A tie-in project with a ten-day outage is really two projects: the months of preparation, and the ten days. They have different crews, different costs per hour and completely different risk profiles. Estimating them as one line hides both.
The habit is understandable. On the drawings it is one scope, one set of tie-ins, one contract. But the moment the plant comes down, the rules of construction economics invert. Time stops being a resource and becomes the scarcest commodity on the project, and every estimating assumption that was true the week before goes stale overnight.
Draw the boundary first
The first estimating decision is not a price. It is a line: what genuinely requires the plant to be down, and what does not. Only the work on the wrong side of that line belongs in the window, and the line is negotiable in ways that are worth serious money. A tie-in weld requires the outage. The scaffold for it, the cable pulled to the tie-in point, the spool fabricated and hydrotested, the valve staged at the gate: none of that does. Every scope conversation in the first month should be an attempt to move work across that line, out of the window. What the day-to-day losses of live-plant work do to the rest of the job is its own subject: see The productivity factor on live plants.
The window estimate
The work inside the window is priced by different arithmetic.
- Hour-by-hour schedule, not day-by-day. Ten days is 240 hours, and the estimate is built against that number, because the window does not round up.
- Crew sizes set by the space available, not by the quantity. Four welders may fit at the tie-in; a fifth adds cost and nothing else.
- Premium time on everything, including supervision, QA and the permit writers, because everyone works the window, not just the trades.
- Standby before the window and demobilisation after it: crews mobilised and waiting while isolation, draining and gas-freeing finish, then the reverse on the way out.
- The cost of a day of overrun, stated, because the owner will ask, and because half the decisions made inside the window are really purchases of schedule at that price.
A worked example
Take a ten-day window, and call the work inside it 8,000 direct labour hours. Compressed to two ten-hour shifts a day, the crews run at premium time; call the loaded labour and equipment cost 2.5 million dollars for the window, with another 400,000 in standby and demobilisation wrapped around it.
Now price the day. If the plant earns, call it, 700,000 dollars a day in lost margin when it is down, then one day of overrun costs more than a quarter of the entire window construction cost. That single number changes how every decision inside the window gets made. A second crane on standby at 15,000 dollars a day is not a luxury; it is insurance priced at two percent of the risk it covers. The estimate that never states the day cost leaves the superintendent making million-dollar decisions with no price list.
The preparation estimate
Everything that can be done with the plant running is priced as ordinary brownfield work, at near-normal productivity: pre-fabrication, modular assemblies, temporary supports, pre-installed valves and spools, scaffold built early, materials staged by tie-in point. This estimate looks boring next to the window, and it should, because boring is the point.
Every hour moved out of the window is bought at a fraction of its window cost, and it pays twice: once in the labour premium avoided, and again in overrun risk that no longer exists.
The cheapest hour on a shutdown is the one that happened last month.
The risk model
Shutdown risk is about time, not money, until it becomes money. So the model runs on the schedule, not on the cost lines: what is the probability the window is exceeded, by how much, and what does each day cost in lost production. Weld repair rates, discoveries at the tie-in points, weather on the crane picks, a late isolation: each risk is expressed first in hours, then converted at the day rate.
Run honestly, that number usually dwarfs the construction contingency, and it should be reported beside the estimate, not buried inside it. An owner who sees that the schedule risk is worth three times the construction contingency makes different decisions about pre-fabrication, spares and standby equipment, which is exactly what the model is for.
Where this goes wrong
One blended estimate. Window and preparation priced together at an average productivity and an average rate. The average understates the window, overstates the preparation, and hides the one number that drives the project.
Crews sized by quantity. The estimate divides the hours by the duration and staffs accordingly, ignoring that the tie-in point only fits four people. The extra crews arrive, stand in each other’s way, and the window slips anyway.
The isolation discovered late. The sequence of isolations, drains and permits gets planned in the week before the outage, and it turns out to consume the first thirty hours of a 240-hour window. It was knowable six months earlier, at estimating time.
No day price. Nobody stated what an overrun day costs, so inside the window every acceleration decision becomes an argument instead of a calculation.
A shutdown estimate is a promise about hours made to a plant that counts them. Split the estimate at the window boundary, price the two projects by their own rules, and put the cost of a day on the front page.
Emerald Group estimates shutdown and turnaround scopes as standalone projects, with the window priced hour by hour and the risk expressed at the day rate. If your next tie-in has an outage in the middle of it, get in touch.