Critical Material Delivery ETA Prediction

The useful output is a risk flag with two kinds of cite: what the supplier last promised, and what logistics status currently shows. It is not a new purchase-order date, not a revised contract milestone, and not a reason to overwrite the committed delivery on the PO.

Procurement still owns the date that vendors, the GC, and trade partners treat as the commitment. The model's job is quality: surface that the committed date and the evidence trail no longer agree, early enough that the lookahead can absorb a decision instead of a surprise.

A flag should name the material, the committed date that still sits on the PO, the predicted window, the gap in working days, and the cites that produced that gap. If a cite is missing, the flag should say so. Silence in the evidence is a quality problem, not permission to invent a delay story.

Do not treat the predicted date as a purchase-order commitment. If planners start scheduling to the model date, the PO date becomes folklore, expedite conversations lose a baseline, and nobody can tell whether a supplier slipped or the project quietly moved the goalposts.

Score items whose lead time can still move the critical path if they slip after award. Typical candidates are mill-order steel, switchgear, elevators, curtain wall, precast, custom air handlers, and any package whose shop-drawing cycle sits inside the remaining float.

Do not score every purchase. Commodity concrete, stock fasteners, and local lumber rarely need an ETA model. Ocean freight alone is not a scoring criterion. If you flag every ocean shipment as late because it is on the water, the board fills with noise and the few factory-slot risks disappear.

A workable list starts from the PO register and the remaining activities that need the material on site. Keep the item if both are true: the supplier lead (fabrication plus transit) is long relative to remaining float, and a late arrival would freeze a pour, a hoist window, or a follow-on trade. Drop the item when the warehouse already holds it, when the supplier has delivered a signed ship notice that matches the PO date, or when the activity has already been re-sequenced off that delivery.

Construction teams already keep POs, submittals, and activity IDs in platforms such as Autodesk and Procore. Use those records as the source of the committed date and the latest supplier document. Do not invent a second materials system. If the PO date and the latest supplier confirmation disagree, that disagreement is the first cite, not a data-cleaning chore to hide.

Build a score from evidence, not from a single predicted calendar day. For each long-lead line, collect three cites if they exist: the PO promised date, the latest supplier confirmation (order acknowledgment, production slot, or factory-test date), and logistics status (ex-works ready, booked transit, in transit, or arrived). Then compare those cites to the first on-site need date on the schedule.

A simple scoring rule is enough to start. Green: supplier confirmation and logistics status both support the PO date with float still covering a normal transit miss. Amber: one cite is missing, or the supplier confirmation sits later than the PO while logistics has not yet booked. Red: both remaining cites point later than the need date, or the supplier confirmation is later than the PO and logistics shows the material is not yet ready to leave the factory.

Never score from logistics status alone. Missing a supplier confirmation is a common way to manufacture a red flag. A tracker that only sees a booking request, or that sees nothing because the mill has not issued an ASN, will treat silence as delay. Silence is incomplete evidence. Hold the score at amber, name the missing confirmation, and give procurement a chase list. That chase is closer to subcontractor compliance chasing than it is to a schedule rewrite: someone still has to get a human promise on record.

Do not invent a port delay to fill a gap. If the ocean AIS feed is quiet, say the ocean cite is unavailable. If the material is domestic trucking from a factory, do not smear it with a generic "international logistics risk" tag. The score should be explainable in one sentence a buyer can read on a call with the mill.

Illustrative example (not a measured case): a hospital electrical room needs 15 kV switchgear before the feeder pour freeze. The PO still shows a delivery that supports the freeze. The supplier confirmation, received after drawing approval, places factory test after the week the install crew is on the lookahead. Logistics status is factory-not-ready; no ocean leg exists. The model should flag red because the supplier cite and the factory status both sit after the need date, and it should cite those two records plus the PO date it did not overwrite. Procurement calls the factory. Planning puts the freeze at risk on the lookahead. Nobody changes the PO until the supplier issues a new written promise.

A red or amber flag is a constraint, not a new activity date. Push it onto the lookahead constraint scan as a named threat against the need-by activity: material, score, cites, and the owner who will close it. The lookahead still uses the PO date as the committed receipt unless procurement has issued a change.

That distinction keeps two conversations from collapsing into one. Planning asks whether the pour, hoist, or area turnover can survive the cited gap. Procurement asks whether to expedite, split the shipment, or accept a later written promise. If you replace the PO date with the model date inside the CPM, both conversations lose their baseline. Recovery options then belong in schedule recovery scenario generation, after someone has confirmed the gap is real.

When the material will arrive, site still has to receive it. A late switchgear crate that now lands in the same week as a tower-crane pick is a laydown problem, not an ETA problem. Hand that to site logistics layout optimisation only after procurement has a confirmed ship window. Do not let the ETA model redraw the gate plan from a predicted day.

Close the flag only when one of three things is true: logistics status now supports the PO date, procurement has issued a PO change from a new supplier promise, or planning has re-sequenced the need date so the cited window no longer threatens the activity.

Watch for three ways a useful flag becomes a false commitment.

Replacing the PO date with the model date leaves the supplier measured against a number they never promised. Keep the predicted window in a separate field. Change the PO only after a new written supplier promise.

Missing a supplier confirmation, then scoring from a blank tracker or a booking request, treats silence as delay. Hold those lines at amber with "confirmation outstanding" and chase the mill before you treat the gap as real.

Flagging every ocean shipment as late because it is on the water fills the board with noise. Score the ocean cite against the booked arrival band and remaining float. A vessel inside its booked window is not a threat. A factory that has not released cargo is.

When a flag is wrong, correct the cites. If the mill still hits the PO date, clear the flag. If the mill confirms later, procurement owns the PO change.