Why site orders slip away, and what to do about it

When material is late on site, somebody goes looking for the culprit. They usually find one: a quote request nobody sent, a reply nobody read, a task everybody assumed a colleague had covered. But after the fifth incident it becomes clear the problem is not the people.

The problem is that the order has nowhere to live.

The order lives in one person’s inbox

The quote request goes out from one person’s mailbox. The supplier’s reply lands there. So does the confirmation. When somebody asks “have we ordered the roof covering yet?”, the answer exists — but only inside one person’s head and one person’s inbox.

The fix is not another shared mailbox. The fix is attaching correspondence to the item it concerns. When a supplier reply pairs itself automatically onto the “roof covering” row of project “Easy 47”, everybody can see the answer.

Nobody can compare the quoted price

A PDF arrives with thirty line items. Somebody opens it, scans the total and says “that looks fine”. That is not laziness — without a tracked market price there is simply nothing to compare against.

And yet this is mechanically solvable. If you know last week’s price for the products you track, you can label every quote line as cheaper, on par or more expensive. A five percent threshold is enough. Suddenly your supplier negotiation has an argument instead of a feeling.

The construction phase lives in the site manager’s head

The site manager knows the roof starts in three weeks, so the covering has to be quoted now. They know it right up until a date moves. Then they know that too — it is just that nobody else does.

Write the construction phase down in one place and tie the item order to it. Software can then say what is behind schedule without anyone having to work it out.

There is a gap between “delivered” and “installed”

Material arrived on site. The invoice is running. Installation is unscheduled, because nobody knows there is anything to install. This gap costs the most money and is the hardest to spot, because in most systems it has no status of its own.

But once you have four dates — requested, ordered, in stock, installed — the gap surfaces by itself. A row with the third date filled in and the fourth empty is precisely the problem.

What follows from this

None of these four problems needs artificial intelligence. They need somewhere for the order to live and a handful of rules that derive the next step from the data. You bring AI in only for what rules cannot handle — reading a scanned price quote, for instance.

The discipline is knowing which is which.