The Tool Inventory Count SOP

Purpose

To confirm, on a rolling schedule, that every tracked tool physically exists and is where the register says it is. This is a count of existence and location only. It deliberately does not assess condition, value, or compliance, because a count that tries to do all four takes a day, gets scheduled once a year, and therefore tells you about a loss eleven months after it happened. The annual tool audit owns condition and valuation; this procedure runs between audits and its whole job is to shorten the time between a tool going missing and somebody knowing.

Scope

Applies to every item on the tool register: shop crib, every service vehicle, and any job box staged on an active site. Applies to all staff who hold tools.

Out of scope: consumables and stock parts, which are counted through the parts process; condition assessment and retirement decisions, which belong to the condition inspection SOP; and any item not on the register, which by definition cannot be counted. If untracked items keep turning up in counts, the register scope is wrong and that is fixed in the register, not here.

Roles and responsibilities

The independence rule in row two is the one that makes this work. A tech counting his own truck knows what should be there and will see it whether it is there or not, which is not dishonesty, it is how memory works under time pressure. Cross-counting is not an accusation and it should be stated that way when the rotation is introduced.

Definitions

• Zone: one countable physical location. One vehicle is one zone. The crib is one zone. Each staged job box is one zone.
• Blind count: the counter lists what is physically present without seeing the register's expectation for that zone. Reconciliation happens afterward, on paper.
• Variance: any difference between the blind count and the register, in either direction. An item present that the register puts elsewhere is a variance too.
• Unlocated: a register item not found during the count and not yet explained.
• Loss: an unlocated item still unexplained at the end of the resolution window.

Procedure

1. Run the rotation so every zone is counted on a fixed cadence

Count one zone per week, rotating in a fixed order. A shop with four vehicles and a crib has five zones and therefore a five-week cycle, which counts every zone about ten times a year. Do not count everything in one week; the whole design goal is a short interval between a loss and its detection, and a rotation delivers that without ever stopping the shop.

Add a staged job box to the rotation as its own zone for as long as that site is active, counted weekly regardless of where the rotation sits, because a site box is the highest-exposure zone you own.

2. Count when the zone is at rest

A vehicle is counted either before load-out in the morning or after return at the end of the day, never mid-shift. A truck counted at noon is missing whatever is in the tech's hands, in the customer's basement, and on the roof, and every one of those looks identical to a loss. Counting at rest removes an entire class of false variance that otherwise makes the whole procedure feel unreliable in its first month.

3. Count blind, then reconcile

The counter walks the zone and writes down every tracked item present by its ID, without the register list in front of them. Only after the list is complete does the register keeper lay the two side by side.

Counting against a printed list produces confirmation, not a count. A person holding a list reads "item 14, pipe threader" and their eye finds the threader-shaped object without ever checking the ID, which is exactly how an item that was swapped between trucks gets counted in the wrong place for a year.

4. Handle any hazard found during the count immediately

Condition is not this procedure's job, with one exception that overrides everything above. If the count turns up an item with a defect that makes it dangerous, deal with it before you continue counting:

• A ladder with a cracked, bent, or split rail, a loose rivet, or a rung that rotates comes off the vehicle at that moment, gets tagged, and goes to quarantine, not back on the shelf.
• A test lead, cord, or insulated handle with a cut, nick, or exposed conductor comes out of the vehicle immediately and does not go back into service pending replacement.
• A sling, strap, or lifting item with broken wires, cut webbing, pulled stitching, or an illegible rating tag is removed from service on the spot, which is what 29 CFR 1910.184(d) requires of a damaged sling found on inspection.

Then record it as removed and continue. Never note a hazard for later and leave the item in the zone, because the next person to open that door is going to use it.

5. Classify every variance before anyone calls anything lost

Each variance falls into one of four classes, and three of them are not losses:

1. Recorded issue: the item is out with a person and the issue log says so. Not a variance at all once checked. Check the issue log before anything else.
2. Unrecorded transfer: the item is in a different zone. Confirmed when it turns up in that zone's count or the holder confirms it. Fix the register, and note that the transfer step was skipped.
3. Unlocated: nobody can say where it is yet. Enters the resolution window.
4. Present but unexpected: found in this zone, registered to another. Same fix as class 2, from the other direction.

Classes 2 and 4 are the most common finding in a shop's first few rotations and they are a process signal, not a loss signal: they mean tools move between trucks without the handoff being recorded, which the issue and return SOP addresses.

