Multi-Tab Browsing

Most browsers were initially designed to help users navigate the web sequentially in a single window. While improvements in software engineering and computational power have allowed modern browsers to concurrently render dozens if not hundreds of webpages as multiple tabs, within each of these tabs, information is often still represented in the same old-fashioned way, as a sequence of webpages.

Opening 20 tabs doesn’t necessarily make navigating the web easier. In fact, a user now needs to mentally keep track of 20 separate threads of information. No wonder many users report that they are overwhelmed by the exploding number of open tabs and feel lost in a swell of information.

We conducted a survey to understand how people browse the web using multiple tabs. The survey is intended to help us better interpret patterns observed in the first Context Graph experiment, and prepare for future work. In the rest of this blog post, we’ll go through the study goals, design, and highlight some findings.

Goals

The first goal of this study is examine if and how users organize multiple tabs, multiple windows, and (when applicable) multiple browsers.

Could we develop aided or automatic means of tab organization for Firefox users? To do so, we need to first understand and model the semantic relationships between the tabs. Thus, a second goal is to explore some open-ended questions: Could we and should we predict how users perceive the similarity between tabs? Could we predict how users group together tabs or open new tabs? If so, what signals might we need to capture (e.g., input features, output objective functions, etc.)?

Study Design

We deployed a Heartbeat study in February 2017 asking about 600 Firefox users to fill out a 5-page 21-question survey.

The survey contains four categories of questions related to (i) tab organization, (ii) window organization, (iii) new tab usage, and (iv) cross-browser usage. We included an additional category of questions on (v) new sessions to isolate timing effect, i.e., any differences in user behavior immediately after starting Firefox vs. after using Firefox for an extended period of time.

Highlights of Survey Results

Tab Organization

Approximately 50% of the survey participants report that they actively rearrange tab placement along the tab bar. Users most frequently organize their tabs by similar subjects, by time, and by webpage domain.

The top reason for tab organization is to have quick access to a small set of important or “anchor” tabs. The remaining tabs are then arranged relative to the anchors. Common strategies to re-find and return to the anchors include pinning the anchor tabs, or dragging an anchor tab to the far left or far right end of the tab bar. Examples of anchoring behaviors include using the email inbox as the root page when engaged in communication, or using an Amazon product page as the root during comparison shopping.

Window Organization

Approximately 40% of the survey participants report that they actively organize tabs by windows. They often use windows to separate work vs. personal browsing. To a lesser extent, they sometimes break out tabs related to a specific task (e.g., email, entertainment, gaming) into a separate window.

Notable Findings

Priority Queue or Countdown Clock

When users organize tabs by time, they often arrange the tabs in the order they expect to close the tab. Users describe the strategy as a priority queue (i.e., to help them keep track of the relative importance of the pages, so they can read them in order) or as a countdown clock (i.e., so that they know when they must complete the task associated with each tab).

Anchor Pages

Users overwhelmingly report that they mentally organize tabs relative to a small number of important or “anchor” tabs. Such an organization strategy is known in psychology as prototype theory and may have implications on how we should model tab organization and design tools to aid Firefox users.

Put differently, if we are to create a graph showing how tabs are related to one another, the prototype representation is akin to connecting related tabs in a spoke-hub structure (with anchor tabs as the hubs) as opposed to creating clusters of tabs (i.e., based on item-to-item similarity between all pairs of items).

Summary + Next Steps

This study was a helpful first step towards understanding how people navigate the web using multiple tabs. We plan to further explore how anchor tabs play into people’s browsing behaviors. The findings on timing are making us rethink how other browser features may be better designed (e.g., browsing history). We will also compare self-reported tab usage with those from other observations and data sources, both to verify the accuracy of the survey responses and to generalize the above findings to a larger population.

For our immediate next steps, however, we will examine a few other topics related to content discovery, so that we have a broader understanding of how users access information on the web.