Building the Scaffold Before the Science

Posted: 8 May 2026 | PROM05 Week 2

One of the quieter lessons of research project management is that the tools you choose to organise the work shape the work itself. This week's focus on visualisation and scheduling has prompted me to think carefully about how I intend to manage the practical development of this project, and to commit to a specific approach before the implementation phase begins in earnest.

The project I am managing is a comparative evaluation of uplift modelling evaluation metrics, specifically the Qini coefficient and the Area Under the Uplift Curve, using a modular Python framework built around the Criteo Uplift dataset. The practical component is substantial: eight discrete framework modules, three uplift estimators, a bootstrap subsampling procedure across fifty iterations, and a rank correlation analysis that will form the evidential core of the dissertation. Managing this without a clear visual structure would be a significant risk.

The Gantt Chart as Planning Discipline

My primary scheduling tool will be a Gantt chart, built in spreadsheet form, spanning the full project timeline from now through to the PROM06 dissertation submission. Phelps, Fisher and Ellis (2007) describe the Gantt chart as the standard instrument for documenting research timelines, and their key observation is that its value lies not in the sophistication of the software used to produce it but in the discipline the construction process demands. Building a Gantt chart forces explicit decisions about every task: what it is, how long it will take, what it depends on, and whether the cumulative duration fits within the available time. That confrontation with arithmetic is where unrealistic plans are caught before they become missed deadlines.

Wingate (2025) adds an important technical dimension to this: a schedule is only partially useful until it is resource-loaded, meaning each task has a named responsible party and an allocated time budget. For a solo project, this translates to explicit hour allocations per module rather than vague duration bands. Assigning eight hours to the data ingestion module and twelve hours to the bootstrap sampling module, for example, produces a materially different schedule than simply noting that both should be done in week three. The Gantt chart I am building will operate at this level of specificity.

The Gantt chart also makes the critical path visible: the sequence of tasks where any delay produces a delay in the overall schedule. For this project, the critical path runs through data preprocessing, S-Learner and T-Learner implementation, and the bootstrap subsampling procedure, because each of those feeds directly into the metric computation and rank correlation analysis that produces the primary results. Tasks on that path will receive priority resource allocation and the closest monitoring.

The Kanban Board as Operational Tool

The Gantt chart handles strategic scheduling. For the implementation phase itself, I will use a Kanban board as the operational task management layer. The standard three-column structure, To Do, In Progress, and Completed, provides a real-time view of workflow state that the Gantt chart does not attempt to capture. Where the Gantt chart asks "are we on schedule?", the Kanban board asks "what is actually happening right now?"

The motivational dimension of the Kanban board is also worth naming directly. Phelps, Fisher and Ellis (2007) note that setting tangible, observable goals is important for researcher morale, and watching a Completed column grow as modules are built and tested is a concrete, visible record of progress that sustains momentum across a project with a distant single deadline. For a solo researcher without the social accountability of a team environment, this self-generated motivational structure is not trivial.

I will use GitHub Projects to host the Kanban board, which has the additional advantage of keeping the task management directly adjacent to the version-controlled codebase. Each card on the board will correspond to a discrete, independently testable module or analytical task, meaning completion is verifiable rather than subjective.

Pre-Registration as Risk Mitigation

A third management instrument that does not fit neatly into the visualisation framework but belongs in this discussion is the OSF pre-registration. Before any data analysis begins, the full analytical plan, including model selection rationale, bootstrap parameters, metric computation approach, and the Kendall's tau threshold of 0.7 for the inconsistency test, will be committed to OSF as a time-stamped public record. Wingate (2025) identifies confirmation bias as a project risk in R&D contexts, noting that a vested intellectual interest in a particular result creates pressure to make analytical decisions that favour that result. Pre-registration removes that pressure by making any deviation from the plan explicitly visible and requiring documented justification.

Taken together, the Gantt chart for strategic scheduling, the Kanban board for operational execution, and OSF pre-registration for analytical integrity form the management scaffold within which the science will be conducted. The scaffold does not do the research. But without it, the research has no structure to stand in.

References

Phelps, R., Fisher, K. and Ellis, A.H. (2007) Organizing and Managing Your Research: A Practical Guide for Postgraduates. London: SAGE Publications. Available at: https://ebookcentral.proquest.com/lib/sunderland/detail.action?docID=354865&pq-origsite=primo (Accessed: 3 May 2026).

Wingate, L.M. (2025) Project Management for Research and Development: Guiding Innovation for Positive R&D Outcomes. 2nd edn. Auerbach Publications. Available at: https://learning.oreilly.com/library/view/project-management-for/9781040326671/ (Accessed: 1 May 2026).