An ERP does not make stock reliable by itself. What it does is make discrepancies visible, movement after movement. That is exactly why the annual stocktake, which crams a massive count and bulk corrections into a few days, stops being enough once the system drives purchasing, sales and accounting in real time.
This guide is written for logistics, supply chain and finance leaders in small and mid-sized companies. It covers two distinct moments: the opening inventory at go-live on the new ERP, then cycle counting that keeps stock accurate afterwards. For the valuation side, we point you to our guide to inventory valuation in ERP: FIFO, weighted average and standard cost, rather than duplicate it here.
Why an annual stocktake is no longer enough with an ERP
Accounting rules in most jurisdictions require a company to verify, at least once a year, that its inventory exists and is correctly valued. In France, for example, this is set out in Article L123-12 of the Commercial Code; IFRS (IAS 2) and US GAAP lead to the same practical need. These texts set a minimum. None of them says that a single count at year-end is enough to run a business.
With an ERP, three things make an annual stocktake alone insufficient:
- Book stock drives automated decisions. Requirements planning, replenishment, customer delivery promises and accounting valuation all rely on the quantity in the database. A quantity error therefore flows downstream for months.
- Errors pile up between two counts. A goods receipt keyed wrongly, a unit of measure mixed up, a transfer never recorded: each anomaly stays invisible until the next count, and the year-end correction blends causes nobody can reconstruct any more.
- A full count freezes the business. Stopping movements for several days to count everything is expensive, especially for a company that ships continuously.
Cycle counting answers all three: count a little, all the time, first where an error costs the most, and analyse each variance while its cause is still recent.
Opening inventory at go-live: counting, valuation, movement freeze
At go-live the goal is simple to state and demanding to deliver: the quantities and values loaded into the new ERP must match what is physically on the shelves, and finance must be able to justify the move from the old system to the new one.
Two approaches coexist:
- A full physical count just before go-live: everything is counted and the result becomes the opening stock. It is the safest option, but it requires a movement freeze and a lot of people.
- Carrying over book stock from the legacy system, followed by intensive cycle counting in the first weeks. It reduces downtime, but it assumes the legacy stock is of decent quality and that you accept corrections early in the life of the new system.
The choice depends on how reliable you know the legacy stock to be, on seasonality, and on how much downtime the business can tolerate. In every case, loading balances belongs to the migration workstream: see our ERP data migration methodology for cleansing, validation and reconciliation.
The D-7 to D+2 count plan
A typical calendar, to adapt to your site:
| Day | Actions |
|---|---|
| D-7 | Appoint counting teams and a lead for each zone. Tidy and label locations. Deal with obsolete items and scrap. |
| D-5 | Clean up item masters: duplicates, inactive items, units of measure. Print count sheets. |
| D-3 | Dry run in one zone: labels, scanning terminals, upload into the tool. Train the counters. |
| D-1 | Last receipts and last shipments reconciled with accounting. Announce the freeze. |
| D | Movement freeze. First count, then an independent second count of the high-stakes zones. |
| D+1 | Recount of significant variances. Second-level validation. Load quantities into the ERP. |
| D+2 | Reconcile with accounting, sign the stocktake report, lift the freeze and resume flows. |
The freeze is the most underestimated point. As long as a receipt or a shipment travels between the old and new systems without a clear rule, the count is wrong the moment it is finished. Decide in advance which date is authoritative for each type of movement, and track the documents “in flight”.
Book vs. physical variances: accounting treatment
The gap between book stock and physical stock appears at go-live as it does at every count. Its treatment always follows the same logic:
- Compare counted quantity and database quantity, item by item and location by location.
- Find the cause before correcting: keying error, unrecorded movement, breakage, theft, unit-of-measure error.
- Post the variance: a negative variance (shortage) records a stock loss, a positive one (surplus) a stock increase, valued according to the method in use (FIFO, weighted average, standard cost).
