Insights — what the panels measure, and how to check them

The generated headline

The panel leads with a written claim about the shape of the network — how fragmented it is, what mode dominates, where the volume sits. Hold onto that sentence rather than skimming it. It is usually the most valuable thing on the page, and the pallet arithmetic further down is where it turns into money.

Start with the checks that are free

Every chart carries a chip with its own total. Add the visible parts and compare. A panel whose bars sum to its own chip is internally sound, and establishing that first is what makes any later disagreement between panels provable rather than merely suspected.

The same data, regrouped

Spend by lane and spend by destination state are usually the same rows grouped two ways, so their chips will differ by whatever one panel includes and the other does not. Work out that difference and look for it: if it equals the lanes visible in one panel but not the other, nothing is wrong — the two simply have different cut-offs. If it does not, you have found something.

Written findings carry their own totals

The key findings compute against a total of their own, which may not appear on any chart. Percentages inside them are often right even when a per-unit figure is not, so check the two separately: recompute one share to confirm the base, then divide the total by the relevant volume to test any per-pound or per-shipment claim before quoting it.

Two panels with the same metric name

If a page shows one metric name twice with different numbers, work out the ratio between them for several categories. A constant ratio means a unit or currency scale, which is a formatting problem. Ratios that vary by category mean the two panels are counting different sets of rows under the same label — a data problem, and the one to resolve before a customer reads the page.

The KPI cards, and how to prove them

The card row is the block most worth verifying, because it is the block you will quote from. Four multiplications settle it, and when they all agree the cards are consistent in every direction.

• Total weight — Shipments × average weight.
• Cost per pound — Total spend ÷ total weight — and it should also equal cost per shipment ÷ average weight.
• Cost per shipment — Total spend ÷ shipments.
• Mode share — Mode count ÷ total shipments.

More than one count of the same thing

A single page can carry several shipment counts — one on a card, one on a chart chip, one inside a written finding. Sum a chart's plotted points to see which figure that panel agrees with, and when quoting money, use the count the money was computed from. Averages are only meaningful over the population that produced them.

Time series with gaps

Read the axis labels before you read the line. A chart can place unevenly spaced periods at even intervals and interpolate straight across a hole of many months, which draws a trend through data that does not exist. When the labels jump, read the chart as two clusters rather than one trend, and note how much of the volume sits in the recent cluster.

Donuts round small shares away

A mode split reading a clean hundred percent can sit beside a card saying not-quite-all. The card is the precise one. Check it before telling a customer their network is exclusively one mode.

The finding that is worth money

This is the calculation to do on every network. Divide average shipment weight by average pallets to get the actual freight riding on each pallet position, then compare it against the tariff's LTL pallet minimum. Billed weight is pallets multiplied by that minimum, so the ratio of billed to actual is the cost of fragmentation expressed as a single number — and on a single-pallet network it is a multiple, not a margin. No panel on the page connects the fragmentation headline to the minimum for you, which is why this is the recommendation most often missed.

• Actual — Average weight ÷ average pallets = freight per pallet position.
• Billed — Average pallets × the tariff minimum per pallet position.
• Ratio — Billed ÷ actual — the cost of fragmentation, and the case for consolidating.

Three smaller checks

Lanes and destinations should agree when every lane leaves one origin, so a mismatch means a destination with two lanes or a lane terminating at its origin. The summary and the deep-dive are generated separately and can be days apart, so check both timestamps before treating the page as one document. And number formatting can vary between panels — locale grouping that changes on larger figures is cosmetic, not a data problem, which is why it tends to show up on six-digit totals and not on the bar labels.

The deep-dive, and how to use the page

Get detailed insights expands several written sections — freight profile and volume, rate structure and cost drivers, contract terms and service levels, network shape and expansion. Read those for the argument, not the arithmetic. And read the whole panel the same way every time: verify the KPI block, do the pallet calculation, check anything that appears twice under one name, treat a gapped time series as clusters, and quote from what reconciles.