After forty-four years and forty-two countries, I have learned that the most expensive problems in any operation are not the ones your systems flag. They are the ones your systems never connect.
Profiling tells you what your data looks like. Reconciliation tells you where the numbers do not agree. But step three is different. Step three asks a harder question: which datasets should be talking to each other and are not?
This is where the money is.
What missing data looks like
Every organisation I have worked in has data sitting in silos that were never designed to speak to each other. The systems themselves are working correctly. The data is accurate. The reports balance. Nobody is doing anything wrong. And yet the combined picture — the one you can only see when you put the two datasets side by side — reveals a loss that nobody has been measuring.
- Purchase orders against inventory receipts. The PO system knows what was ordered. The warehouse system knows what arrived. When you join them, you find the gap — goods that were paid for and never received, or received and never matched.
- Equipment location logs against maintenance schedules. The GPS or RFID system knows where the asset is. The maintenance system knows when it was last serviced. Neither talks to the other. The result: equipment running past safe limits in the field while the scheduler believes it is in the yard.
- Freight invoices against carrier contracts. The accounts payable system processes what the carrier bills. The contract management system holds the agreed rates. When you join them — often for the first time — you find years of overbilling that no one disputed because no one was comparing the two.
- Labour time records against project codes. Payroll knows who worked and when. Project management knows what was budgeted. The join reveals which projects are absorbing undeclared overhead and which are being systematically undercharged.
The rail car engagement
The clearest example I can give you came from a rail operation in the Middle East. The company had two systems that were both functioning well. The first tracked the physical location of every rail car across the network — GPS-based, updated continuously, accurate to the hour. The second managed the load schedule: which cars were allocated to which runs, on which dates, carrying what.
These two systems had never been connected. Nobody had asked them to be. Each team — operations and scheduling — worked from their own system and their own reports. The operational team knew where cars were. The scheduling team knew where cars were supposed to be. They just never looked at both at the same time.
When we joined the two datasets, the picture was immediate. A significant proportion of rail cars were sitting idle in locations that the schedule had not accounted for. Not broken. Not lost. Not delayed by any recorded event. Simply parked — because the scheduling logic had never been tested against actual physical position.
The recoverable utilisation loss was approximately ten million dollars per year. The answer had been sitting in the data for years. Both systems were doing exactly what they were supposed to do. The gap was the space between them — and no one had thought to compare the two.
Why this never surfaces in normal reporting
Standard management reporting is built around what your systems were designed to measure. KPIs are drawn from single-source data. Dashboards reflect what one system knows. Variance analysis compares this period to last period in the same dataset.
A missing connection, by definition, does not appear in any of these. You cannot build a KPI around a relationship that no one has defined. You cannot show a variance against a baseline that does not exist. The loss is invisible not because of bad reporting, but because the question was never asked.
This is why the diagnostic has to be systematic. If you only look at what your existing reports show you, you will only find the problems that your reports were designed to find. Step three forces the question outward: what should we be measuring that we are not?
How we approach it
The process is not complicated. It requires someone who understands both the operational workflows and the data architecture. We start by mapping every major data-generating system in the operation: ERP, scheduling, logistics, maintenance, HR, procurement, finance. We then work through the operational flows and ask a simple question at each handoff point: is there a system on each side of this transaction, and are they connected?
Where they are not, we treat that as a hypothesis to test. We extract data from both systems, build a provisional comparison, and look at what it shows. Many provisional comparisons reveal nothing unexpected. Some reveal exactly what we suspected. Occasionally, one reveals something that changes the entire picture of where the operation stands financially.
The fixed-fee data diagnostic that AJM runs is structured precisely around this process. We do not presuppose what we will find. We work through the data methodically — profile, reconcile, join — and report what is there. If all the connections are in place, we tell you that. If there is a ten-million-dollar gap sitting between two systems that were never compared, we show you where it is and what it would take to close it.
The question worth asking
In forty-four years of operational work, I have never seen a large organisation where every relevant data join was already in place. The question is not whether there is a missing join in your operation. The question is how much it is costing you, and whether you have looked.
Both systems working. Never connected. The answer sitting in the data for years.
That is step three.
Related: What I look for in the first data export — step one of the diagnostic. And why a fixed-fee pilot is the right way to start.