelphia”
Click on ORGANIC result “Sam Wexler Plumbing”

Started here on
November 4th

November 16th
No change
I’ve been checking
regularly, but not
seeing any change

Using your mobile phone and cellular network,
Search for “accountants cincinnati”
Click on “Mellott & Mellott”

I asked them to complete the task on our website
so I could track their user agent and make sure
they were actually using mobile.

Started here on
November 4th
Position E

A little early, but
possibly trending
towards higher
rankings?

Checking from different
IPs and browsers. Appears
to also be trending towards
improved rankings.

Using your mobile phone and cellular network,
Search for “electricians cincinnati ohio”
Click on “Queen City Electric”

So far, no ranking changes
with the click-to-call test, but it has
only been running for 6 days.

Pogo-sticking Test
Can we demote a listing through
automated pogo-sticking signals?

This locksmith spammer has
created a fake listing with
the angieslist category page.
I wonder if we can take
him out with microworker
pogo-sticking…

Search for “cincinnati locksmiths”, click on that
spammy listing, click back, then click on the top
ranked listing.

So far, 166 people have done it, and
no ranking changes, but I sure as
hell hope this doesn’t work.

Can you imagine?
Pogo-sticking Negative SEO…

Why no driving directions test?

Google has plenty of data to determine if you
actually drove to the location. Just searching for
directions isn’t a signal they need to use.

Ranking Fluctuations
What is going on?

October 4th
Position E
Remember the first test where we
didn’t see any ranking change?
November 16th
Position E

We captured over 1700 screenshots
from October 3rd through November
15th. There was a ton of variation.

Oct 24th

Oh, and look here
on Nov 12th

What’s causing the variation?
Different browsers?
Different data centers?

Notice that there wasn’t nearly as
much variation in the los angeles
personal injury lawyer results.

Theory
Google deliberately mixes up the
results to test if users like them

Google has much more data on “los
angeles personal injury lawyer” than
“willmington nc accountants” so
they already know the preferred
choices, and there is less variation.

Multi-armed bandits and the epsilon-greedy method
http://bit.ly/11cMyxt

http://bit.ly/1ybvkLE

http://bit.ly/1wMyoMa

Let’s check our test…

Take Aways

1. Google *is* using click-through
data, and sadly, it *is* easy to
manipulate.

2. The ranking fluctuations since the
Pigeon update are deliberate
(I think)

3. Weekly or monthly rank tracking
data is useless. Need to track at
least daily to measure trends.

Thank you!
Any questions?

Darren Shaw - User Behavior and Local Search - Dallas State of Search 2014

Recommended

Recommended

More Related Content

What's hot

What's hot (13)

Viewers also liked

Viewers also liked (20)

Similar to Darren Shaw - User Behavior and Local Search - Dallas State of Search 2014

Similar to Darren Shaw - User Behavior and Local Search - Dallas State of Search 2014 (18)

Recently uploaded

Recently uploaded (20)

Darren Shaw - User Behavior and Local Search - Dallas State of Search 2014

Editor's Notes