The tool crib as a control point: what the exit read catches, and what it cannot

01 / FIELD NOTE

Keep the decision tied to the operating context.

A tool crib is the one place in a tool’s day where its location is known without anyone having to look for it. Tools are kept there, issued from there and returned there, so a read point at the crib is the closest thing tool tracking has to a checkpoint — a place where a tool’s state changes and where that change can be observed rather than reconstructed. Most of what a tool-tracking project promises rests on that observation, which is why it is worth being precise about what it can and cannot decide.

What the read decides is identity and custody. It can establish that a particular tool left the crib at a particular time, that it came back, and — where the transaction is bound to a person — whose custody it was in while it was away. That is a real and useful record: it answers who held it last, which is the question that turns a missing tool from an unexplained absence into an investigation with a starting point.

What the read does not decide is fitness. A calibration that expired last week, a cracked handle, a missing guard, a cutting edge ground past its limit — none of those are properties a tag carries, and none of them are visible to a reader. A system that treats "the tool left and came back" as equivalent to "the tool is serviceable" has answered a different question from the one the workshop is asking, and the gap between the two is where the failures happen.

The read point itself decides what can be observed, and the common arrangements are not interchangeable. A fixed cabinet reads its own contents with the doors closed, so what it knows is the set of tools inside it at a moment — which makes it the only arrangement that can report an absence with confidence. A portal at the crib doorway reads a crossing rather than a store, so it knows what passed and not what remains. A reader built into a bench or workstation reads the tools present at that station, which is a question about the task rather than about custody. A handheld is the fallback that covers the ground none of the fixed points reach.

Custody and authorisation are separate questions, and only one of them is answered by a read. The record can say who took the tool. It cannot say whether they were entitled to take it, whether they hold the qualification the tool requires, or whether the job they are going to perform permits that tool. Where a site needs those answers, they have to be enforced in the issue workflow — the read supplies the identity the workflow acts on, but it cannot supply the rule.

Calibration and maintenance state belong to whatever system already owns them, and the crib read’s contribution is timing rather than knowledge. A tool leaving the crib is the moment when its calibration status is most useful to the person taking it, because that is when the decision to use it is still reversible. Surfacing that status at the point of issue is a workflow choice, and whether it warns or blocks is a policy decision that depends on what the tool is used for.

Durability and the read environment are the same problem here as anywhere else metal is involved, with one addition: the storage itself is part of the read design. Tools are metal, packed densely, and kept in a steel cabinet or on a steel rack, and both the tools and their container act on the signal. The cabinet is not a neutral box that happens to hold tagged objects; it is a shielded enclosure, and the read inside it is a contained-zone problem rather than an open one.

A missed read at a control point means something different from a missed read in a count, and the difference is what the workflow has to absorb. In a count, a miss is a variance — the next count corrects it and nothing was decided in between. At a control point the miss is the event itself: a tool taken without a record, or returned without one. If the only two outcomes the system can produce are a clean read and silence, then the record and the crib drift apart with nobody watching, and the drift is discovered at the next audit.

The recovery path is therefore part of the design. What happens when a tool does not come back within the interval its job should take, and what happens when the issue record and the return record disagree, are questions the workflow has to answer in advance. The two useful shapes are a reconciliation at a defined point — every tool accounted for before the job closes — and an exception queue with a named owner. Without one of them, a missing tool is indistinguishable from a missed read.

Where a tool left inside finished work is a hazard rather than a loss, the control point changes character. The question stops being where the tool is and becomes whether every tool is accounted for before the work is closed out, which is a reconciliation between what was issued and what was returned, performed at a defined moment. That is the case where the crib read does something a periodic count cannot do at all, because a count performed afterwards is a search rather than a check.

The scope of the installation should follow the environment rather than the ambition. A fixed crib with a cabinet suits a workshop where tools return at the end of a shift and the high-value items are few. A site where the crib travels — a van, a container, a temporary store on a job — has no doorway to instrument and has to work from handheld reads and a defined reconciliation instead. Choosing the arrangement the site can actually operate is what keeps the record current, and a record that is not current is worse than no record, because it will be trusted.