A purchase order gets raised, the material shows up eventually, and at some point it's installed — but most firms can't answer, for a specific delivery, exactly when it arrived, whether it matched what was ordered, or which crew installed it and on what day. Each gap in that chain is a place a shortage, a defect, or a billing dispute goes undetected until it's expensive.
Key takeaways
Tracking a material from order to install cleanly takes three separate, linked records: the purchase order itself, a delivery confirmation, and an installation record. Most firms have the first one reliably — a PO gets raised in the system before material is ordered — and lose the thread somewhere in the other two.
One field platform's own guidance is explicit about what a PO is for: a financial commitment against pre-determined costs already sitting in the budget, not an open account to draw against as material trickles in. That distinction matters because a firm treating the PO as the only record ends up unable to answer a simple question later — did this specific delivery match what was ordered, on this specific date — because the PO alone was never designed to carry that information.
The running-tab mistake is common enough that at least one vendor addresses it directly in its own documentation: the Commitments tool's own guidance states plainly that if a firm's practice is to invoice as items are delivered rather than against a single pre-agreed price, the right tool is a direct cost, not a purchase order — treating a PO as an open-ended running account is described as a misuse of the tool, not a variant use of it.
The practical read for material tracking: a PO fixes the price and quantity at the point of order. Everything that happens after — partial deliveries, delivery-by-delivery confirmation, quantities actually received against quantities ordered — needs its own tracked record layered on top of the PO, not folded into it as running edits.
A daily log's Deliveries category is the natural point to record that a shipment arrived, separate from the PO that authorized it. That single entry — date, quantity received, condition on arrival — is what later confirms a delivery happened as ordered, rather than leaving the question to whoever remembers unloading the truck.
Installation is the third and often weakest link: the same platform's Quantities category, covered in more detail in our companion piece on tracking installed quantities, is where material actually going into the building gets counted. Without a consistent link between a PO line, a logged delivery, and a logged installed quantity, a firm can end up with material sitting on site for weeks with no record of when it actually went in — which matters the moment a shortage or an installation defect needs to be traced back to a specific delivery.
Material substitutions and mid-order price changes are common enough on a live job that they need their own record rather than a silent edit to the original PO. A Commitments tool that separates change orders into their own tracked view — status, amount, due date — keeps the original order intact as a reference point while the substitution or price change is tracked as its own dated event.
That separation matters for the same reason a field note needs to be tied to its originating RFI: a reviewer checking a final invoice against the original budget needs to see what changed, when, and why — not a PO that's been quietly revised with no trace of the original figure it started from.
Everything above assumes the PO itself was priced well, which is a separate problem that gets solved earlier in the process. Our companion piece covers negotiating supplier terms with real data rather than assumption, and our companion piece on long-lead items covers the schedule-risk side of ordering material with a long lead time far enough ahead that a tracking gap doesn't also become a schedule delay.
Material tracking breaks down fastest where responsibility for each step isn't assigned to a specific person. A PO raised by an office estimator, delivered to a site the estimator never visits, and installed by a crew reporting to a superintendent who never saw the original order is three people who each know one-third of the story.
Naming, for each PO, who confirms the delivery and who confirms the installed quantity — not as a general job description but as a specific line on the PO itself — closes that gap before the order is even placed, rather than trying to reconstruct who was supposed to be watching after a shortage is already discovered.
A worked example
A mechanical package raises a single PO for $145,000 covering piping, fittings and specialty valves, delivered across four shipments over six weeks. Tracked only against the PO, the budget shows $145,000 committed from week one, and there's no record distinguishing which of the four deliveries arrived, or whether the third shipment's valve count matched the order — until an installer flags a shortage in week five with no paper trail to check it against.
Tracked with delivery and quantity records layered on the PO, each of the four shipments is logged on arrival with quantities received against the PO line; a shortfall on the third delivery — 18 valves received against 24 ordered — is caught the day it arrives, not the day an installer runs out mid-task. The PO stayed the fixed reference point throughout; the three intermediate records are what actually caught the gap.
No. Per one platform's own documentation, a PO is a financial commitment against a pre-determined cost, not a running record of what's been delivered or installed — those need their own separate, linked entries.
A direct cost, not a purchase order. The vendor's own guidance is explicit that a PO is "not designed to be used as a running tab" for that kind of billing pattern.
As its own tracked change order, not a silent edit to the original PO. A Commitments tool that keeps change orders in a separate view — with their own status, amount and due date — preserves the original order as a reference point while the substitution is tracked on its own dated record.
A 30-minute call is enough to tell you whether AI pays for itself here.
