Skip to content
From Fragmented Docs to Visualized Asset Graphs: A Bank's Requirement Management Path
R&D Management 16 min read

From Fragmented Docs to Visualized Asset Graphs: A Bank's Requirement Management Path

China Post Savings Bank shares its practice of building a requirement asset library with visualization graphs, addressing fragmented requirements through entry-level extraction, intelligent tagging, and a three-phase rollout to enable traceable, reusable assets and intelligent search, diagnosis, and closed-loop data integration.

BanTech Think Tank
BanTech Think Tank
BanTech Think Tank

Pain Points in Requirement Asset Management Under Digital Transformation

In the context of continuous financial business innovation and accelerated digital transformation, commercial banks have accumulated massive requirement assets during information system construction. The article identifies three core pain points:

Fragmented and scattered assets: Business requirements are stored as unstructured text, making intrinsic associations between requirements difficult to maintain. Association chains break, preventing a global view of requirement context. Information silos cause departments to repeatedly propose similar requirements due to information asymmetry, wasting resources.

Incomplete precipitation and traceability mechanisms: Requirement changes lack complete traceability records; change history and content cannot be queried. During multi-project parallel execution, cross-project requirement associations are unclear. When core requirements change, impact on related business modules, upstream/downstream processes, and overall system architecture cannot be accurately assessed, leading to business logic conflicts or functional deviations.

Low sharing and insufficient value mining: Without multi-dimensional classification (business domain, system module, requirement type), asset retrieval relies heavily on keyword memory. Precise positioning based on business scenarios or functional modules is impossible, hindering quick location and sharing of requirement assets in complex scenarios, affecting work efficiency and user experience.

Requirement Asset Library Construction: Making Assets Precipitable, Referenceable, and Shareable

Through key stages of requirement asset extraction, construction, and visualization graph display, the bank builds a library that supports precipitation, reference, and sharing, turning tech assets into valuable resources driving business innovation and core competitiveness.

Requirement Asset Extraction: From Document-Level to Entry-Level

The process shifts requirements from document-level to entry-level, where each entry becomes a clear, traceable, manageable minimum unit, achieving a transition from "document-level" to "content-level" management.

Asset collection: Mine and extract valid requirement information from various requirement documents and historical versions, covering all potentially valuable content without omission.

Cleaning and standardization: Refine and structure requirement content, transforming scattered requirements into independent knowledge units such as business requirements, function points, business rules, and business processes. These units have high reusability and strongly support future project development.

Requirement Asset Construction: Intelligent Tagging and Evolution Tracking

Macroscopically, assets are systematically divided by business sectors, functional types, and other dimensions to build a clear catalog for quick location and flexible invocation. Microscopically, requirement documents are decomposed into level-1 modules, level-2 function items, and level-3 function points, with a fine-grained tag system:

Basic attribute tags (e.g., new, optimization, deletion) portray the full panorama of system functional status and evolution trajectory.

Business attribute tags are automatically extracted via large language model semantic analysis, precisely covering core processes, involved products, and other dimensions.

Thus, requirement assets transform from static text into dynamic data assets supporting intelligent retrieval, precise matching, and efficient reuse, laying a solid foundation for intelligent application of functional evolution graphs.

Requirement Asset Visualization Graph Display: Building Functional Panoramic Views

Visualization is key to converting tacit knowledge into explicit assets. The core goal is to obtain a system functional panoramic view, forming a reusable, inheritable requirement asset knowledge system. Combining requirement asset tags and function lists, multi-dimensional views are built, linking relationships among business requirement functional modules to fully grasp the asset landscape. After each full iteration of system function points, the system automatically extracts contained functional modules and specific function items, analyzes interaction relationships between modules and other systems, and dynamically draws system-level functional panoramic views, intuitively presenting internal functional panorama and external interaction modes.

Implementation Path: Low Manual Maintenance Cost, Easy to Implement, Sustainable

The implementation follows the core principle of "low manual maintenance cost, easy to implement, sustainable," using scientific methods and clear paths to turn construction concepts into operational results.

Implementation Methods

Stock asset "slicing" sorting: For existing text-level requirement assets, avoid rebuild-from-scratch. Instead, use lightweight technical means for fine-grained slicing and intelligent parsing, achieving standardized, structured storage of stock requirements, minimizing manual sorting cost, and fully activating stock asset value.

