Chapter 04
Evidence saturation and pruning
When more capture stops improving decisions, how to recognise saturation in the time-to-find and tagging metrics, and how to prune the plan without weakening the workflows it supports.
01
When more capture stops improving decisions
There is a point at which adding more capture stops improving decision quality. The records exist, the storage runs, the dashboards populate, and the workflows still cannot use the evidence faster than it arrives. A claims analyst with twenty fixed views has more material than one with six, but the time to assemble a defence packet is the same or longer because the search burden grows with the source count. A safety officer with helmet feeds from every operative has every minute on tape and still relies on the supervisor walk for the events that matter. Evidence saturation is real, and recognising it early is what separates a capture plan that compounds in value from one that compounds in cost. The threshold is not a number of cameras; it is a ratio between the evidence produced and the evidence acted on.
02
Reading the time-to-find metric
Saturation shows up in the time-to-find metric — how long it takes from a workflow trigger to the relevant record being on screen for the person who needs it. Under twenty seconds for a recent event suggests the indexing is healthy. Two minutes suggests the plan has surface but the tagging and route discipline are weak. Ten minutes means the plan is producing evidence the team cannot retrieve in the window the workflow allows. The first response to a slow time-to-find is not to add more capture or more storage; it is to improve indexing, tagging, and route discipline. Better tags on the existing capture usually close the gap. Once tagging is tight, the saturation point becomes visible — adding a new source does not move the time-to-find at all, and the source is a candidate for pruning rather than a candidate for the plan.
03
Pruning without weakening the workflows
Pruning the capture plan means removing sources that no workflow uses, sources that have been superseded by a better source, and sources that the project has stopped tagging consistently. The test is the workflow log: pull the last quarter of evidence retrievals from each workflow and see which sources contributed. A source with zero retrievals across the quarter, on a workflow that ran ten times, is not load-bearing. The pruning conversation tends to be uncomfortable because the source was placed deliberately at the start, but the discipline is to recognise that a capture plan written at mobilisation is a hypothesis, not a contract. Pruning is not about saving money, although it usually does; it is about keeping the plan defensible at audit and useful for the team. A plan with twenty sources that all serve workflows is stronger than one with forty, half of which run for habit.
04
Failure modes when nothing gets pruned
Three failure modes appear when pruning is skipped. Storage costs creep until the project quietly drops retention windows to fit the budget, which weakens the claims position years later. Tagging discipline collapses across all sources because the team cannot keep up, and time-to-find degrades on the sources that matter as well as the ones that do not. The capture plan becomes ceremonial — referenced at handover, never updated — and the team works around it rather than from it. Each of these is recoverable individually, but together they describe a plan that has lost its grip on the workflows. The remedy is the quarterly prune, which is short, structured, and unsentimental about sources that have stopped earning their place.
Practice
01. Pull the last quarter of evidence retrievals from one workflow on your project. Tabulate which capture sources contributed and how often. Identify any source with zero contribution and propose whether to prune, redeploy, or retain.
Look for: A typical exercise reveals that one or two fixed cameras contributed to almost every retrieval, two more contributed occasionally, and one contributed never. The recommended action is usually to redeploy the unused source to a workflow with no dedicated coverage, rather than prune outright, unless the workflow gap has already been closed by another source.
02. Measure your current time-to-find for a recent workflow trigger. Identify whether the bottleneck is capture, indexing, or tagging. Propose the cheapest intervention that would halve the time-to-find.
Look for: On most projects the bottleneck is tagging rather than capture. The cheapest intervention is to introduce a small set of mandatory tags — workflow, package, front, permit number — applied at capture rather than after the fact. The intervention costs almost nothing in tooling and pays back the first time a claims defence runs.
Checkpoint
Are there sources on your capture plan that have not contributed to a named workflow this quarter, and what is stopping you removing them?
Recommended reading