Seo state of the union 2015

The SEO State of the Union

Anirvan Chatterjee • Brian Turner • Leslie YuanClinical & Translational Science InstituteUniversity of California, San Francisco

Agenda

1. Why SEO works

2. How we tested SEO for 52 websites

3. The SEO leaderboard

4. Three things we learned from the results

5. How to boost your rankings

1. WHY SEO WORKS

Back in 2010…

Back in 2010…

Back in 2010…

Back in 2010…

Back in 2010…

UCSF Profiles launch promotion

• Email every single member of faculty

• Postcards to faculty• Free iPad contest

UCSF Profiles traffic

October 2010

• 5,185 visits

July 2015

• 97,864 visits

UCSF Profiles visits, 2010-2015

Sep-1

0

Dec

-10

Mar

-11

Jun-1

1

Sep-1

1

Dec

-11

Mar

-12

Jun-1

2

Sep-1

2

Dec

-12

Mar

-13

Jun-1

3

Sep-1

3

Dec

-13

Mar

-14

Jun-1

4

Sep-1

4

Dec

-14

Mar

-15

-

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

Daily visits to UCSF Profiles

UCSF Profiles visits, 2010-2015

Sep-1

0

Dec

-10

Mar

-11

Jun-1

1

Sep-1

1

Dec

-11

Mar

-12

Jun-1

2

Sep-1

2

Dec

-12

Mar

-13

Jun-1

3

Sep-1

3

Dec

-13

Mar

-14

Jun-1

4

Sep-1

4

Dec

-14

Mar

-15

-

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

Daily visits to UCSF Profilessearch vs. everything else

The dream

Latina breast cancer

The dream

The dream

The dream

!

The theory

Photo: Joseph Wu

The reality

Photo: Jim Deane

How search-driven users navigate

People search for names, e.g.:

• kristine yaffe • lawrence fong ucsf • eric vittinghoff • aaron fields ucsf • jack taunton • aimee kao ucsf • joe derisi • abul abbas

How search-driven users navigate

How search-driven users navigate

Homepage vs. search engine

Visits that start off on our home page…

Visits that go from search engine to a profile…

Homepage vs. search engine

Visits that start off on our home page…

• 68% look at 2+ pages per visit

• n = 3,222 in July 2015

• 2,190 visits of 2+ pages in July 2015

Visits that go from search engine to a profile…

• 18% look at 2+ pages in a visit

• n = 92,743 in July 2015

• 16,245 visits of 2+ pages in July 2015

Why does SEO matter?

• Usage• Usage• Usage

UCSF Profiles search traffic, 2010-15

Sep-1

0

Dec

-10

Mar

-11

Jun-1

1

Sep-1

1

Dec

-11

Mar

-12

Jun-1

2

Sep-1

2

Dec

-12

Mar

-13

Jun-1

3

Sep-1

3

Dec

-13

Mar

-14

Jun-1

4

Sep-1

4

Dec

-14

Mar

-15

-

500

1,000

1,500

2,000

2,500

3,000

Visits per day from search engines

We think we’re doing great

And other sites are too

2. HOW WE TESTED SEO FOR 52 WEBSITES

Inclusion criteriaWe picked 52 research networking sites that are…

• Associated with a single institution– exclude trade groups, collaborations, etc.

• Based in a majority English-language locale– exclude France, Germany, etc.

• Accessible to the public and search engines– exclude systems behind a firewall

• Running on a host on port 80– exclude 54.213.177.247 or hostname.edu:8080

InstitutionsAlbert Einstein College of Medicine

Arizona State University

Boston University

Case Western Reserve University

Clinical Translational Science Institute at Children's National

Cornell

Duke University

Georgia Regents University

Harvard University

Indiana University

Johns Hopkins University

Michigan State University

Montana State University

Northern Arizona University

Northwestern University

Ohio State University

Oregon Health & Science University

Penn State

Scripps Research Institute

Stanford University

Temple University

Texas A&M

Thomas Jefferson University

University of Arizona

University of California, Davis

University of California, San Diego

University of California, San Francisco

University of Colorado Boulder

University of Colorado Denver

University of Florida

University of Hawai‘i

University of Illinois - Chicago

University of Iowa

University of Maryland-Baltimore

University of Massachusetts

University of Melbourne

University of Miami

University of Minnesota

University of Montana

University of Nebraska

University of Nevada, Las Vegas

University of Nevada, Reno

University of Pennsylvania

University of Rochester

University of South Africa

University of Southern California

University of Utah

Wake Forest

Washington State University

Wayne State University

Western Michigan University

Platforms

• 21 SciVal Experts• 13 VIVO• 11 Profiles RNS• 4 Elsevier Pure• 3 Home Grown

Methodology, step 1 of 4Get a list of ~every profile

Methodology, step 2 of 4Pick 500 random users per site

Methodology, step 3 of 4Search Google for “name institution”

