Here’s what we have learned about the item pages, broken down by source:

Web Analytics

1. Our most-accessed items get viewed mostly via external links, especially from social media tools (like StumbleUpon) and Google Images.
2. More than 3/4 of item page views are for the medium image view as opposed to the details view.

Usability Tests (Spring 2008)

1. Users recognized the social media tools we chose to put in the ‘Add To’ section, but didn’t anticipate using those particular services.
2. Users were able to easily toggle between the details and the image view.

Feedback Forms and Other User Feedback

1. It’s good to have multiple image sizes available.
2. It’s unclear how to get to all of the details for an item.

Interviews with our Collection Sponsors

Desired improvements:

1. A comment box on items for users to contribute information.
2. Better page-turning for multi-paged items.
3. A way to get a citation for each item.
4. Connection to info about item’s archival source collection.
5. Easier printing, especially for multi-page items.
6. Clear display options for seeing the item and/or its metadata.
7. Multiple export options–especially for multi-page items.
8. Cleaner metadata display, separating technical from descriptive metadata.
9. Customized presentation of items for certain collections or item types.
10. Clarity to user what the rights / acceptable uses are for each item.
11. A way for items to be identified as part of more than one ‘collection’, especially broader collections like ‘documentary photographs’ or ‘moving images.’

CIT’s Ideas for Helping our Images be Used for Instruction

1. Integration w/social media tools (like ShareThis) and blogging platforms.
2. A ‘download’ button (where allowed) w/multiple size options, esp for use in PPT.
3. Copyable ‘Embed code,’ including citation.

Other Ideas from the Implementation Team

To accompany all the ideas and info gathered above, here are a few additional ideas we’ve had in our implementation team discussions:

1. Ability to export items in PDF, especially for multi-paged items
2. Automatically-generated ‘Related Items‘
3. Integration with our internal metadata editor so staff can edit from public interface
4. Permalinks with URL handler
5. Full-text display alongside images where available
6. Aesthetics: page should be wider; item title needs more prominence

Now, what do we do with all this feedback? First, we’ll share good example interfaces from around the web that tackle some of these same issues in various ways. Once we’re feeling fully inspired, we’ll draw up some prototypes to share. Stay tuned!

Not surprisingly, some search queries are common and reveal the subject strengths of our digital collections.  For example, the top three queries that bring users to our collections are “sheet music,” “ad access,” and “history of advertising.”

After scanning through thousands of these search queries, several distinct categories emerge: the known-item query (an exact title in quotes), the URL as query (e.g.  http://library.duke.edu/digitalcollections/adaccess/), and the format query (e.g. “diaries” or “manuscripts”), among others.  The most entertaining category, however, is the query issued in the form of a question.

Below are some of the important questions our users have asked with links to where they’ve found answers to those questions in our digital collections.

• “what is electronics?”

• “can you rent cars at Hertz for fun?”

• “what to wear with a corset”

• “what is a funeral chair?”

• “what are 1955 war bonds worth today”

• “what are chemical properties of Listerine?”

• “will I be sick tomorrow?”

The slides are very basic. Included are some figures extracted from previous blog posts (http://library.duke.edu/blogs/digital-collections/category/assessment/) as well as ‘greatest hits’–the most-accessed item from each collection.

• videos uploaded to iTunes U the week of September 21, 2008
• videos uploaded to YouTube the week of December 14, 2008

Each service provides some distinct advantages over the other. A basic matrix of differences can be found here:
http://www.oit.duke.edu/web-multimedia/multimedia/YouTube/index.html#faq

Usage

To gauge use, we looked at about 8 weeks of data in both systems following the publication of the videos in YouTube. There were 16,412 YouTube views, 993 iTunes downloads, and 392 iTunes previews.

Diamonstein-Spielvogel Video Archive Usage Stats
Dec 14, 2008 – Feb 8, 2009

Notes:

• An iTunes Preview is when someone double-clicks a video to watch it without downloading it.
• An iTunes Download is when someone clicks the “GET MOVIE” button to download the video to their computer.
• YouTube does not allow downloads
• We can’t tell how many times iTunes U videos have been watched after they have been downloaded.

Where Are All These Views Coming From?

The iTunes views & downloads all happen in the same place: the collection interface in iTunes U. We don’t know how many originated with an external link (URL’s are available for individual iTunes tracks). With the YouTube videos, people are finding them in a multitude of ways. Here are example stats for our most-viewed video (as of today, Feb 25):

But no single discovery method for YouTube videos stands out as ‘primary’ across the collections. Of our five most-viewed videos (see Top 10 below), Winogrand has been viewed mostly through an embedded player (32.0%, primarily from a street photography blog), Meyerowitz was discovered most frequently as a ‘related video’ to other YouTube videos (26.0%), Maloof was viewed most frequently via external links (35.0%, mostly from a woodworker community site), and Dine was found mostly via keyword searching in YouTube (34.0%).

