On September 15 the White House widened its 50% tariffs on Canadian goods to dozens of additional product lines, including some steel and aluminum, furniture and paper. On September 29, parts of that list stop being a tariff and become an outright import ban. Canada has had counter-tariffs of 15% to 50% running on more than 700 US product lines since September 8, and Washington has confirmed the ban list covering alcohol, dairy and motorcycles. Meanwhile, on the other side of the world, Asia-Europe schedule reliability has fallen to roughly 10% after four typhoons worked through Shanghai, Ningbo-Zhoushan and Yantian, and carriers have pulled around 350,000 TEU out of weeks 39 to 43 (The Loadstar), the kind of blank sailing wave that rewrites an arrival plan without anyone being told.
Two regions, two completely different failure modes, hitting in the same month. Sooner or later someone in a quarterly review asks the obvious question: which region is actually hurting us more, and where should we put the people? In most companies the honest answer is that nobody can tell. Not because the data is missing, but because the North America number and the Asia-Europe number were never built to be compared.
Two shocks, one scoreboard
The instinct is to put both regions on the same slide. On-time delivery by region. Cost per shipment by region. Exceptions per hundred orders by region. It looks like a fair fight, and it almost never is.
North America's number is probably assembled from a TMS that sees domestic truckload well and cross-border poorly. The Asia-Europe number probably comes from a forwarder's portal, arrives in weekly batches, and counts a different set of milestones. One region's "late" starts at the customer's promised date. The other's starts at the carrier's revised ETA, which means the carrier gets to grade its own exam. Ranking those two against each other does not produce an insight. It produces a decision made on an artefact of the plumbing.
This is the layer underneath the reporting problem, and it is why a shared data model matters more than another dashboard. Before you can compare regions, the regions have to be speaking the same language. That is a data normalization problem rather than a reporting one, which is why it gets solved in a command center sitting above the regional systems, the way Orkestra does, instead of inside whichever system happens to own each region.
The same KPI name, two different clocks
OTIF is the usual offender. Every region reports it. Almost no two regions define it the same way.
The questions that decide the number are boring and rarely written down. Does the clock start at the original promise date or the most recent revision? Is a delivery on time if it lands inside a two-hour window, a same-day window, or any time before the receiving dock closes? Does a partial shipment count as a miss, a fraction, or a separate line? Does a customer-requested delay count against you?
Each of those choices can move OTIF by several points on its own. Stack four of them and you can move a region's score by more than any operational improvement it made that quarter. If you want the mechanics of pinning these definitions down before they reach a scorecard, the layered KPI approach is the place to start. The same discipline applies to transit time, where the clock boundaries decide the number as much as the lane does (how to measure transit time variability).
So the first act of benchmarking is not analysis. It is writing down, for each region, the exact event that starts the clock and the exact event that stops it, and then making them match.
The same cost line, two different contents
Cost per shipment has the same problem in a more expensive form.
Ask each region what is inside that number. One will give you linehaul only. Another will include fuel, one will include terminal handling and the peak season surcharge, and one will have quietly folded demurrage into an accrual that never reaches the operational report at all. In a month where surcharges are doing most of the moving, the region with the most honest cost accounting will look like the worst performer.
Tariffs make this sharper. A North America lane that just absorbed a 50% duty on a component is not operationally worse than an Asia-Europe lane. It is carrying a policy cost that no planner in that region caused or can fix. If your comparison cannot separate the duty line from the execution line, you will conclude that a team is underperforming when what actually happened is that a trade rule changed on September 15.
Carve the number into components that each region reports identically: linehaul, fuel, accessorials, duty and tax, penalty charges. Compare the components. The totals can stay on the slide, but they should not be what anyone argues about.
The region that looks best is often the one reporting least
This is the failure that costs the most, because it rewards exactly the wrong behaviour.
Exception counts, delay hours and dwell time all depend on somebody sending an event. A region whose partners push clean milestone data will surface more exceptions than a region whose partners send a spreadsheet on Fridays. On the scoreboard, the second region looks calm. It is not calm. It is dark. Then the quarter closes, the claims and detention invoices arrive, and the calm region turns out to have been the expensive one all along.
The defence is to report coverage next to every operational metric: what percentage of shipments in that region produced a full milestone set, and how old the newest event is. A region running at 40% event coverage does not get a performance ranking. It gets a data remediation plan. Response times deserve the same caution, since a slow median is often a coverage gap in disguise: check how exceptions hand off across time zones before concluding one region simply works slower. This is the same gap that makes monthly reports read like history rather than management, only worse, because here it also misallocates headcount.
Normalize first, rank second
Everything above points at the same conclusion: the comparison is only as good as the normalization underneath it, and normalization is not a reporting task. It happens upstream, at the point where carrier, forwarder, ERP, TMS and WMS feeds arrive, and it has to hold continuously rather than being redone by an analyst each quarter.
That is the job a command center layer like Orkestra is built to do before any chart renders. Feeds from 200 plus pre-built connectors land in one event model, so a milestone carries the same meaning whether it came from a European forwarder's EDI or a North American carrier's API, and a cost field contains the same components on both continents. The comparison stops being a reconciliation exercise and starts being a live view.
DBW Advanced Fiber is a useful example of what that changes. Running EU to NA flows on Orkestra, the team moved from spreadsheets to real-time visibility across the full corridor and took 18% out of supply chain cost. The gain did not come from a new carrier. It came from being able to see both ends of the same lane in the same terms, and acting on what that showed.
What to do when the numbers still will not line up
Sometimes a region genuinely cannot be normalized yet, usually because a major partner does not send structured events. Say so on the slide.
A benchmark with a clearly marked gap is far more useful than a complete-looking table that quietly compares a measured region to an estimated one. Mark the region as unranked, state the coverage percentage, and put a date on when the feed will be live. Leaders can work with a known unknown. What they cannot work with is a confident number that turns out to have been assembled from whatever was available.
Where to start on Monday
Take the last regional performance slide your team presented and pick the single metric that drove a decision. For each region on that slide, write down three things: the event that starts the clock, the event that stops it, and the percentage of shipments that produced both. You will usually find the definitions diverge in at least one region, and that coverage varies more than performance does.
Fix the definitions first, publish coverage alongside every metric, and only then rank anyone. A benchmark that survives that test is one people will actually act on. Keeping it true between quarters, rather than only on the day it was built, is what cross-system analytics running on normalized data does: in Orkestra the definitions live in the model, so the regional comparison is a live view instead of something an analyst rebuilds every time a trade rule or a schedule changes.