Methodology, step 3 of 4 cont’dHow we picked institution names

1. First, look at the domain name of the homepage for the institution– profiles.ucsf.edu → www.ucsf.edu → ucsf– vivo.experts.scival.com/indiana → www.iu.edu → iu

2. If the site’s not on the main domain, and the RNS URL includes a common variation of the name, use that– vivo.experts.scival.com/indiana → indiana

3. If #1 and #2 give different results, use both (with OR)– indiana OR iu

Methodology, step 4 of 4:Record the search ranking position

#1

Methodology, step 4 of 4 cont’d:Why only consider the top 3 results?

The first 3 organic search results made up 71% of all non-mobile Google organic clicks.

Source:“Google Organic Click-Through Rates in 2014”Philip Petrescu, Advanced Web Rankinghttps://moz.com/blog/google-organic-click-through-rates-in-2014

Methodology

• Looked at search rankings for 24,583 profile pages across 52 sites

• Used the RankTank keyword rank checker tool to automate the process– http://www.ranktank.org/

In summary:

What % of a site’s profiles appear in the top

3 Google results for“First Last Institution”?

(e.g. “brian turner ucsf”)

3. THE SEO LEADERBOARD

The Top 101. findanexpert.unimelb.edu.au (95%)2. profiles.umassmed.edu (91%)3. profiles.ucsf.edu (88%)4. vivo.med.cornell.edu (81%)5. profiles.ucdenver.edu (76%)6. profiles.bu.edu (74%)7. connects.catalyst.harvard.edu (72%)8. gru.pure.elsevier.com (67%)9. experts.umn.edu (64%)10. profiles.psu.edu (63%)

Between 25-60%11. scholars.opb.msu.edu (59%)12. vivo.scholars.northwestern.edu (55%)13. scholars.northwestern.edu (55%)14. ohiostate.pure.elsevier.com (54%)15. experts.scival.com/unisa (45%)16. experts.scival.com/ctsicn (44%)17. profiles.ucsd.edu (42%)18. vivo.colorado.edu (42%)19. profiles.stanford.edu (38%)20. profiles.tsi.wakehealth.edu (38%)21. umaryland.pure.elsevier.com (37%)22. www.icts.uiowa.edu (35%)23. profiles.sc-ctsi.org (35%)24. experts.scival.com/uic (30%)25. hawaii.vivo.ctr-in.org (27%)

26. experts.scival.com/cwru (23%)27. profiles.jefferson.edu (23%)28. experts.scival.com/jhu (22%)29. vivo.scripps.edu (20%)30. unlv.vivo.ctr-in.org (20%)31.

experts.scival.com/wayneresearchconnect (19%)

32. experts.scival.com/wsu (18%)33. vivo.upenn.edu (17%)34. umt.vivo.ctr-in.org (16%)35. experts.scival.com/wmich (15%)36. scholars.duke.edu (15%)37. www.urmc.rochester.edu (15%)

Between 15-25%

Under 15%38. vivo.ufl.edu (14%) 39. msu.vivo.ctr-in.org (13%)40. experts.scival.com/nebraska (13%)41. unr.vivo.ctr-in.org (12%)42. temple.pure.elsevier.com (11%)43. experts.scival.com/nau (11%)44. experts.scival.com/arizona (10%)45. experts.scival.com/asu (9%)46. experts.scival.com/ucdavis (9%)47. experts.scival.com/miami (7%)48. experts.scival.com/einstein (4%)49. experts.scival.com/utah (2%)50. vivo.library.tamu.edu (1%)51. experts.scival.com/indiana (1%)52. experts.scival.com/ohsu (1%)

4. THREE THINGS WE LEARNED

#1. Use your own domain

Institutional domain? (e.g. vivo.cornell.edu)• average score = 49%

Unrelated domain? (e.g. experts.scival.com/asu)• average score = 21%

#2: Software counts

1. findanexpert.unimelb.edu.au Custom2. profiles.umassmed.edu Profiles3. profiles.ucsf.eduProfiles4. vivo.med.cornell.edu Vivo5. profiles.ucdenver.edu Profiles6. profiles.bu.edu Profiles7. connects.catalyst.harvard.edu Profiles8. gru.pure.elsevier.com Elsevier Pure9. experts.umn.edu SciVal Experts10. profiles.psu.edu Profiles

#2. Software counts

• Custom software average score = 56%

• Profiles average score = 56%

• Elsevier Pure average score =

42%

• Vivo average score = 26%

• SciVal Experts average score = 23%

#1+2: Software and domain matter!