- Document it: the journal entry must point to a count document, an author and a date.
This documentation is what the external auditor or the accountant will look at first. The exact supporting evidence expected depends on your file and on local rules: agree it with them before go-live, not after.
Cycle counting: principle and regulatory framework
Cycle counting means counting subsets of the stock throughout the year on a defined schedule, so that every item is counted at least once over the period and the most sensitive ones far more often.
From a compliance standpoint, what matters is your ability to demonstrate the existence and value of inventory at closing. A perpetual or cycle-count programme can support that, provided it is organised, documented and traceable: count calendar, count sheets or entries, variances found, approvals. Whether such a programme exempts you from a full physical count at year-end is a question to settle with your auditor or accountant, depending on your situation and jurisdiction. Do not decide it alone.
ABC classification and count frequency
ABC classification ranks items by importance, usually their consumption value or their number of movements:
- Class A: few items, high value or high criticality. Frequent counts.
- Class B: intermediate volume and value. Periodic counts.
- Class C: many items, low value. Rare counts.
As an example of the logic (not a standard), a company might count its A items every month, its B items every quarter and its C items once a year. Real frequencies depend on stock value, turnover and available resources.
SAP documents this logic: materials carry a cycle counting indicator in the material master that sorts them into categories (for example A, B, C, D), with a count interval defined per category in configuration, and an ABC analysis program can update the indicator based on consumption or requirements (SAP S/4HANA help, cycle counting method).
Recalculate the classification regularly: an item can move from C to A after a big contract, or the other way round at end of life.
Event-triggered counts
A calendar is not enough. Complement it with event-triggered counts:
- Zero stock in the system while an order calls for it: check physically whether there really is none left.
- Negative stock, a sign that an issue was posted before the matching receipt.
- Receipt of a critical item or a large lot: count right after inspection.
- Repeated variance on an item or a location: bring the next scheduled count forward.
- Dispute with a customer or supplier over a quantity.
These targeted counts pay off the most: they land where the error has just occurred, so the cause can still be identified.
ERP configuration: SAP, Odoo, Dynamics 365 Business Central
All three ERPs below offer a cycle counting mode. The vocabulary differs, the logic does not: a frequency reference, generation of count documents, entry, then approval of variances.
SAP: physical inventory documents and cycle counting
In SAP, the process relies on physical inventory documents: creation (transaction MI01), entry of count results (MI04), recount and variance handling with difference lists and posting (SAP help, physical inventory). For cycle counting, you set the indicator in the material master, define intervals per category in configuration, and a dedicated program generates inventory documents for the items whose count is due (SAP help, cycle counting). Exact transaction codes and screens vary by version (ECC, S/4HANA, Fiori): check them in the documentation for your release.
Odoo: inventory adjustments
In Odoo, the feature goes through inventory adjustments. You select the product lines to count, then “Request a Count” lets you assign a user and a scheduled count date. The counted quantity is entered in the dedicated column, then the adjustment is applied line by line or in bulk. By default, the scheduled date is December 31 of the current year (Odoo 18 documentation, inventory adjustments). The Barcode app lets you enter counts by scanning and run the assigned counting tasks (Odoo documentation, adjustments with barcodes).
Dynamics 365 Business Central: physical inventory journal
Business Central provides the physical inventory journal: you calculate the expected inventory, print the count list, enter actual quantities, then post. For cycle counting, you define counting periods, assign them to items, and the “Calculate Counting Period” action selects the items to count. Microsoft advises keeping the lines calculated at the start rather than recalculating expected stock mid-count, because it may change in the meantime (Microsoft Learn, count, adjust, and reclassify inventory). To split the work between several counters, the documentation also describes a document-based approach (physical inventory orders and recordings).
What to decide before configuring
Whatever the ERP, settle these points up front:
- classification rules and frequencies per class;
- the counting unit and the handling of locations, lots and serial numbers;
- tolerance thresholds and approval levels (next section);
- permissions: who creates, who counts, who approves. The counter must not be the person who posts their own variances, which ties in with our recommendations on ERP access governance.
