Y. Xing, C. Dondi, R. Borgo, and A. Abdul-Rahman. Visualizing historical book trade fata: An iterative design study with close collaboration with domain experts. IEEE Transactions on Visualization and Computer Graphics, 30(1):540-550, 2024. DOI.
This Case Study Report was authored in 2026 by Yiwen Xing, University of Oxford, UK
This case study is based on the development of the BookTracker [1], a visual analytics system designed to support historians in studying the provenance and circulation of early printed books. The work was conducted through a close collaboration between visualization researchers and historians specializing in the history of the book trade. The data used in this work is derived from the Material Evidence in Incunabula (MEI) database [2], which records the provenance evidence and institutional holdings of books printed in Europe during the early age of print.
The entire design process can be summarized as seven design iterations following repeated design–implement–feedback loops, as illustrated in Fig. 1. From the overall design and collaboration process, we extracted key domain requirements as follows:
search and visualize the provenance history of one or multiple books;
visualize the provenance paths on the geo-map;
present the provenance history through animation;
visualize the detailed statistics of the provenance records;
support the detection of similar patterns among book records.
Fig. 1 The design iterations of BookTracker.
These requirements were not identified in one go. Instead, throughout the collaboration, we observed that domain experts do not always have clearly defined analytical problems in mind, nor are they always familiar with articulating their needs in computational terms [1]. Consequently, the design process evolved largely through discussions during feedback meetings. Importantly, the visualization tasks addressed in each iteration were not always refinements of previous designs; rather, new analytical directions often emerged, leading to branching problem explorations.
Here, we revisit our iterative design process retrospectively using the SCRS-SAVI framework. The Workflow for Optimizing Workflow (WF4OWF) is illustrated in Fig. 2. Considering the baseline approach used by historians to search provenance records, the interface is heavily text-based, with no visualization, no algorithmic assistance, limited interaction, and minimal statistical support. Users must select each book entry from the catalogue individually and open a new tab to view its provenance records. The key symptom and cause here is that it is cognitively costly for users to memorize and compare the provenance histories of multiple books simultaneously, while the search interaction itself is tedious and inefficient. The information compression in the current interface is low.
To optimize this workflow of searching and memorizing provenance information, the first remedy is to introduce geo-based visualizations to represent provenance data as an alternative to textual records. In this representation, the provenance locations of a book are compressed into points (cities) on a map, while the movement sequence is represented as line segments connecting these points. This allows users to view the provenance histories of multiple books simultaneously within a single visual space. However, a side effect of this approach is that when the number of books becomes large, the visualization may become cluttered.
A second remedy is to introduce an overview visualization that supports pattern identification across the entire dataset and enables users to select subsets of books more efficiently. Nevertheless, this approach may again result in visual clutter when too many provenance paths are displayed simultaneously (side effect). To address this, two alternative remedies can be considered. 1) The Edge-Bundling algorithm can be introduced to reduce visual clutter by grouping similar paths together. While this reduces display complexity, a potential side effect is that the resulting curved lines may introduce information distortion, making it harder for users to determine which paths correspond to individual books. 2) A heatmap representation can replace the paths on the geographical map by displaying only the frequency of transfers between locations. This approach mitigates map clutter and compresses the provenance information, but introduces a different side effect: the detailed provenance information of individual books becomes distorted.
We observe that the improved visual analytics workflow introduces a fundamental trade-off between close- and distant- reading of provenance information (symptom). In other words, this reflects a trade-off between higher Alphabet Compression (AC) and the resulting increase in Potential Distortion (PD) and cognitive cost (cause). To address this issue, additional interactions can be introduced to bridge summary visualizations and detailed views (remedy), allowing users to move seamlessly between overview and detail. However, the side effect is that users may need to perform multiple interaction operations, increasing the human labor cost.
A further remedy is to introduce animation, which can automate sequences of interactions and allow users to observe the system revealing the provenance dynamically. Interestingly, these analysis echoing the requirements from our real-world collaboration: the domain experts explicitly requested animation as an engaging way to present their data. But here it can also serve as a means of reducing the human effort required to manually perform multiple interaction steps.
Fig.2 The Workflow for Optimizing Workflow (WF4OWF) of the case of visualizing book trade data.
[1] Y. Xing, C. Dondi, R. Borgo, and A. Abdul-Rahman. Visualizing historical book trade fata: An iterative design study with close collaboration with domain experts. IEEE Transactions on Visualization and Computer Graphics, 30(1):540-550, 2024. DOI.
[2] C. Dondi. “15cBooktrade”: An evidence-based assessment and visualization of the distribution, sale and reception of printed books in the renaissance. Gazette du livre médiéval, 60(1):83-101, 2013. DOI.