Source-linked AI summary
Drillboards: Adaptive Visualization Dashboards for Dynamic Personalization of Visualization Experiences
Sungbok Shin, Inyoup Na, Niklas Elmqvist
TL;DR
Dashboards are often designed for a specific audience and purpose, limiting adaptation when users differ in expertise or goals. Drillboards address this by organizing charts into a merge-based hierarchy that users can drill through, and the authors report favorable outcomes from expert-authored drillboards shown to casual end-users.
Problem
Existing dashboards are designed for specific audiences and purposes, creating a need to accommodate users with different purposes, expertise, and desired effort.
Method
Drillboards use a formal vocabulary and merge rules to build hierarchical dashboards, with DrillVis supporting authoring and reader-side drill-down and roll-up.
Results
Favorable qualitative results came from 3 domain experts authoring drillboards and 10 casual end-users assessing them, with novices swiftly grasping experts’ intentions.
Takeaways & Limitations
Drillboards support personalized interaction with complex datasets by offering multiple levels of detail and abstraction for experts and casual end-users.
Takeaways & Limitations
User confusion arose from visual instability during drill-down, and aggregation policies require modification for different chart types.
Abstract
from arXiv · showhide
We present drillboards, a technique for adaptive visualization dashboards consisting of a hierarchy of coordinated charts that the user can drill down to reach a desired level of detail depending on their expertise, interest, and desired effort. This functionality allows different users to personalize the same dashboard to their specific needs and expertise. The technique is based on a formal vocabulary of chart representations and rules for merging multiple charts of different types and data into single composite representations. The drillboard hierarchy is created by iteratively applying these rules starting from a baseline dashboard, with each consecutive operation yielding a new dashboard with fewer charts and progressively more abstract and simplified views. We also present an authoring tool for building drillboards and show how experts users can use to build up and deliver personalized experiences to a wide audience. Our evaluation asked three domain experts to author drillboards for their own datasets, which we then showed to casual end-users with favorable outcomes.
1 INTRODUCTION
Drillboards make dashboards adaptive by organizing coordinated charts into hierarchies that users can explore at different levels of detail. The authors provide an authoring system and report favorable outcomes from expert-created drillboards evaluated by casual end-users.
- 1 INTRODUCTION: Drillboards organize a baseline dashboard into a hierarchy whose higher levels merge charts into progressively more abstract views, while drilling down restores detail for users’ expertise, purpose, and effort.The hierarchy has a single compact root and a high-detail baseline; predefined levels such as novice, intermediate, and expert are also possible.
- 1 INTRODUCTION: Authors create hierarchies by selecting charts, applying merge operations, and updating titles, explanations, and the hierarchy view.The authoring workflow is paired with a reader workflow in which users expand charts into child charts and collapse them back into parent charts.
- 1 INTRODUCTION: The drillboards algorithm supports exploration through drill-down and roll-up of aggregated visualizations for different levels of expertise.This is presented as a core contribution of the technique.
- 1 INTRODUCTION: DrillVis provides an authoring environment for creating aggregate drillboards and a reader mode for exploring them through drill-down and roll-up interactions.The system supports both constructing the hierarchy and navigating the resulting data views.
- 1 INTRODUCTION: In a qualitative study, 3 domain experts authored drillboards and 10 casual end-users assessed them, with favorable results.The authors report that drillboards effectively supported communication and that novices swiftly grasped experts’ intentions.
2 BACKGROUND
The background positions drillboards as a dashboard-specific form of hierarchical aggregation that combines end-user flexibility with preservation of detail. Unlike automated dashboard customization, drillboards retain an invariant underlying structure while exposing different amounts of detail.
- 2 BACKGROUND: The work builds on research about multiple coordinated views, dashboard design, hierarchical aggregation, and dashboard authoring, while explicitly claiming no contribution to multiple views.These perspectives establish the surrounding literature rather than constituting the drillboards contribution itself.
- 2 BACKGROUND: The technique targets information loss in single-page dashboards by using drill-down and roll-up without requiring more screen space or additional pages.The authors frame this as a way to improve data presentation and comprehension.
- 2 BACKGROUND: Drillboards extend hierarchical aggregation by applying drill-down and roll-up to component charts within dashboards, a combination the authors describe as novel.Prior work explored hierarchical data presentation and drill-down or roll-up, but not this dashboard-specific integration in the authors’ account.
- 2 BACKGROUND: Drillboards address dashboard users’ varying expertise by enabling flexible exploration while preserving a common dashboard structure.This responds to the need for end-user flexibility identified in prior dashboard research.
- 2 BACKGROUND: Unlike automated methods such as QualDash [15], drillboards keep the dashboard core invariant while offering more or less detail according to user needs.The authors associate this invariance with consistency across users and support for users who develop expertise.
3 DESIGN FRAMEWORK: DRILL-DOWN DASH-
Drillboards adapt one dashboard to users with different expertise, tasks, and changing skill levels by organizing charts into a navigable aggregation hierarchy. Authors build this hierarchy from a baseline dashboard using formal merge operations that progressively reduce and abstract chart detail.
- Drillboards provide one dashboard with different detail levels for users whose expertise, tasks, and skill change over time.
- 3.2 Model: A drillboard is a dashboard containing an aggregation hierarchy whose root is a pile recursively containing chart atoms or other piles.
