Every shipment carries two landed costs: the estimate used to price and buy, and the actual that arrives across invoices over the following weeks. The gap between them is where margin explanations live. This guide covers how to measure landed cost variance and what to do with each kind.
Variance only means something at the level where costs are incurred: the shipment, broken into its standard components of goods, freight, insurance, duty, and fees. Comparing a month's total estimated freight against a month's total actual freight mixes products, lanes, and timing into a fog. Shipment level rows let each difference be traced to a cause a person can name.
The actual side needs a closing rule, because invoices arrive over weeks and a few straggle for months. A practical system declares a shipment's actuals complete at a set age, books late arrivals to a variance reserve, and reports the reserve honestly instead of letting one stray invoice rewrite old product margins in a later quarter.
Rate variance says the market moved: the same lane cost more than the estimate assumed. Volume and mix variance says the shipment itself differed: more weight, a different mode, a split delivery. Assessment variance says the duty or tax base differed from the estimate's assumption. Error variance, the smallest category in a healthy system, says somebody typed or mapped something wrong.
Each cause has a different owner and a different fix. Rate variance updates the estimating tables. Mix variance goes back to planning and supplier packing. Assessment variance goes to classification and valuation practice. Lumping them into one red number produces one vague action, usually an argument, whereas the split produces four small, obvious ones.
Estimating tables should carry a memory: the rolling actual cost per lane, mode, and product family, refreshed from closed shipments, so next quarter's estimates start from evidence rather than from the rates someone remembers negotiating. The variance report is the mechanism that moves that memory, and skipping it is why some companies' standard costs fossilise at the year they were first set.
Set a threshold for investigation rather than reviewing every line. Shipments beyond a variance band, by percentage or by money, get a named cause before they close. That single habit creates a learning loop small enough to sustain and strict enough to catch the systematic leaks, which are always worth more than the random ones.
Product margin reviews change character when variance is on the table. A product can be a good product badly estimated, or a bad product accurately estimated, and the response differs completely: fix the table in the first case, fix the price or the product in the second. Without the split, teams defend products with costing arguments and attack costing with product arguments, in a circle.
Over time, shrinking variance is itself a performance measure for the supply chain function. It says forecasts are honest, suppliers ship what was ordered, and classifications hold up. A business that can predict its landed costs within a few percent prices with confidence, and confidence, in margin decisions, is worth money.
Keep reading: SKU level landed cost allocation: choosing a method and sticking to it, Landed cost data quality: fixing the supplier invoice errors that poison your numbers, and How can landed cost data improve your next Incoterms negotiation with suppliers? cover the nearby cases in this series.
The difference between the landed cost estimated for a shipment, the figure used for buying and pricing, and the actual cost once all freight, duty, and fee invoices settle. It is measured per shipment and per cost component so each gap has a traceable cause.
Monthly totals mix lanes, products, and timing, so differences cannot be attributed. Shipment level comparison lets each variance be classified as rate movement, shipment mix, duty assessment, or plain error, and each cause has a different fix.
Use a closing rule: declare a shipment's actuals complete at a fixed age and book late arrivals against a variance reserve reported openly. That prevents one stray invoice from silently rewriting old margins months later.
By feeding closed shipment actuals back into the estimating tables as rolling lane and product costs, and by forcing a named cause on any shipment outside a set variance band. Estimates start improving from evidence instead of memory.