I recently purchases William Gibson’s “Zero History” Kindle edition for the iPhone and read through a hundred pages, ocassionally encountering underlined sections but not giving them any attention. When a particularly Gibson zinger was highlighted, I tapped: “Some very considerable part of the gestural language of public places, that had once belonged to cigarettes, now belonged to phones.” … 56 other people highlighted this part of the book.

You can read this annotation, and copy paste as I did, from Amazon’s top highlighted page for this book. Suprisingly, this content is not available from the Zero History product page (book or kindle), as it seems to me to complimentary to, and more impartial and informative, than many reviews. These sections form the same kind of human smart summarization that link text (e.g. the text inside hyperlinks) provides for web search engines.

Trompsing through top highlights is rewarding for popular titles like Freaknonomics and “outliers: the story of success”. This use case looks to be specifically not supported by Amazon, so you’ll have to use precise query syntax, rendering results for short titles like “Blink” a mess — adding the author name to the query cleans the results up.

So how’s the book? Good, if a far cry from his intensely futuristic cyberpunk roots, with the iPhone playing such a critical role that when the MacBook in play was identified as an Air, it seemed like Apple should be paying for the product placement!

Table 1: Comparison chart of three long-term studies

The replication of the finding is welcome, and the lack of novelty didn’t prevent Techmeme buzz from sprouting up, probably helped significantly by the excellent visual heatmap. Nice work Mozilla peeps! Here’s a zoom on the back button findings by clicks per user and % of users:

There’s been supposition that the growing use of tabbed browsing is reducing the importance of the back button. Certainly, opening search results into new tabs reduces the need to go back to get to search results as do parallel scenarios in non-search browsing. My recent analysis of tab usage from SRP and non-SRP pages shows that open in new tab is common but still a minority use case for search result browsing.

The TestPilot analysis doesn’t specifically address the relative prominence of back button versus tab switches or links inside pages because it only includes browser chrome — defined as the area surrounding the tabbed browser. I’d love to see a replication of this study which counted clicks on tabs as well as interior to the browsed page UI.

In addition to improvements to logging to allow individual tabs to be tracked accurately as users open and close tabs around them, the latest study includes tagging of search result pages (SRPs). This is particularly interesting to me as clickstream patterns around search results are highly useful for evaluating search quality and even generating data for machine learning, but tabbed browsing (and new window strategies) may seriously affect how a server side log appears and complicate accurately reconstructing the user sequence.

How is tabbed browsing used in search?

I found that 96% of the users in the 2000 user data set had a URL loaded from an SRP. 86% of these users also opened a new tab from an SRP, suggesting that at least in this audience, opening links in tabs from an SRP is a very well known strategy available to over 90% of users.

The denominator for the following stats are *either* tab focus changes or page loads in a tab. For those SRPs, 40% of changes were page loads in the same tab and 20% were open in new tabs. Thus, open in new tab doesn’t seem to be the predominant strategy. However, I would expect navigational queries, estimated by research at 30% of web search engine usage, not to invoke a open in new tab strategy.

General behavior following a page load

Looking at successive events from a load event to specifically to understand search and tabs, we see users are only about 8.5% more likely to open a page following an SRP in a new tab than for any other page load.

My SQL for creating a MySQL table, importing the data, and creating a derived sequence table is available on github:

For my analysis of previous tab usage TestPilot datasets, see this post and my visualization galllery.

Back to tracking SRP behavior, it looks like tabbed browsing is only barely more of an issue for pathing in search behavior than in general web activity. Anybody know of an analytics solution that tackles this? Creative use of window.name would likely maintain pathing integrity even when tabs are opened in the background.

• Language Learning
  •  ClozeFox
  •  LangLadder (aka MemWord)
• Game-based Learning
  •  NetDetective
  •  Studytroll
• Annotation
  •  HooverNotes
  •  Expression Widgets
• Trails
  •  Mupple
  •  LineHive

I’m personally especially excited to see the Trails implementations as I’ve long regarded that as a missing piece of the Web 2.0+ world.

A couple projects break the mold here: Cohere is superset of annotation in the knowledge construction & management space and part of a larger .edu project. Rubrick facilitates assessment in a collegiate environment.

We’re having great fun at the Design Camp leading into SXSWi, track it via #jet4learning. Looking forward to spending more time on the future of JetPack today and the final presentations.

