1. 1. using web analytics to measure baselines
   2. putting researcher profile pages front and center
   3. ensuring your site’s getting indexed by search engines
   4. thinking like Google—tweaking URLs and metadata to attract search engine users
   5. getting inbound links to establish reputation

Notes

• have grown traffic roughly 20-fold since first launch - have learned what can do and what not to do. No magic -- all the pieces work together

• http://profiles.ucsf.edu launched in October 2010 with lots of publicity -- started at 5,000 hits a month but has grown to 100K hits/month

• 37% CA

• 30% other USA

• 83% from search engines

• Growth Hacking 101

• measure baselines with analytics -- “Web Analytics 2.0” by Kaushik

• put people pages front and center

• register with Google Webmaster Tools

• segment on-campus vs. off-campus traffic

• 15% new visitors last month overall, but is 72% of on-campus users while only 20% of outside users

• ignore your home page -- users don’t land there -- 2.1% of Profiles users start on the home page, and 98% totally avoid the home page

• and only 21% of 2+ minute users start on the home page

• growth of home page views is pretty static while the rest of the traffic is growing

• people type a name into Google and want to go to the person’s profile

• so have cleaned up those profile pages

• more inviting, more data fields, bigger photos, etc -- look at good examples

• get indexed -- make sure search engines can see it

• have a dynamic site map of all the pages on your site -- check with Google webmaster tools

• make sure you’re not blocking via robots.txt (www.robotstxt.org)

• tweak URLs and page snippets -- search results show title, URL, and snippet

• customize the URL to make it more appealing -- e.g., profiles.ucsf.edu/eric.meeks

• make the snippet more readable -- not random pieces of text

• make the <title> on profile pages so it’s short and globally unique

• the <meta name=”description”> something like “Jane Doe’s profile, publications, research topics, and co-authors

• with Schema.org can add in a line of professional metadata

• follow instructions

• use google.com/webmasters/richsnippets

• make pretty URLs -- see a lot of researchers putting that URL in their email signature -- feels more personal

• and the pretty URL should be the “real” URL, not just a redirect

• Get inbound links -- self-reinforcing

• doing a lot with campus news department so every news story mentioning a researcher includes a link to the profile

• and have the link in the directory

• get links to Profiles on departmental pages

• work with departments to give them a feed -- or just link by saying “view on UCSF Profiles”

• include links in narrative bios

• create APIs for people to use

• document them on a developer-centric website

• online discussion group

• outreach to campus groups

• ask for attribution via a link back to Profiles -- you save them time and money and they give you links

• over 2 dozen sites now using Profiles data and linking back

• media mentions also link to individual profiles (instead of what is often a pretty crappy lab page)

Questions

• Do some people/departments see as competition for their web space?

• Profiles had some buy-in but no top-down direction to use it -- work closely with departments

• sometimes makes the Profiles higher than the department page, but few of them measure

• helps to be giving them data, and data that is clean

• E.g., UCSF School of Pharmacy partnered with the Profiles team and re-used, and worked out editing in Profiles -- don’t have to push it on them

• Pretty URLs?

• VIVO has the URI vs. the display URL -- don’t want to use our URIs as links on the display pages since appears then to have 2 URLs for the same thing

• The logistics -- how to assign a pretty URL to two people with the same first and last name

• could have preferred URL

• Eric -- still have the numeric URIs in Profiles, same as VIVO

• do content negotiation to display the HTML page with the pretty URL

• the links within pages are linked by URI, not URL

• Profiles had a pretty URL /websites/profiles/name -- better to shorten

• have a strategy to avoid name collisions that does a good job -- if 2 Eric Meeks, one of whom has middle initial

• have shared that algorithm -- pretty easy to get uniqueness without having to deal with

• but is worth doing

• BU has taken the UCSF code and modified it but put it into practice

• Jim -- do you retire the pretty URL?

• Eric -- the same case can be made for persistent URLs as well as persistent URIs

• Anirvan -- a lot of people have built this capability into

• having 2 URLs pointing to the same content

• if look at the accept headers -- if wanting XML or JSON, gets the data; otherwise direct to the pretty URL

• if use “link rel=canonical” to the pretty URL on the ugly URI “page” -- in the header -- do view source and do a search for canonical

• if a robot has scraped a URL with a bunch of query arguments

• do you do this for grants, equipment, etc.?

• No -- people come to Profiles overwhelmingly for people

• Patrick -- Scholars@Duke has encountered some resistance to competition -- stronger in the humanities than in medical world, perhaps?

• Eric -- we’re biomedical and the Profiles system does a good job for people. Resistance got smaller especially when offered background

• But had been other systems at other schools that used the word “profiles” so is a more confusing

• To be a successful research networking tool have to be a good website, and isn’t easy to do

• Duke has taken an aggressive approach with the widgets, but don’t underestimate how important it is to make a successful website to be a successful tool

• Jim -- did this presentation at an IFest and we haven’t done a lot yet -- thanks for helping to raise our vision up

• Anirvan -- worked in eCommerce and was excellent training

• Jim -- Nate Prewitt implemented microformats at the Hackathon

• Anirvan -- will the descriptions change? A short, one-line item

• Will be happy to share the slides via SlideShare

• Paul -- do you have before and after stats on the use of Schema.org tags?

• Anirvan -- have general stats; added Schema.org tags in late 2011 -- easy, and if Google is telling us to do it we should do it since they represent 82% of our traffic