The instruction that was not one.

What that firm actually is

The company running the site isn't building anything. Their agreement with the employer is for services — management, engineering, procurement on the employer's behalf — and it is usually reimbursable or fee-based rather than lump sum.

Two consequences follow immediately and neither is obvious from the site.

They carry no delivery risk. A main contractor who lets the programme slip pays for it; a firm paid a fee for managing it doesn't, and on a reimbursable basis is paid for the extra months. Nothing follows from that about how any particular firm behaves. What follows is that the pressure a main contractor feels on the programme has no equivalent here, and if it is going to exist it has to be put in the agreement.

And they are not a party to any of the package contracts. Their construction manager can be the most senior person on site, chair every meeting, and still have no contractual relationship with the crew taking his instruction.

Named, or not named

Whether an instruction binds anybody comes down to one thing: whether the package contract names that firm.

Where it does — as the Engineer, as the Employer's Representative, or by an express delegation — the instruction has the standing Contract Week 3 describes, and everything in Contract Week 6 applies unchanged.

Where it doesn't, that person is a third party with an opinion. A reasonable, expert, well-informed opinion, from somebody the employer is paying to have it. It is still not an instruction, and work done on it is work done at the contractor's own risk.

The uncomfortable part is that the same firm can be named in one package contract and not in another. Same person, same site, same meeting: binding on the civil contractor and advisory to the mechanical one. Nothing on the site distinguishes the two, and neither of them has read the other's contract.

The version where the question cannot arise

This arrangement has a family and a sibling, and the sibling is worth knowing because it is the case that defines the rule.

The family is construction management, and the firm described so far is the version that acts for the employer. It coordinates the designers, the trade contractors and the suppliers, and it isn't contractually answerable for the works being finished. A managing contractor on a process job and a construction manager on a building one hold the same position under different names, in different markets, and the week's question applies identically to both.

The other version closes the gap on purpose. There the manager coordinates as before and also takes on the construction phase, carrying the risk of completion that a main contractor carries — and it signs the construction contracts itself. Design and the preconstruction appointments stay with the employer; everything from the ground up runs through the manager.

Which settles this week's question before anybody asks it. A manager holding the package contracts is a contracting party, and its instructions bind because there is nothing else left for it to be. The question of whether somebody is named only exists in the version where they hold nothing.

Why the second version was invented is the sharpest support anywhere for the claim this article started from, that the firm running the site carries no delivery risk. Where the manager acts purely as coordinator, a decision it takes in running the site is the employer's decision. Impose a procedure across the whole site, have it turn out to be wrong, and the exposure lands on the party the manager was acting for rather than on the manager. That is the loophole the at-risk version was built to close, and it is the same absence of delivery risk seen from the employer's side rather than the contractor's.

So there is a question before the one this week has been asking. Not whether your contract names them — whether they signed anything at all.

Why it holds together anyway

None of this surfaces while the instructions are small, and the reason it stays quiet is also the trap.

The instructions are generally sound, and structurally they would be: somebody with sight of the whole site sees a clash the individual packages can't, so moving the pipe run is the correct answer. A contractor complies because the person asking is running the job, because refusing costs goodwill, and because the work is small enough that arguing about it costs more than doing it.

So instruction after instruction is absorbed without complaint, and a working convention forms: that firm instructs, the packages comply. The convention is real and the site runs on it.

What it isn't is a contractual position, and the difference only appears on the instruction that turns out to cost money. By then the convention has been operating long enough that everybody on site believes it is how things work — and in the only sense that matters day to day, it has been.

What a refusal would expose

It is worth working through what happens if one contractor declines — reads their contract, sees that the instructing party isn't named in it, and asks for the instruction to come through the employer.

They are within their rights, and it will be read as obstruction, because every previous instruction went through without this. Resolving it takes a confirmation from the employer or a delegation being produced, and work restarts.

The useful part is what the episode makes visible. One party knowing where it stands is enough to establish that the others didn't, and the confirmation that follows is often the first time the authority has been written down at all. If it happens on a package you are planning, the days lost are not obstruction. They are the unwritten convention being tested for the first time.

What a planner does with this

Not much about the authority itself — that is a commercial question and it belongs to the contracts team. What sits with project controls is the record.

Every instruction that changes what gets built or when needs to be captured with three things: who gave it, on what date, and under which contract they were acting. The third is the one nobody writes down, and it is the one the argument turns on six weeks later.

Where the answer to the third is unclear, that isn't a reason to stop work. It is a reason to say so at the time, in writing, to the party you do have a contract with. Reporting Week 23 made the case for the written record before the phone call; this is the same habit where the stakes are contractual rather than numerical.

None of that requires taking a position on whether the instruction was valid. It requires only that the question can be answered later by reading something rather than by remembering a site walk.

System design

None of this is a judgement about whether an instruction was valid. It is the set of records that lets somebody answer that question later by reading rather than remembering.

Record	Produced by	Required quality	Verified against	Feeds
Authority register	Contracts	One line per party: named, named without powers, or not named	The clause in each package contract	Whether an instruction is valid
Instruction log	Project controls	Who, when, and under which contract they acted	Minutes, marked drawings, email	Variation · delay event
Query on unclear authority	Project controls	Sent to the party you are in contract with, at the time	The instruction itself	Entitlement, months later
Convention note	Project controls	What the site does in practice, recorded as practice	Observation	Explaining a year of compliance

The last row is unusual and worth keeping. A year of compliance with an unnamed party is itself evidence of how the project was run, and it is the only thing that explains why nobody objected until the instruction that cost money.

Practical insight

Find out this week whether the firm running your site is named in your own contract. It is one clause, somebody in commercial can point you at it in a minute, and you will get one of three answers: named with defined powers, named with powers nobody defined, or not named at all. Where you sit changes what your instructions are worth.

Then take last month's instructions — your meeting minutes, your marked-up drawings, your emails — and count how many of them record who gave the instruction and under what authority. On most multi-contract jobs your answer will be close to none.

Add the column this month rather than reconstructing it in a claim two years from now, when the person who gave you the instruction has left and you are working from memory.

Key takeaways

• A managing contractor holds a services agreement, not a works contract, and builds nothing.
• They carry no delivery risk. Time lost is time paid for, which changes the incentive to compress.
• Whether their instruction binds depends on whether the package contract names them.
• Named as Engineer or Employer's Representative, everything in the contract track applies unchanged.
• Not named, the same person is an expert third party and compliance is at the contractor's own risk.
• The same firm can be named in one package contract and not in another, and nothing on site shows which.
• A working convention forms because the instructions are sound and the work is small enough not to argue about.
• Capture who instructed, when, and under which contract. The third is what the argument turns on.
• The same delivery method has a version where the manager signs the construction contracts itself. There its instructions bind, because it is a contracting party rather than a third one.
• Where the manager only coordinates, its decisions are the employer's decisions and the exposure lands there. That is why the other version exists.

Records born here. The authority register, one line per party · the instruction log with its authority column · the written query when the authority is unclear.

What is coming next

If a managing contractor is named as the Engineer in three package contracts and not in the other two, there isn't one Engineer on the project. There are three, or none, depending on which contract is being read.

Next week: determination and impartiality when every package has its own Engineer.