The Research Debt Register: Making Shortcuts Visible

Every research process accumulates debt. A data shortcut is accepted to move faster. A validation gap is deferred because the first result is promising. A monitoring field is planned but not yet implemented. A reference cohort is good enough for the current question but not for every future use. None of these choices are automatically wrong.

They become dangerous when they disappear. A shortcut that remains visible can be managed. A shortcut that becomes part of the background eventually looks like an approved assumption. The team forgets why the choice was made, what it affects, and what evidence would be needed to retire it.

A research debt register is a simple governance artifact. It records the shortcut, the owner, the affected evidence, the consequence of leaving it unresolved, and the condition that would close it. The purpose is not bureaucracy. The purpose is memory.

The register is also a protection against accidental promotion. If a shortcut remains in use long enough, people begin to treat it as part of the framework. A visible debt record interrupts that drift by keeping the original limitation attached to every result that depends on it.

Calling something debt should not imply that the research is invalid. Early exploration often requires imperfect data, narrower universes, simplified costs, or provisional thresholds. A team that refuses every shortcut may never learn enough to ask the next question.

The distinction is whether the shortcut is labeled. A labeled shortcut can be scoped: this result is useful for exploration, not final validation; this comparison excludes low-liquidity cases; this cost model is conservative for one channel and incomplete for another. The label preserves the boundary of the claim.

This distinction also changes review tone. The question is not whether the team was wrong to accept a shortcut. The question is whether the shortcut is still appropriate for the current use. Debt can be rational at one stage and unacceptable at another.

A useful debt item needs at least four fields: assumption, consequence, owner, and retirement condition. The assumption states what was simplified. The consequence states what could be distorted. The owner states who is responsible for resolution. The retirement condition states what evidence closes the item.

Without a consequence, the item is just a note. Without an owner, it is a wish. Without a retirement condition, it can remain open forever. The register should make each shortcut actionable enough that it can be reviewed during normal governance.

A fifth field is often useful: dependency. If no live decision depends on the debt item, it can remain low priority. If monitoring, capacity, exposure, or retirement review depends on it, the same item deserves more attention. Dependency turns a flat list into a prioritized register.

The same debt item can be acceptable in week one and unacceptable in month six. Age matters because other decisions begin to depend on the shortcut. A provisional universe definition may be harmless during exploration. If it becomes the basis for live monitoring, the same shortcut now affects interpretation.

The register should therefore include age and dependency. An old item with no downstream use may remain low priority. A younger item that sits under capacity, exposure, or monitoring decisions may require faster resolution. Priority follows consequence, not just time.

Aging should be visible even when nothing breaks. Many debt items become dangerous precisely because they are quiet. The absence of an incident does not prove the shortcut is safe; it may only show that the system has not yet been tested in the condition where the shortcut matters.

Debt should not stay in the middle indefinitely. Some items are promoted into approved assumptions after evidence shows the shortcut is acceptable for the intended use. Other items are retired because better data, stronger validation, or improved monitoring replaces them.

Promotion needs evidence. It should not happen because the shortcut has survived long enough to feel normal. Retirement also needs a record, because future readers should know which earlier limitations no longer apply. Both paths keep the register from becoming a stale list.

Before relying on a result that contains known shortcuts, the record should answer the following questions.

Research debt is the cost of moving before every question is fully resolved. That cost can be rational. It becomes hazardous when the shortcut loses its label and starts behaving like an approved part of the framework.

Keep a register with assumption, consequence, owner, age, dependency, and retirement condition. Review it when the framework changes state. Promote items only with evidence, and retire them with a record.

The purpose of the register is not to eliminate shortcuts. It is to keep shortcuts from becoming invisible.