Cycle Counts and Physical Inventory: How Coin Dealers Resolve Variances Before They Grow

The count is a test of the system

A physical count does more than confirm quantity. It tests whether purchasing, receiving, holds, locations, transfers, sales, returns, refining, and adjustments have produced a record that matches reality. A variance is therefore not just a number to correct; it is evidence that some part of the operating process may have failed.

Annual inventory alone lets small errors compound and makes root-cause investigation difficult. After months of movement, staff may no longer remember the transaction, camera evidence may be harder to retrieve, and several offsetting errors can hide one another. Cycle counting divides the inventory into controlled scopes reviewed throughout the year, while a full physical inventory provides a broader periodic confirmation.

The objective is trustworthy location-level, item-level inventory with fewer disruptive shutdowns. That requires disciplined preparation, independent observation, protected count results, and a variance workflow that preserves evidence. Counting faster is not success if the expected quantity is simply copied into the result.

Define the inventory universe

List every place controlled property can exist: vault shelves, safes, retail cases, buying trays, quarantine, statutory hold, customer pickup, shipping, receiving, show cases, employee custody, grading or assay, refining, repair, consignment, supplier return, and in transit. Include customer-owned property and consignment where the business must safeguard or report it, but keep ownership categories distinct.

Each location needs an identifier, physical boundary, authorized custodians, allowed item states, and count method. “Back room” is too broad. A shelf, bin, tray, show case, sealed transfer case, or external custodian should be identifiable enough that a count scope can be frozen and reproduced.

Reconcile the location list with insurance schedules, alarm and camera zones, access roles, third-party confirmations, and accounting. An unrecognized temporary location is a recurring source of apparent loss. If the shop routinely puts intake on a desk overnight, that desk is functioning as a location and needs either control or elimination.

Schedule by risk, not convenience

Classify inventory using value, liquidity, movement frequency, attractiveness, portability, control history, external custody, and transaction complexity. Count the highest-risk groups most often. A small tray of high-value certified coins may deserve more frequent review than a heavier but lower-risk sealed bulk holding.

Use several triggers. Calendar counts test stable inventory. Event counts follow a show, move, employee change, break-in, system outage, major receiving event, or material variance. Zero-balance counts confirm that empty locations are actually empty. Negative quantities, missing locations, dormant items, and repeated adjustments should automatically enter a review queue.

Publish enough of the schedule for staffing and operational preparation, but preserve surprise where policy requires it. Prevent custodians from “cleaning up” unexplained differences before independent review. The program should cover every controlled location over a defined cycle and retain evidence that coverage occurred.

Prepare and freeze the scope

Define the exact locations and item states included, count team, start time, expected finish, supervisor, and movement rule. Finish or pause open transactions cleanly. Clear unrelated paperwork and containers without moving controlled property outside the scope. Verify scales and scanners according to procedure.

Freeze movement in the selected area during the count. If the business cannot stop all movement, use a cut-off timestamp and controlled movement log with sender, receiver, item, source, destination, and time. The count leader reconciles every in-window movement before results are compared. Unrecorded movement can make both the physical count and system balance individually correct at different moments while appearing to disagree.

Generate the count assignment without exposing expected quantities where a blind count is intended. Preserve a snapshot of the ledger at cut-off, including item identifiers, location, status, ownership, quantity, and relevant unit. Do not let the same user alter the snapshot after counting begins.

Count what is physically present

Work systematically through the location. Identify each item or homogeneous lot, record observed quantity, verify serial or unique identifier where present, note condition or seal issues, and scan the location. Use approved units—pieces, tubes, troy ounces, grams, face value, or lots—without casual conversion. A sealed container can be counted as one controlled unit only if policy and prior verification support that treatment.

Do not assume an item belongs in the location simply because the system says so. The first count starts from the shelf outward. Record unexpected items as found property within the count, not as a reason to search for a matching expected line and force it into place. Likewise, leave a missing expected item unresolved until the comparison stage.

For bulk material, define whether the control is piece count, gross and tare weight, fine weight, sealed package count, or another method. Record the method and device. Sampling should be described honestly and used only where approved; a sample is not a full physical count.

Use blind recounts for differences

When the first observation differs from the ledger, assign a second qualified counter who does not see the expected balance or first result. Reset and recount the exact scope. For unique items, verify identifiers; for weight-based inventory, recheck unit, tare, and scale; for packages, confirm seals and inner-container rules.

Independence matters. Two employees counting together and agreeing on one number is a dual observation, not two independent counts. Both can be useful, but the procedure should name which control is required. High-value locations may require one counter and one verifier even when the first result matches.

Record both results, people, times, methods, and evidence. Never overwrite the first count with the second. If they differ from each other, contain the scope and involve the count supervisor before movement resumes. A repeated recount until someone produces the expected number is not investigation.

Classify the variance before adjusting

Common variance types include missing item, unexpected item, wrong location, wrong quantity, serial mismatch, unit-of-measure error, ownership mismatch, duplicate record, unposted receipt, unposted sale or return, transfer timing, refining or transformation gap, damaged seal, and catalog or master-data error. Classification guides the investigation and corrective action.