One notable issue industry wide for faceted browsing systems is the affordance for reducing the constraints. In this non-ebay example, only the bread crumb enables me to remove my price filter directly. eBay extends the facet removal affordance with checkboxes beside attribute filters. I like download.com’s summary of applied filters and X buttons better as a more salient version of a breadcrumb bar. This model is akin to the early Flamenco project rendering. Read more about faceted browsing interfaces at Marti Hearst’s at searchuserinterfaces.com. I’ll be studying the refinement patterns at eBay for ways to advance the state of the art.

I’ve crafted a couple firefox chrome templates for use in paper prototyping and will be talking through choosing the right Jetpack UI option as well as larger issues around UI design in Seminar 2 tomorrow. As part of the larger context, I’m encouraging pre-coding efforts at evaluating concepts and UI expressions.

Get the templates and my presentation materials at lab.mozilla.jetpack. The entries cover a wide range of educational contexts and I’m confident the competition will set a high bar!

All of the basic rules of HCI apply: keep the user informed, match between system and real world, recognition not recall, speak the users language, provide shortcuts, etc. Then we get a whole slew of new concerns around facilitating learning.

I see great opportunities for browser addons to aid sense-making in the early stages of learning a space as well as in elaboration and chunking. Chunking is how humans get beyond the limits of short term memory (the common 7 plus or minus 2 rule) to create richer mental representations.

Even more so, learning ehancements could facilitate connections between people with shared learning objectives, enable novice-expert interactions, and create other social opportunities abound beyond the more obvious rehearsal tools (e.g. flashcards) and ease of information access utilities. Getting the right information at the right time is critically important for the integration of new knowledge into existing understanding.

The focus of the design challenge is on JetPacks but full firefox extensions can also be entered. While add-ons have a huge range of user interface options, the JetPack solution set of UI is more constrained. This should be a nice sandbox for UI patterns. I’ll be teaching a couple classes along the way to a finalist workshop preceding SXSW Interactive.

I’ve got one consumer ready JetPack out there for rotating tabs for a display kiosk but look forward to crafting many more! I’m even dabbling with a multi-touch enabled Firefox on a HP tx2 tablet w/ Windows 7.

Check out the official challenge website for more info. You’ve only got a month to make your submissions. I’ll see the winners in Austin right before SXSWi.

Timing & Event Sequences

Here’s a look at what happens *after* a page loads in a firefox tab:


~83% of the time another page is loaded with an average view time of 6 seconds. Opening a new tab makes that view time shorter. Closing a window is a much more deliberative action than closing a tab, averaging ~60 seconds
.

Timing & Sessions

An early study of Netscape 1.0 browser files determined that browsing episodes tend to fall within 20 minutes. This finding seems to hold true today as indicated by the graph inflection point and has technical bearing on in-memory cookie retention.

The graph is a zoom on the 80% of sessions under 90 minutes. We see 75% of session fall within half an hour. There’s a definite research opportunity to explore overall versus site versus task session durations, though this data can only inform on site and overall duration.

Getting to the Nitty Gritty

At some point in a complex data analysis it becomes useful to go down to the individual data and make sure you understand what’s happening. Here’s a view of a single user engaged in a very length 4 hour browsing session.

The table shows how this user compares to the average in terms of the frequency of top sequences. This user browses a bit more aggressively than the average user, loading more pages per tab. He or she also is more likely to load a new page when a switching to a tab. On the right is a visualization (click for all 4000px in height via image or view html) of the sequencing of these actions. Each column represents an open tab and the colored cell is the tab of current activity. Height increases beyond 20 seconds to a maximum of about 8 minutes in this viz.

It’s interesting that the visualization shows the user working orderly through tabs, largely moving a single tab at a time away from the current focus except when jumping to tab 1. I’m still working to fully understand what different types of transitions communicate around tab closes, open in new tab foreground and background, etc.

Here are a some static HTML views of sessions starting with a “tab addict” who’s carry around over a dozen tabs that don’t get visited in either this short or longer session. On the other hand, here’s an example of a quick 5 minute episode in which the user revisits all of his open tabs

Sharing

My R and MySQL code is on etherpad and I’ve made a 66mb MySQL database dump available with session tagging and sequenced data for the Vanilla 30 users.