6. Work the resolution window, then close

An unlocated item stays open for 5 working days. During that window it is checked against the next zone counts, the holder is asked directly, and the last three jobs it was used on are checked. At the end of 5 working days the manager closes it as located or as a loss. Nobody else closes it, and it is never left open indefinitely, because a permanently open variance list is the same as no list.

7. Apply the escalation rule

Per zone, per count, escalate when either of these is true: (a) unresolved items at the end of the window exceed 2% of that zone's tracked item count AND number at least 2 items, or (b) any single tier-A item is unlocated, where tier A means anything that produces a reading trusted for a safety decision, or anything that bears a load or a person.

The AND inside branch (a) is deliberate. On a zone of 40 to 60 items, one item is already above 2%, so a bare percentage gate escalates on every single loss in a small zone and the escalation stops meaning anything within two months. Branch (b) has no percentage at all, because one missing voltage tester or one missing harness is a real problem at any zone size.

Escalation is: a full recount of that zone within 2 working days, a review of that zone's issue and return records for the period since the previous count, and a conversation with the holder about process rather than blame.

8. Record the rotation results and read them as a series

Log per count: zone, date, counter, items expected, items counted, variances by class, and losses closed. One line per count. The value is entirely in the series. A single count tells you about a week; ten counts tell you which zone leaks, whether unrecorded transfers are falling as the handoff habit takes hold, and whether losses cluster around a site type or a job type.

Worked example: a five-zone shop's first rotation

The setup. Four vehicles plus a crib, five zones, 214 tracked items in total.

Zone 3, week three of the rotation. The register expects 47 tracked items in that zone. The counter, a tech from another truck, counted at end of day with the vehicle back at the shop.

Blind count: 44 tracked items present. Reconciled against the register: 42 of those 44 matched the zone-3 list, and 2 were items the register places in other zones. So 5 register items were not found, and 2 unexpected items were present.

Classifying the 5.

• 2 were on the issue log, out with a tech on a multi-day job. Not variances.
• 1 was in the crib awaiting repair, recorded. Not a variance.
• 1 turned up 3 weeks later in truck 1's count, which made it an unrecorded transfer, and the register was corrected.
• 1 was never found and closed as a loss after the 5 working days.

The 2 unexpected items were both unrecorded transfers in the other direction, so between them and the one above, three transfers had happened that month with no handoff recorded. That is the actionable finding from this count, and it is a process finding, not a loss finding.

Running the escalation rule on the real numbers. One unresolved item out of 47 tracked items is 2.13%, which is above the 2% figure in branch (a). But branch (a) also requires at least 2 items, and there was 1, so branch (a) is not met. The lost item was a hand tool, not a reading instrument and not load bearing, so branch (b) is not met either. No escalation. It was closed as a loss and logged.

Now the counterfactual that shows the other branch working. Had the unlocated item been the truck's clamp meter, branch (b) fires on its own, at 1 item and 2.13%, and the zone gets a full recount within 2 working days plus an issue-log review. The percentage never enters the decision, which is the entire reason branch (b) exists.

The rotation total. Across all five zones in the first full five-week rotation: 214 tracked items, 3 items closed as losses, and 9 unrecorded transfers corrected.

Reading that number carefully. Three losses out of 214 tracked items is about 1.4% per rotation, and a rotation is five weeks, not a year. If that rate held for a full year of roughly ten rotations it would be on the order of 14% of the tracked fleet, which is the number worth reacting to. Stating it as "we lost 1.4%" without the rotation attached is the error that makes a serious leak sound like a rounding difference.

The honest caveat runs the other way too: a first rotation always sweeps up drift that accumulated over months or years before anyone was counting, so the first cycle overstates the ongoing rate. The number to steer by is the second and third rotation, and the shop should expect the count to fall before it settles.

Records and retention

Keep the count log for at least three years, because the series is the product and three years is enough to show seasonality and to support an insurance claim. Keep the closed-loss entries permanently against the item ID in the register; an item closed as lost that later reappears must be reopened rather than quietly re-added, or the loss history stops being trustworthy.

References

• OSHA 29 CFR 1910.184(d), slings inspected each day before use, with damaged slings removed from service immediately
• OSHA 29 CFR 1910.242(a), employer responsibility to maintain hand and portable powered tools in safe condition
• See related: The Annual Tool Audit SOP; The Tool Issue and Return SOP; How to Build a Tool Inventory That Stays Current; How to Track Tools Across Multiple Trucks