Incremental asset "native structured" construction: For new requirements, guide standardized writing via templates and normalized tools from the source, realizing entry-level and structured writing. Embed standardized writing into the requirement preparation process so each new requirement automatically forms a standardized, normalized asset, achieving routine, convenient precipitation.

Requirement panorama "visualization" display: Integrate all stock and incremental assets to build a globally unified, clearly hierarchical requirement panorama view at system/project dimensions. Present system functional architecture, functional associations, and iteration evolution context through graph-based visualization, facilitating quick retrieval for business personnel and global control for managers.

Functional evolution "automated" tracking: Through model semantic understanding, associate and aggregate identical function points across multiple requirement documents of the same system. Precisely view the complete evolution chain of a function point's iterations and changes, making function-point-level evolution traceable and queryable, effectively avoiding information gaps and silos caused by partial personnel knowledge.

Implementation Path: Three-Phase Strategy

Pilot first, small-scale verification, solid foundation: Select representative systems (rich requirement content, diverse system types, covering both business and technical systems) for small-scale verification. Focus on two tasks: (a) verify model extraction rules against pilot system requirement context and functional evolution history, ensuring key function points are extracted without omission or bias; (b) verify visualization display effectiveness via real business scenarios, collect user feedback, and iteratively optimize model rules and display effects for full rollout.

Full promotion, full coverage, build requirement asset system: Based on pilot experience and optimized model rules, conduct structured extraction for all systems, sort out requirement context and function point evolution history for each, forming a panoramic view covering all bank requirement assets, achieving "one system one view, one function one archive" fine-grained management with traceable sources, clear queries, and traceable context. Simultaneously, use structured requirement writing tools to standardize preparation quality from the source, making incremental requirements achieve structured, entry-level storage at writing stage — "writing is entry, entry is standardized" — continuously precipitating high-quality assets for efficient use and intelligent analysis.

Effect evaluation, continuous research and optimization, deep value mining: Build a scientific quantitative evaluation system and long-term operation optimization mechanism, forming an "evaluate-feedback-optimize" virtuous cycle, upgrading the asset library from "usable" to "easy-to-use" to "willing-to-use." Evaluation uses three dimensions — usage frequency, application breadth, user feedback — with core metrics like retrieval rate, citation rate, likes/dislikes for quantitative analysis, identifying high-frequency reuse and low-efficiency idle assets. Optimization collects qualitative user suggestions on function usage and operation processes, combining user evaluations to insight practicality and usability issues, and targetedly improve system functions and convenience. Dual-driven by quantitative evaluation and qualitative optimization, continuously mine asset value to support digital-intelligent transformation.

Building an Intelligent Requirement Asset System to Enhance Management Efficiency

Solidify the requirement asset library base and empower management with intelligent means, achieving closed-loop control and efficient operation of requirement assets.

Intelligent search empowerment: Leverage data mining and intelligent analysis for automatic extraction, intelligent recommendation, and precise matching of similar requirements, shifting from "passive retrieval" to "proactive empowerment." Assets actively adapt to business scenarios, reducing manual retrieval cost. Build intelligent asset combination generation capability, supporting drag-and-drop assembly of existing knowledge units during requirement writing to quickly form new requirements, significantly improving writing efficiency.

Intelligent diagnosis and warning: Build intelligent diagnosis and pre-warning capabilities. During requirement writing, real-time connect to the asset library for intelligent retrieval and matching, issuing timely warnings for duplicate submissions, high-frequency changes, low-value requirements, effectively reducing resource waste from low-quality or duplicate requirements, continuously improving requirement preparation quality and overall R&D efficiency.

Data link connectivity for closed-loop value: Relying on the asset library, connect data links with project management, R&D implementation, and other links, achieving full-chain association mapping of project initiation scope, business function points, and system function points. Build an integrated data closed-loop from requirement precipitation, reuse implementation, to effect evaluation. Drive requirement asset value flow across the entire R&D process, providing continuous, stable power for high-quality requirement governance and efficient R&D delivery.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

knowledge managementdigital transformationbanking ITrequirement managementrequirement traceabilityasset libraryintelligent taggingvisualization graph
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.