There is a generic Vanilla 30% data table (vanilla30) and a table prepped for session grouping and self-joins for chaining (vanilla30seq). The live session visualizer and data-browser is not quite ready for sharing. I’ve also imported the TreeStyle data set.

Moving through this we can observe:

• **A long tail of average tabs, with an inflection around 20 tabs. 75% of users average under 8 tabs, the 50% mark is at 4.1 tabs. Note the color coding is explained by this first cell.
• **In the top right cell, we see a bimodal distribution, in which most users with high maxTabs have high average tabs, but a subset of “tab bingers” who more occasionally use large numbers of tabs. More below…
• **One puzzling aspect is the presence of 22 distinct values for day. Though the timespan of the study is one week, apparently the start and end times are not the same.

This dataset is pretty large, with 7749 points analyzed by day by user. I imported the data into mySQL and am using a hybrid of SQL and R to conduct exploration (see the code).

I moved to a more granular analysis of (the first 25,000 rows) of hourly averageTabs and maxTabs.Here we see the bimodal distribution more strongly. Some minority of users break with the general relationship between average tabs and max tabs. These are the “tab bingers” or alternatively speaking, the clean uppers, who go from few tabs to many and back to a few. Futher analyses will have to be done to identify the pattern here.

In fact, looking at the speed of change of open tabs for the 50% of heavy users with more than 4 average tabs open is one of the more intriguing opportunities discovered so far (see the # tabs per navigation action by Dubroy). This is possibly just an artifact of using average, as MySQL doesn’t have a median function, but it does suggest there may be spiky versus constant tab “junkies”.

My first tweet on this dataset said that 50% of users never go beyond 13 tabs. In fact, the number looks lower than that in this 50% subsample. Seven +-2 likely holds for number of tabs open for most users. There are two interesting exceptions to this rule, called out in the “projection pursuit” ggobi video below:

In the video, spiky tabbers are in grey and tab addicts are plotted with hollow squares:

It’s challenging to derive meaningful and concise conclusions from complex datasets like this one. To that end, I’m attempting with this analysis, like I did with the Places Stats project, to share my analysis code for open source collaboration. I’ve even done a video on using R + GGobi.

This first round of analysis barely touches the surface of the interesting aspects to this data set and suggests two tentative areas for further inquiry: Average tab users currently only use a handful of tabs, can we make life easier on them? Two types of users venture into the >10 tab world, habitual tabbers and spiky tabbers (e.g. addicts and bingers in my unPC terminology) — why?

Further work with this dataset is going to require more sequential analyses, walking the data to generate more granular metrics on growth and constriction of # of tabs, as well as looking at spawning methods and the role of windows. That doesn’t even begin to get into the folks who optimize tabs by using extensions! Avoiding averages is also critical in general and to connect to Dubroy’s thesis study. Does MySQL do UDFs?

About:Search

A new version is up and now on addons.mozilla. Here’s the premise… Search is *hard* to evaluate. Trust me. It’s especially hard to evaluate subjectively, where brand and other esoteric factors can skew your impression. Which search engine really works best for you?

Here’s a composite of the drill down pages for the big players, with the live.com -> bing transition represented.

The queries with no clicks metric is one of my favorites. You can watch a lecture from Thorsten Joachims that discusses it and other search metrics at videoLectures.net. Basically, these queries often represent an expectation violation where you the human thought you had provided the search engine with enough information but it was unable to respond satisfactorily.

I developed about:search with intention of helping people get objective stats on the performance of the search engines they use. I found that my usage was too skewed to Google for the stats to be comparable. Beyond more QA on the abandoned query metric, I’m working on an engine rotator for the navbar search box to help folks give a collection of search engines regular usage.

About:Shopping

Shopping is another interesting area of important and richly faceted data in your browsing history. With a new gig focused on understanding shoppers, deal hunters, and modern day traders at eBay, I became reflexively interested in my own online shopping.

Hence, the alpha of about:shopping. It uses a small list of the big shopping sites and a heuristic around URL names to find your top shopping sites. Your feedback is welcome (in the comments, via via twitter, etc).

UPDATE: About:Shopping now includes the ability to tweet your top 4 shopping sites automagically and is available on addons.mozilla. Follow my ab0ut:you fun on twitter.