Landed cost data usually lives in a month end spreadsheet nobody opens. A weekly dashboard with a small set of honest numbers changes that: problems surface while the shipments causing them are still moving. This guide picks the five metrics worth a weekly slot and explains what each one is for.
Freight per unit, tracked for the main lanes and modes, is the fastest signal that the market or your shipping pattern moved. Averages hide mode shifts, so the number should be shown split by ocean and air at minimum; a lane whose air share quietly doubles announces itself here weeks before it announces itself in the freight invoice total.
Two: duty paid as a percentage of goods value. This ratio should be boring. When it moves, something real happened: a classification changed, a preference claim failed, a rate took effect, or valuation practice drifted. A boring metric that suddenly is not boring is the best kind of alarm, because it has so few innocent explanations.
Estimate accuracy, the share of recent closed shipments landing within a set band of their estimated cost, tells the business how much to trust the numbers it prices from. It belongs on the same page as the costs themselves, because a precise looking dashboard built on loose estimates is decoration, and this metric says so every week.
Port and warehouse dwell time, with demurrage and storage charges as its money shadow, captures the operational friction that never appears in rate cards. Rising dwell precedes rising charges by a week or two, which makes it one of the few genuinely predictive numbers in the set.
Every week has them: the air freight rescue, the split shipment, the re routed container. A single line totalling exception spend, with a one line note naming the biggest case, keeps exceptions visible as a category instead of letting them dissolve into lane averages. Exception spend that becomes normal spend is how freight budgets quietly double, and a weekly line is a cheap vaccine.
Keep the dashboard to one page and the same five numbers, reviewed in the same short meeting, with last week and the rolling average beside the current figure. Dashboards die from ambition: twenty metrics, customised per viewer, refreshed late. Five numbers that arrive every Monday without fail will change more decisions than a perfect system that ships twice a year.
Each metric needs a named owner and a pre agreed response: the freight number moves, bookings get reviewed; the duty ratio moves, classifications get checked; exceptions spike, someone explains the biggest case in the meeting. A dashboard without responses is a screensaver, and the five numbers earn their place only because each one is wired to an action someone has already agreed to take.
Review the set itself quarterly. Lanes change, a new market opens, a metric that alarmed usefully for a year goes quiet because the problem got fixed. Replace it. The discipline is not the five specific numbers in this guide; it is the habit of watching landed cost while it can still be changed.
Keep reading: Landed cost variance: comparing what you estimated against what the shipment really cost, SKU level landed cost allocation: choosing a method and sticking to it, and Landed cost data quality: fixing the supplier invoice errors that poison your numbers cover the nearby cases in this series.
Because month end explains costs after the buying and shipping decisions that caused them have already repeated. A weekly view built from operational data surfaces problems while the next shipment can still be changed.
It is a stability alarm. The ratio should barely move, so when it does, a real cause sits behind it: a classification change, a failed preference claim, a new rate, or valuation drift. It has very few innocent explanations.
The cost of rescues and irregular moves: emergency air freight, split shipments, re routing. Tracked as its own line it stays visible as a category; folded into lane averages it quietly becomes normal spend and inflates the freight budget.
Few enough to review in one short weekly meeting, five is a working number, each with a named owner and an agreed response when it moves. Dashboards fail by ambition and lateness, not by simplicity.