As the complexity of data analysis increases, even well-designed data interfaces must guide experts in transforming their theoretical knowledge into actual features supported by the tool. This challenge is even greater for casual users who are increasingly turning to data analysis to solve everyday problems. To address this challenge, we propose data-driven, contextual, in-situ help features that can be implemented in visual data interfaces. We introduce five modes of help-seeking: (1) contextual help on selected interface elements, (2) topic listing, (3) overview, (4) guided tour, and (5) notifications. The difference between our work and general user interface help systems is that data visualization provide a unique environment for embedding context-dependent data inside on-screen messaging. We demonstrate the usefulness of such contextual help through two case studies of two visual data interfaces: Keshif and POD-Vis. We implemented and evaluated the help modes with two sets of participants, and found that directly selecting user interface elements was the most useful.

Visualization of ranked lists is a common occurrence, but many in-the-wild solutions fly in the face of vision science and visualization wisdom. For example, treemaps and bubble charts are commonly used for this purpose, despite the fact that the data is not hierarchical and that length is easier to perceive than area. Furthermore, several new visual representations have recently been suggested in this area, including wrapped bars, packed bars, piled bars, and Zvinca plots. To quantify the differences and trade-offs for these ranked-list visualizations, we here report on a crowdsourced graphical perception study involving six such visual representations, including the ubiquitous scrolled barchart, in three tasks: ranking (assessing a single item), comparison (two items), and average (assessing global distribution). Results show that wrapped bars may be the best choice for visualizing ranked lists, and that treemaps are surprisingly accurate despite the use of area rather than length to represent value.

Effective data analysis ideally requires the analyst to have high expertise as well as high knowledge of the data. Even with such familiarity, manually pursuing all potential hypotheses and exploring all possible views is impractical. We present DataSite, a proactive visual analytics system where the burden of selecting and executing appropriate computations is shared by an automatic server-side computation engine. Salient features identified by these automatic background processes are surfaced as notifications in a feed timeline. DataSite effectively turns data analysis into a conversation between analyst and computer, thereby reducing the cognitive load and domain knowledge requirements. We validate the system with a user study comparing it to a recent visualization recommendation system, yielding significant improvement, particularly for complex analyses that existing analytics systems do not support well.

General purpose graphical interfaces for data exploration are typically based on manual visualization and interaction specifications. While designing manual specification can be very expressive, it demands high efforts to make effective decisions, therefore reducing exploratory speed. Instead, principled automated designs can increase exploratory speed, decrease learning efforts, help avoid ineffective decisions, and therefore better support data analytics novices. Towards these goals, we present Keshif, a new systematic design for tabular data exploration. To summarize a given dataset, Keshif aggregates records by value within attribute summaries, and visualizes aggregate characteristics using a consistent design based on data types. To reveal data distribution details, Keshif features three complementary linked selections: highlighting, filtering, and comparison. Keshif further increases expressiveness through aggregate metrics, absolute/part-of scale modes, calculated attributes, and saved selections, all working in synchrony. Its automated design approach also simplifies authoring of dashboards composed of summaries and individual records from raw data using fluid interaction. We show examples selected from 160+ datasets from diverse domains. Our study with novices shows that after exploring raw data for 15 minutes, our participants reached close to 30 data insights on average, comparable to other studies with skilled users using more complex tools.

A standard (single-column) bar chart can effectively visualize a sorted list of numeric records. However, the chart height limits the number of visible records. To show more records, the bars could be made thinner (which could hinder identifying records individually), and scrolling requires interaction to see the overview. Treemaps have been used in practice in non-hierarchical settings for dense visualization of numeric data. Alternatively, we consider wrapped bars, a multi-column bar chart that uses length instead of area to encode numeric values. We compare treemaps and wrapped bars based on their design characteristics, and graphical perception performance for comparison, ranking, and overview tasks using crowdsourced experiments. Our analysis found that wrapped bars perceptually outperform treemaps in all three tasks for dense visualization of non-hierarchical, sorted numeric data.

Datasets commonly include multi-value (set-typed) attributes that describe set memberships over elements, such as genres per movie or courses taken per student. Set-typed attributes describe rich relations across elements, sets, and the set intersections. Increasing the number of sets results in a combinatorial growth of relations and creates scalability challenges. Exploratory tasks (e.g. selection, comparison) have commonly been designed in separation for set-typed attributes, which reduces interface consistency. To improve on scalability and to support rich, contextual exploration of set-typed data, we present AggreSet. AggreSet creates aggregations for each data dimension: sets, set-degrees, set-pair intersections, and other attributes. It visualizes the element count per aggregate using a matrix plot for set-pair intersections, and histograms for set lists, set-degrees and other attributes. Its non-overlapping visual design is scalable to numerous and large sets. AggreSet supports selection, filtering, and comparison as core exploratory tasks. It allows analysis of set relations inluding subsets, disjoint sets and set intersection strength, and also features perceptual set ordering for detecting patterns in set matrices. Its interaction is designed for rich and rapid data exploration. We demonstrate results on a wide range of datasets from different domains with varying characteristics, and report on expert reviews and a case study using student enrollment and degree data with assistant deans at a major public university.