Top 10 Videos

YouTube:

1. Visions and Images: Garry Winogrand, 1981 (2,576 views)
2. Visions and Images: Joel Meyerowitz, 1981 (918)
3. Visions and Images: Elliot Erwitt, 1981 (607)
4. Handmade in America: Sam Maloof (546)
5. Inside New York’s Art World: Jim Dine (470)
6. Inside New York’s Art World: Lee Krasner, 1978 (454)
7. Visions and Images: Harry Callahan, 1981 (377)
8. Inside New York’s Art World: Motherwell (358)
9. Visions and Images: Arnold Newman, 1981 (331)
10. Visions and Images: Duane Michaels (291)

iTunes:

1. American Architecture Now: Frank Gehry, 1980 (60 downloads)
2. American Architecture Now: Peter Eisenman, Jaquelin Robertson, 1984 (35)
3. Visions and Images: Garry Winogrand, 1981 (28)
4. American Architecture Now: Stanley Tigerman, 1984 (28)
5. Visions and Images: Elliot Erwitt, 1981 (27)
6. American Architecture Now: Philip Johnson (27)
7. Visions and Images: Joel Meyerowitz, 1981 (26)
8. American Architecture Now: Richard Meier (24)
9. Visions and Images: Arnold Newman, 1981 (22)
10. American Architecture Now: Michael Graves, 1980 (21)

Comparison:

Only four videos appear in both Top Tens, so there is disparity in which videos are most popular in both systems. 6 of the top 10 videos in iTunes are from the American Architecture Now series yet not a single one of these even cracks the top 10 in YouTube. Does this mean architecture enthusiasts are more apt to be iTunes users? Probably not. A more likely explanation for this phenomenon is the arrangement of tabs for the various series in the iTunes interface:

American Architecture Now is the first and default tab, thus videos in the other series are initially hidden from view. We may reconsider the tab arrangement for future collections.

Comments & Ratings

YouTube’s comment and rating functions have both been used considerably in the collection. As of today (Feb 25) we have received 35 total comments on 24 videos (High: 3). 84 ratings have been cast on 41 videos (High: 9). 36 of the 41 rated videos have a 5/5 star average rating, 1 has a 3/5 average, and 4 have a 1/5 average. The high ratings indicate that people are finding the videos to be valuable, and that is reinforced by the grateful messages that comprise the majority of the comments received.

Other Observations

It’s unclear to what extent the services complement each other, although we have already had a patron contact us through YouTube inquiring about how s/he could download the videos (iTunes makes this possible). It’s also conceivable that someone might download a video from iTunes, watch it, and then use the YouTube version to discuss or share it.

What do people actually do when they come to the Duke Digital Collections homepage as it is now?  One way to tell is to review our server logs.

Here’s a look at the year in review.

2008 At a Glance:  68,325 homepage hits

38,481 (56.32%) leave the page without following links or searching (or in a handful of cases follow a link to a site outside of our library.duke.edu or scriptorium.lib.duke.edu domains).

Observations:

From this page, people were nearly three times more likely to visit individual collection homepages than they were to do a cross-collection search. This is consistent with what we found earlier: Three-quarters (76.7%) of queries in 2008 were done within a single collection, one-quarter (23.3%) were cross-collection.

This stands as another good reminder that while we have gone to great lengths to build an integrated discovery system across the collections, users more frequently gravitate toward the individual collections themselves.

* We can’t really determine how many of the recursive referrals back to the homepage were clicks to anchored links like this vs. clicks on the Digital Collections logo in the header.  (edited 2/23/09 –SCA)

Header & Footer Links

How often did people click our header and footer links from our homepage?  As it turns out, not very often:

Header About/Help:

Footer:

People might be more inclined to seek the kind of information linked to in the header and footer at a more immediate point of need, for example: clicking Help or Ask Us Now following an unsuccessful search or clicking Use & Reproduction from an item page.

A Tale of Two Layouts

In April 2008, we modified the homepage to group collection links by theme, introduced an A-Z list, and made it clearer which collections are included in cross-collection searches.  The two versions are not radically different, but the data above is a combined view of about 3.5 months of data from the old and 8.5 of the new.  Click to enlarge for a closer look:

Jan 08 homepage (approx.)

