A standard landed cost is a plan. An actual landed cost is a fact. The variance is the difference, and it is not noise to be averaged away: it decomposes into named causes, a freight rate that moved, a duty rate applied differently at liquidation, a cube assumption that was never true, an allocation base that drifted from volume. Importers who review variance monthly fix their standards while the fix is cheap. Importers who do not are pricing next year from last year's imagination.
A single variance percentage per product teaches nothing, because opposite errors cancel inside it. Split the variance by component: product price, freight, duty, insurance, brokerage and fees, allocation effects. A product can show a calm total while its freight doubled and its duty halved on a reclassification, two facts that demand opposite responses. Component variance turns a shrug into an agenda, each line with an owner and a question attached.
Rate variance and usage variance deserve separate names too. Paying more per kilo is a rate problem, negotiated or market driven. Shipping more kilos per unit than the standard assumed, because the carton spec changed in production and nobody told the cost model, is a usage problem, and it will recur on every shipment until the standard or the carton is corrected.
An actual cost can chase corrections forever: the freight invoice, the amended invoice, the post summary correction, the liquidation two years later. Set a closing rule per period: actuals are struck from documents received by a set date, later corrections flow into the next period's variance as named items, and liquidation differences on old entries get their own line rather than being smeared across current products. Without a rule, every month's actuals are permanently provisional and the variance review dissolves into an argument about timing.
Materiality belongs in the rule as well. Correct what moves a decision, note what does not, and never let the pursuit of a perfect actual consume more than the variance it explains. The review exists to improve pricing and sourcing, not to produce archaeology.
A variance review that changes nothing is a ceremony. Its outputs are decisions: a standard revised because the market moved, a carton respecified because usage variance exposed it, a supplier discussion opened because their price variance is a pattern, a duty position revisited because actual applied rates keep differing from the model. Track the decisions and whether the next quarter's variance narrowed. That, and not the spreadsheet, is the control working.
Keep the standard's history. A cost model that silently overwrites its standards cannot show its own learning, and the third time a standard is revised upward for the same cause is a story someone should have to tell out loud.
Related reading: Can you claim duty drawback on returned and resold imports?, Can you claim duty drawback on destroyed or unsold inventory?.
On a cycle that matches your volatility, quarterly for most importers, and immediately when a structural fact changes: a tariff action, a freight contract renewal, a supplier price change. Standards should be stable enough to price against and honest enough to trust, and those two needs are managed by scheduled review, not by editing whenever a number looks unflattering.
Both, for different questions. Per shipment variance finds operational causes, a bad consolidation, a misrated invoice. Per unit variance over a period finds pricing causes, the standard the catalogue is actually being sold against. Reviewing only one view systematically misses the other view's problems.
There is no universal threshold, but a working rule is to investigate component variances that persist two periods running or that would change a pricing or sourcing decision if they held. A large one month freight spike in a known disruption may need no action. A small duty variance repeated every month is a misclassification or a wrong rate in the model, and it compounds.
Whoever owns product margin, with the broker's data and the freight invoices in the room. Variance dies when it is owned by a reporting function alone, because the causes live in purchasing, logistics, and customs decisions, and the review only matters where those decisions can be revisited.