The Water Cannot Stop

A water-treatment plant is an awkward place to run a project. Pumps have to be replaced, switchgear has to be isolated, concrete has to cure, new controls have to be proven. And throughout all of it the plant is still expected to produce safe water for the people who did not agree to be part of anyone’s project.

The project can stop. The water cannot. That single asymmetry is what makes the work interesting, because it means the schedule is not really a plan for building something. It is a plan for how long a town can safely go without part of its plant.

Three days that take four months

Replacing an intake pump and its switchgear needs a three-day shutdown. Three days is the visible part — the part with cranes in it. The work that earns those three days starts months earlier, and almost none of it belongs to the people doing the installation.

The equipment has to be built and delivered. Its certification has to be accepted. A temporary bypass has to be installed and proven. The reservoir has to hold enough operating reserve to cover the outage. Electrical isolation has to be agreed. The laboratory has to have staff free to test, and the regulator has to have someone ready to read the results. Customers have to be told, neither too early nor too late.

Every one of those belongs to a different organisation, and every one of them is perfectly well managed inside that organisation. The manufacturer knows its delivery date. The laboratory knows its rota. Nobody is being careless. They simply keep separate calendars, and separate calendars only meet at the moment they were supposed to have agreed.

One Project Page, and what sits under it

The instinct with work this cross-organisational is to give each party its own project and then hold meetings to reconcile them. That is the arrangement that produced the problem. The alternative is one Project Page for the whole operation, divided the way the responsibility is actually divided:

Each team holds its own work and its own conversation. The dependencies are drawn only where one team’s output is genuinely another team’s starting condition — which, on an operation like this, is a much shorter list than the meeting invitations suggest.

The shutdown window is itself a node, and it is worth being precise about what kind. It is not a piece of equipment and it is not anybody’s labour. It is the span of days during which the plant may safely be taken out of its ordinary state, and everything above it on the board is a condition that must be true before that span can open. Read that way, the schedule stops answering “when does construction happen” and starts answering something more useful: is this organisation ready to permit construction to happen at all.

Everyone sees enough, nobody sees everything

A shared board fails the moment it becomes a broadcast. The manufacturer should not be reading the utility’s internal cost discussions; the laboratory does not need the mechanical sequence. But each of them does need to know what their own lateness would cost, and that is a different question from “who is allowed to see the project”.

The External Collaborator seat is the one worth dwelling on, because it is the seat that makes an outside supplier honest without making them a member of anything. The manufacturer can see that seven activities across three teams are waiting on their certificate. They cannot see the contract, the internal chat, or the rest of the programme. That is usually enough — a supplier who can see what is queued behind them rarely needs to be chased.

Two days out

Two days before the window, the electrical team finds that the switchgear certification package is incomplete. The manufacturer’s file arrived without the final protection-relay test certificate. The switchgear is physically on site. The contractor is mobilised. The bypass has passed its test. The public notice has gone out.

In the ordinary arrangement this is not yet a project problem. It is a procurement problem, or a document-control problem. Procurement chases the manufacturer. The electrical contractor waits. Operations carries on preparing. Communications does not know the date is in question. The laboratory and the regulator hold time they may not need.

There is no single problem yet — there are five organisations each holding a different version of one. The cost of that arrangement is not the missing certificate. It is the days that pass before anyone can see the whole shape of it.

So the electrical lead files a Status against the certification node: the package arrived without the relay certificate, and approval cannot complete until it is issued. Filing it moves nothing on its own. What it does is put the problem where the dependencies already are, so the question “what is waiting on this?” has an answer that nobody has to assemble by hand.

Novus does not produce the certificate and does not have an opinion about relay testing. The document stays in the utility’s own document system, where it belongs. What the board contributes is narrower and harder to get anywhere else: what the missing certificate is doing to time, expressed in the only unit everybody in this operation shares.

The decision that structure makes possible

Now the owner has a choice, and it is a real one — hold the date and gamble, move the whole window and remobilise everybody later, or keep going with the parts that are still safe while the electrical chain waits.

The third option is the valuable one, and it is invisible without the structure. It requires knowing precisely which work depends on the certificate and which does not — that the temporary bypass, the mechanical staging and the laboratory preparation are untouched, and only isolation and installation are gated. That is a question about the dependency graph, and it is answerable in seconds on a board and not at all in a meeting.

The certificate arrives one day late. Because the unaffected work carried on, the whole operation moves by a day instead of collapsing into a full remobilisation. Records keeps the original Status, the decision the owner took, the reason, and the time the controlled overlap gave back.

What a day is worth

Suppose this operation carries $18,000 a day for the temporary bypass, $22,000 a day in contractor and specialist standby, and $5,000 a day in laboratory, regulatory and logistical rescheduling. Then a shutdown day is about $45,000 of time-dependent exposure, and the three days that the controlled overlap protected are worth roughly $135,000.

That is not a claim about Novus and it is not a saving anyone can promise you. Those rates belong to a particular utility, a particular contract and a particular plant, and there is no cost field anywhere in this product — every figure above is the reader’s own arithmetic. What Novus supplied was the number those rates get multiplied by, early enough to still be a decision rather than a post-mortem.

Where the credit belongs

Novus did not issue the certificate, did not judge the relay test, and did not decide anything. What the structure did was stop a missing document from staying a procurement problem long enough to become an operational one — by putting the people who had to act in front of the same dates at the same time, each seeing enough to understand the consequence of waiting and no more than that.

The schedule was never the hard part of this operation. Four organisations keeping four honest calendars was the hard part. One board, divided the way the responsibility is actually divided, is what turns four calendars into one operating day.