Apr 08 homepage

Here are two separate charts (below): one for the first half of 2008 (Jan – June) and the other for the second half (July – Dec). The one on the right includes more collections (we introduced several throughout the year) and may be a more representative look at the usage.  Also keep in mind that the collections vary in size (larger collections have more items *to be viewed* and often have more ways to formulate queries).

Click to enlarge:

Jan – June 2008

July – Dec 2008

Interesting findings

1. Cross-collection vs. Within-collection searches.

   Our system was designed to enable easy searching across multiple collections. Our users definitely have used that capability, though most queries were still performed within the context of a single collection. For all of 2008, 23.3% of total queries (1,436,604) were searching across multiple collections; 76.7% were within a single collection. The second half of 2008 saw more cross-collection querying (27.1%) than the first (20.2%). Of course, the more collections we build, the more useful (and interesting) it is to search across them. Of all the cross-collection queries (334,425), only 2.25% (7,533) directly originated from our Digital Collections homepage. The rest of the query action happened inside–as followup (new searches or refinements) to other searches, on item pages, or on individual collection home or contextual pages.

   8.08% (27,026) of cross-collection queries came from individual collection homepages.  In fact, the “Search All Collections” search box on both the Ad*Access homepage (3.49% — 11,657) and Historic American Sheet Music homepage (2.97% — 9,916) was used even more frequently than the master search box on the Digital Collections homepage. We built our system interface with the challenge of maintaining the branding and the identity of the individual collections while integrating their discovery and display into a common site.  Beyond putting in an overarching homepage with a common search box to accomplish this feat, it appears we’ve gotten even more significant leverage from the reverse angle: connecting the individual collection websites back out to our full breadth of digitized collection content.

2. Details view vs. Medium Image.
   
   Of all the overall item views in 2008, 21.5% were for the details (metadata-centric with thumbnail) view and 78.5% were for a medium image (image-centric with some metadata on the left). This was a fairly consistent finding for most of the collections. Note: Large image hits were not tracked, and multi-page items like sheet music have more images to be viewed than details.

   

   Does this mean that users are more interested in the visual content of the collections than the descriptions? Perhaps, although we found that the details page was used heavily in one particular collection, the Sidney D. Gamble Photographs. (50.6% details view, 49.4% medium image). Why? The metadata is fairly sparse, although to supplement it, we introduced linked terms (tags) in the details view for this collection, extracted from the item titles. Clicking “above,” “boat,” “gorge,” or “pagoda” here will retrieve other items in the collection with those terms. It’s possible that this extra navigation/discovery feature is what makes the details view more desirable.

3. Comparing Queries, Item Views, & Homepage Views

   These are obviously imperfect metrics open to a lot of interpretation, but the following may prove useful. Looking only the collections in our new system (’Tripod’) in the aggregate:

   	Jan – Jun 08	Jul – Dec 08	Total 2008
   Homepage Views 1	316,829	386,532	703,361
   Queries 2	796,255	640,349	1,436,604
   Item Views 3	1,107,593	1,328,068	2,435,661
   Item Views per Query	1.39	2.07	1.70
   Item Views per Homepage View	3.50	3.44	3.46
   Queries per Homepage View	2.51	1.66	2.04

   1. Homepage Views: includes master homepage for all digital collections plus homepages of individual collections.

   2. Queries: any page view with ‘/search/’ in the URL string. Includes cross-collection & within-collection queries. Is not just search box submission (re-sorts, facet limit selections, page 2+ of search results, canned searches, & some browse pages all count).

   3. Item Views: includes either detail or medium view. Also includes page 2+ view of multipage item, so count is likely skewed for collections with paginated objects (i.e., sheet music or comics)

   So, very generally speaking, for every homepage view, two queries are performed and for every set of search results, nearly two items are viewed. Also, about three and a half items are viewed for every one homepage view.

   Let’s look at examples of individual collections that deviate from the averages.  These figures may ultimately only help to illustrate that user interaction differs across the various collections:

   Item Views per Query. HIGH: vica (39.14), blake (15.46), bloomsbury (11.99), strong (8.80), hmp (5.13). // LOW: mma (0.80), gedney (0.93), diap (1.09), songsheets (1.26).

   Item Views per Homepage View. HIGH: hmp (11.45), duc (10.97), gamble (10.08), diap (5.95), hasm (5.22). // LOW: mma (1.46), songsheets (1.73), vica (1.97).

   Queries per Homepage View. HIGH: duc (5.44), diap (5.44), gamble (3.98). // LOW: vica (0.05), blake (0.17), bloomsbury (0.27), strong (0.35), eaa (0.90).