Barcodes and RF terminals. Scanning reduces keying errors and speeds up counting: the item, location and lot are identified by reading rather than typing. It is the most profitable investment in a cycle counting programme, well ahead of advanced features.
Handling variances: root causes, tolerance thresholds, two-level approval
Cycle counting without rigorous variance handling produces figures, not reliability.
Root cause analysis. Before posting the adjustment, classify each significant variance: counting error (hence the value of a recount), keying error at receipt or shipment, unit-of-measure problem, unrecorded movement, location transfer not entered, breakage or shrinkage, theft, bill-of-materials error for consumed components. Keep this classification up to date: the dominant cause tells you where to act (training, procedure, receiving inspection).
Tolerance thresholds. Define variance thresholds, in quantity and in value, by item class:
- below the threshold: adjustment approved by the warehouse manager;
- above it: mandatory recount, then root cause analysis;
- far above it: approval by finance, or even an investigation.
Thresholds are management choices. A class A item is monitored more strictly than a class C item.
Two-level approval. The principle is to separate the person who counts, the person who approves the quantity adjustment, and the person who approves the accounting impact once a value threshold is crossed. This segregation of duties is configured in the ERP roles and leaves an audit trail: who counted, who approved, when, on which document. It is also what makes the programme defensible in front of an auditor. Our article on ERP audit trails and legal obligations (in French) details these requirements.
KPIs: stock accuracy, variance rate, variance value
Three indicators are enough to get started:
- Inventory record accuracy (IRA): the share of items or locations whose counted quantity matches the database quantity, within the tolerance you set. Define the tolerance before measuring, otherwise two sites cannot be compared.
- Variance rate: the number of lines with a variance divided by the number of lines counted, by ABC class and by zone.
- Variance value: the sum of valued variances, with positives and negatives kept separate (netting them hides problems), expressed against total stock value.
Track these indicators by month, by class and by cause. Their trend matters more than their absolute value: if the accuracy of A items improves after scanning is introduced, you have proof of the programme’s effect.
The link with stockouts is direct. Book stock higher than real stock hides the need to replenish until the customer runs into a stockout. Conversely, book stock that is too low triggers needless purchases and ties up cash. For the replenishment mechanics, see our guide ERP supply chain: WMS, TMS and demand planning, and for multichannel environments the article on ERP for omnichannel retail and real-time inventory.
Common mistakes and checklist
The mistakes we see most often:
- Counting without freezing: movements slip through during the count and make the result unusable.
- Recalculating expected stock mid-count, which skews the comparison.
- Adjusting without finding the cause: the variance disappears from the system but not from the floor.
- Letting the same person count and approve.
- Never recalculating the ABC classification.
- Ignoring locations, lots and serial numbers, then wondering why variances cancel out in the total.
- Forgetting valuation: a quantity adjustment always has a value impact, to be understood before approving.
- Neglecting the unit of measure between purchasing, storage and sales.
Implementation checklist
- Documented choice between a full physical count and carrying over book stock at go-live
- D-7 to D+2 count plan with a freeze rule
- Method validated with the auditor or accountant
- ABC classification and frequencies per class configured
- Event triggers defined (zero stock, negative stock, repeated variance)
- Barcodes or RF terminals on critical locations and items
- Tolerance thresholds and two-level approval configured in roles
- Variance cause classification and monthly review
- IRA, variance rate and variance value indicators published
- Quarterly review of classification and thresholds
Going further
An inventory variance is never just a matter of quantity: it changes the value of stock on the balance sheet. Read our guide to inventory valuation in ERP (FIFO, weighted average, standard cost) to understand the accounting impact of each adjustment, and subscribe to the newsletter from the home page to receive our next practical guides on ERP data reliability.