Calculate the quantity and financial significance using a documented reference. A one-piece difference can be more important than a large bulk count difference. Apply escalation based on value, product risk, suspected custody break, recurring pattern, and legal or insurer requirements—not only quantity.

Contain affected inventory. Prevent sale, transfer, payment, refining, or location change if continued activity would compromise the evidence. Assign an investigator and due time. Notify only the roles required by policy; protect the integrity and confidentiality of the review.

Investigate the transaction chain

Begin with the last verified event and move forward. Review purchase intake, receipt, testing, label creation, hold release, location moves, customer reservations, sales, returns, transfers, shipments, show activity, grading, refining, and prior counts. Compare timestamps, users, item identifiers, camera references, scale records, carrier records, payment or invoice links, and physical neighboring locations.

Search for symmetrical errors. A shortage in one bin and overage in another may be an unposted move. A missing one-ounce bar and an extra ten-ounce bar may indicate a unit or product selection error rather than equal financial impact. A sold item still on the shelf may be a fulfillment failure, while an empty package may be more serious.

Keep facts, hypotheses, and conclusions separate. Record each check and result so another reviewer does not repeat work or inherit an unsupported assumption. Preserve evidence before retention windows or daily system processes remove it. Escalate safety, theft, fraud, or legal concerns through the dealer’s incident procedure.

Correct with an attributable adjustment

If the variance is a timing or transaction error, complete or reverse the missing event using the appropriate workflow. If it is a wrong location, post the controlled move. If quantity truly requires adjustment, use an inventory adjustment with reason, evidence, financial value, initiator, approver, and reference to the count case. Do not edit the original purchase or prior count until history matches today.

Require approval based on value and cause. The counter should not approve their own material adjustment. A repeated small difference can deserve higher review than one isolated event. Closed accounting periods may require a current-period correction under the accountant’s policy rather than a backdated change.

After posting, verify the location physically and in the system, release any hold, and confirm the count case reconciles to zero open units or clearly documented unresolved loss. Closing the adjustment is not the same as closing the root cause; assign the process correction separately.

Find and fix the root cause

Use categories that lead to action: inadequate location design, label failure, training gap, confusing product setup, wrong unit, shared account, delayed posting, unauthorized movement, weak handoff, transfer design, return disposition, show process, integration failure, equipment problem, policy gap, or unresolved incident. Avoid “human error” as the final answer; ask what allowed the error to become inventory.

Choose a corrective action, owner, due date, and verification. Examples include relabeling bins, changing scanner prompts, separating similar SKUs, requiring receiver acceptance, adding a second-person threshold, revising a unit conversion, training a role, repairing an integration, or shortening the count cycle. Verify the change by reviewing later transactions and counts.

Look across cases. Several locations with the same product variance suggest master data or packaging. Variances after shows suggest dispatch or reconciliation. Differences near shift changes suggest handoffs. The count program is a feedback system for operations, not only a ledger-cleaning exercise.

Plan the full physical inventory

A periodic full count still has value for financial reporting, insurance, governance, and independent assurance. Plan far enough ahead to control sales, purchases, shipping, transfers, shows, and third-party custody. Define cut-off, locations, teams, independence, count sheets, system access, evidence, recount thresholds, external observers, and reopening criteria.

Reconcile goods in transit, customer-owned property, consignment, grading, refining, supplier returns, and other external custody before the count. Obtain confirmations where physical observation is not possible. Avoid double counting property dispatched from one location but not received at another.

Use lessons from cycle counts to reduce year-end disruption. Clean locations, resolve negative balances, close stale transfers, and investigate old exceptions before the freeze. The full count should validate a controlled year, not attempt to reconstruct it in one weekend.

The inventory-accuracy dashboard

Track locations due and overdue, count coverage by risk class, first-count accuracy, blind-recount agreement, variance quantity and value, time to resolution, adjustments, root cause, repeat rate, unlocated inventory, negative balances, and property outside controlled locations. Show both absolute and relative results; a tiny percentage can still contain one material item.

Measure process health around the count. How many movements occurred during freeze? How often was expected quantity visible to the counter? Which counts lacked evidence or independence? How many adjustments were posted without completed investigation? Accuracy claims are only credible if the procedure was credible.

Share trends with the operating team while protecting sensitive incident detail. Celebrate reductions in repeat causes and faster controlled resolution, not merely fewer reported variances. A team that surfaces small issues early is strengthening the system.

A 30-day cycle-count launch

1. Week 1: Map every physical and external location, ownership state, custodian, and count method.
2. Week 2: Classify risk, set count frequency, create blind assignments, and define freeze and movement controls.
3. Week 3: Pilot counts in one high-risk and one high-movement area; investigate every difference fully.
4. Week 4: Correct root causes, publish the recurring calendar, and launch variance and coverage dashboards.

Inventory accuracy is not the absence of reported differences. It is the ability to detect, explain, correct, and prevent them while the evidence is still fresh.

Count against one location-aware ledger

bullionOS connects item identity, custody, transfers, locations, invoices, and adjustments so physical counts can verify the operating record instead of rebuilding it.

Connect this workflow in bullionOS

Bring this operating process into one connected dealer record. Explore Coin Inventory Software for Dealers, or visit the bullionOS dealer operations resource hub for related guides.