• SciVal Experts + institutional domain average score = 59%

• Profiles + institutional domain average score = 58%

• Custom + institutional domain average score = 56%

• Elsevier Pure + unrelated domain average score = 42%

• Profiles + unrelated domain average score = 35%

• Vivo + institutional domain average score = 31%

• Vivo + unrelated domain average score = 18%

• SciVal Experts + unrelated domain average score = 14%

#3. Get incoming links

<a href="http://your.site.here/">

#3. Get incoming links:inside pagerank

#3. Get incoming links:search engine ranking factors

#3. Get incoming links:search engine ranking factors

Links to the page

Links to thesubdomainor domain

#3. Get incoming links:linking root domains

A linking root domain is a domain under a public suffix that includes links to your sites.

• *.cnn.com CNN• *.ox.ac.uk Oxford

University• *.anoka.k12.ca.us Anoka School

District, Minn.

#3. Get incoming links:the top 3 sites

1. findanexpert.unimelb.edu.au (95%)

2. profiles.umassmed.edu (91%)3. profiles.ucsf.edu (88%)

#3. Get incoming links

findanexpert.unimelb.edu.auhas 488 linking root domains:

• newscientist.com• f1000.com• anl.gov• duraspace.org• electionwatch.edu.au• and 483 more root domains…

#3. Get incoming links

profiles.umassmed.eduhas 249 linking root domains:

• en.wikipedia.org• grants.nih.gov• theguardian.com• bloomberg.com• nih.gov• and 244 more root domains…

#3. Get incoming links

profiles.ucsf.eduhas 858 linking root domains:

• sourceforge.net• harvard.edu• ucsf.edu• universityofcalifornia.edu• ucsfhealth.org• and 853 more root domains…

#3. Get incoming links:why (diverse) incoming links matters

1 10 100 10000%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Linking Root Domain Countvs. Google Top 3 results %

# Linking Root Domains (log)

% o

f "N

am

e [

School]

" in

top 3

Google

searc

h

resu

lts

5. HOW TO BOOST YOUR RANKINGS

Photo: Ryan McFarland

1. Be worthy of love

• Most people care about people, not generic information-finding site

• Make profile pages beautiful and chock-full of information, so people will want to link to them

2. Establish benchmarks

• Install Google Analytics on every page

• Learn how to use it– read Web Analytics 2.0 by Avinash

Kaushik

3. Get good with Google

• Add a sitemap.xml (sitemaps.org)

• Register on Google Webmaster Tools to:• register your sitemap• catch indexing errors early• link to your Google Analytics account

4. Look good with Google

<title> tag

<meta name="description">

Schema.org people metadata

URL

Virtuous cycle

1. Some people link to you2. You show up on Google3. More people see you, and link to

you4. You do even better on Google

5. Show the world you’re worthy of love• Get campus sites to link to your

homepageas a trusted campus resource

• Get campus sites to link to individual profiles from departmental profiles, news stories, directory, etc.

• Encourage reuse of your data via APIs, and ask for a link back as attribution

If that works…

• Some researchers will link to their profile pages on their own sites

• Some blogs and social media will link to your profile pages as authoritative sources

• Some departments may link to your profiles because your data is more current than theirs

The state of

the union is

strong.

The state of

the union is

weak.

The state of

the union is

mixed.

Thank you.Anirvan Chatterjee· @anirvan

– profiles.ucsf.edu/anirvan.chatterjee– [email protected]

Leslie Yuan· @leslieyuan– profiles.ucsf.edu/leslie.yuan– [email protected]

Brian Turner– profiles.ucsf.edu/brian.turner– [email protected]

Eric Meeks– profiles.ucsf.edu/eric.meeks– [email protected]

Find this online

bit.ly/rnsseo