4. Homepage Views
   
   The Digital Collections homepage views increased by 34.7% from the first half of the year to the second (29,239 to 39,373). This may be a result of better marketing or perhaps the addition of several new collections after July created both more attention and places from which to link back to the homepage.Though the most popular individual collection homepages are still hit more frequently than the overall homepage. From July – December, the homepages of Ad*Access (119,376), Emergence of Advertising in America (76,847), and Historic American Sheet Music (48,141) were all visited more frequently than the Digital Collections homepage (39,373).
5. New Collections on the Block
   
   The Sidney D. Gamble Photographs collection certainly made a splash, introduced in the first week of July. From July to December, it had the 2nd-most queries of all the collections (103,047), 4th-most item views (261,019), and 5th-most homepage views (25,898).

   Of our five collections introduced in October, American Song Sheets is off and running with the strongest usage (3,719 homepage views, 5,080 queries, 6,417 item views in a little over two months).

6. Legacy Collections
   
   Finally, let’s not forget our existing digitized collections that remain in flat HTML files and have not been migrated into our new system. Several of these collections are still being used more heavily than many of the collections that we have created or migrated in 2008. The Duke Papyrus Archive stands out in particular: on the whole year, it had the 4th-most homepage views (55,739) and item views (342,319).

Related post: Referrer Stats for 2008

Now that 2008 is over, we’ll be posting a few charts & graphs that illustrate some interesting trends in how our digital collections (and our shiny new system) have been used in the past year. This post focuses on “referrers,” or, those other sites that people come from that directly lead them to land on our pages.

OK, so what are we counting?

• Any referrals from outside of Duke Libraries websites that resulted in a Page View anywhere in library.duke.edu/digitalcollections/*, including our digital collections homepage, individual collection homepages, contextual info pages, item pages, or search results.

How many?

• 890,000 referrals from 10,000 unique external domains (all Duke library web sites/pages excluded). Only the top 9 individually account for more than 1% of external referrals, so there’s quite a long tail.

Notable External Referrers

Of the 10,000, some stand out in particular…

* Hundreds of Google domains are also peppered throughout (google.co.uk, google.ca), though this is just http://google.com/

It’s no surprise that Google leads the pack and far outpaces any other external referrer.  More than 28% of the times people landed on one of our pages from somewhere else, Google was that somewhere else.  On the other hand, Google Images (images.google.com) didn’t produce much traffic to our site, despite it being a potentially great discovery resource for the kinds of things we’ve digitized.  Some further SEO research and tweaks might help us ensure we’re being found there in the future.

Battle of the Social Bookmarking Sites

Our system has easily bookmarkable URL’s for any item or set of results, so we put a handful of social bookmark quick-links at the bottom of the screen (example above) to encourage people to share and discuss. Whether or not our links were used for bookmarking, here’s a look at the traffic these and other bookmarking sites drew to our collections:

* Estimated, after combining stats from multiple domains

Astonishingly, the social bookmarking site/service StumbleUpon was our #2 external referrer, second only to Google.  It drove far, far more traffic to Duke Digital Collections pages than the rest of the social bookmarking tools combined, and indeed, our own library website (see below).

Examples of StumbleUpon in action: This page (caution: potentially sensitive content ahead) made this item by far the most-viewed item of our entire digital collections during 2008 (over 10,000 views; stay tuned for more info in a future post).  And here are StumbleUpon pages for our Digital Collections homepage, a collection homepage, and a result set.

Traffic From Our Very Own Library Website(s)

To put the external referrers in context, here are examples of how many referrals to Digital Collections pages came from other pages in Duke Libraries websites.

* Rank & Percentage not calculated as these are “internal” referrers/referrals

What jumps out here is a very small tally for hits coming from our library catalog interface (aka OPAC).  There’s currently little integration between the collections (and items within) and our catalog interface(s); currently some of the collections are cataloged as electronic resources (here’s an example), though that’s about it.

Our library homepage has also not been a particularly good facilitator of traffic to Digital Collections.  We started the year with a “Digital Collections” link in the CHECK THIS OUT section, then the link moved into LIBRARY SERVICES in the fall (see below).

We’re hoping for a more effective and engaging representation of Digital Collections on the library homepage in 2009.

Traffic from Elsewhere at Duke

* Estimated, after combining stats from multiple domains

Our friends at Duke News & Communications have featured several of our collections throughout 2008.  Here’s an example article from November.

More info to come

We have more facts, figures, and observations to share, so be on the lookout for more soon.  Let us know if you’ve noticed any interesting trends in your referrer stats, too.  Feel free to leave us some comments, observations, or questions here!