- 3.2 Model: Users drill down by replacing a pile with its children and roll up by replacing charts with their common ancestor, with predefined views optionally capturing specific detail or skill levels.
- 3.2 Model: Authors create the hierarchy by repeatedly replacing two or more charts or groups with a single pile, while the root and original dashboard serve as novice and expert views.
- 3.3 Aggregation Operations: The framework defines six merge operations: labeling, summarizing, archetyping, projecting, juxtaposing, and overlaying charts.
- 3.3 Aggregation Operations: Summarization abstracts similar data with calculations such as averages, sums, or differences; archetyping selects one child, while projection maps dimensions to scatterplot or parallel-coordinate axes.
4 SYSTEM: DRILLBOARDS
DrillVis is a web-based authoring environment for constructing, navigating, and presenting drillboards through coordinated drillboard and tree views. It implements several constrained aggregation operators and supports multiple authored views, while remaining limited to tabular data and selected operations.
- 4.1 Visual Interface: DrillVis provides a drillboard view for the current representation and a tree view for navigating the aggregation hierarchy in author and reader modes.
- 4.1 Visual Interface: Visualization cards represent chart atoms individually and piles as card stacks, with fixed-size, space-filling, or width-resizable layout options.
- 4.2 Aggregation Operators: The implemented operators include arithmetic summarization, archetype selection, two-dimensional projection, and overlaying charts with merged or separate y-axes.
- 4.3 Author Mode: Authors can generate charts from tabular data, select dataset groups through multilevel menus, add annotations, and create or save multiple views for different expertise levels.
- 4.2 Aggregation Operators: Aggregation requires compatible axes or dimensions, and the implementation does not support labeling, juxtaposition, parallel-coordinate projection beyond two charts, or complex network and bespoke visualizations.
- 4.3 Author Mode: The authors emphasize that aggregation policies are not uniquely determined and should remain consistent to avoid difficult-to-interpret results.
5 USER STUDY
The four-phase study examined how three domain experts authored drillboards and how 10 casual end-users explored the resulting personalized hierarchies. End-users completed questions accurately and quickly, while usability feedback was generally positive despite some reported mental effort.
- Study Design: Three domain experts authored drillboards for datasets concerning international funding, workforce shortages, and regional household wealth.The study separated expert authoring from casual end-user evaluation across four phases.
- Usability Study: Participants generally found DrillVis satisfactory and manageable, although mental demand and effort were mixed for some users.PSSUQ scores were mostly between 1 and 2, while NASA-TLX responses indicated that visual stability may have increased cognitive load.
- Domain Experts: Experts created hierarchies differently: P1 emphasized macro-level aggregation, P2 used arithmetic operations to detect issues, and P3 selected representative charts for wealth comparisons.Their authoring choices reflected distinct analytical goals across financial, labor, and regional household datasets.
- Casual End-Users: 9 out of 10 casual end-users completed all three questions within 10 minutes, and all participants answered correctly with supporting reasons.Average completion times were 8m 39s, 7m 56s, and 7m 49s for the three drillboards.
- Casual End-Users: All end-users preferred novice mode for P3’s visualization, while novice and expert modes were mixed for P1 and P2.P3’s novice mode contained three charts, compared with nine in the most abstracted expert mode.
6 DISCUSSION AND LIMITATIONS
The discussion finds that drillboards support expert-driven abstraction and rapid end-user understanding, while identifying visual stability and chart-specific aggregation as important limitations. Future work should improve guidance, automation, chart coverage, and evaluation across alternative uses.
- 6.1 Explaining the Results: Experts used progressive abstraction to match distinct analytical mental models, supporting macro-level trends, issue detection, and wealth-disparity comparisons.The authors interpret these different aggregation strategies as evidence that drillboards can mirror how experts summarize and abstract data.
- 6.1 Explaining the Results: Casual end-users answered correctly in under 10 minutes, and consistent completion times across chart counts suggest that hierarchical aggregates helped them understand data context efficiently.Participants’ preference for novice mode also indicates that author guidance helped them interpret unfamiliar large datasets.
- 6.2 Limitations: Dynamic chart expansion and collapse sometimes disrupted users’ mental maps, causing confusion during drill-down and roll-up navigation.Freezing a drillboard would eliminate this problem, but freezing would remove the central drilling interaction, so layout and transition improvements are needed.
- 6.2 Limitations: Drillboards are not universally applicable across chart types because aggregation operations require visualization-specific policies and some operations are unintuitive for treemaps or node-link diagrams.Each additional chart implementation therefore requires tailored adaptations, which the authors leave for future work.
- 6.3 Future Work: Future work includes improving guidance or automating aggregation, adding unsupported chart types and aggregation mechanisms, and evaluating drillboards as analytical or presentation tools.The authors specifically call for comparative studies with existing data-exploration tools.
7 CONCLUSION
The paper introduces drillboards and DrillVis as an adaptive visualization technique for personalized interaction with complex datasets. Its evaluation reports support for data exploration across different levels of user expertise and abstraction.
- 7 CONCLUSION: Drillboards and DrillVis provide personalized interaction with data through multiple levels of detail and abstraction.The technique is presented as an adaptive visualization approach implemented in the DrillVis system.
- 7 CONCLUSION: The evaluation validated DrillVis for facilitating data exploration among both experts and casual end-users.The conclusion attributes the reported improvement in engagement and understanding to tailoring visual representations to user needs and cognitive styles.