EPiCCircleshared sheet

Interactive learning for EPC. Connecting disciplines.

The Cost-Influence Curve — Why "We'll Optimise It in Detailed Design" Is Already Too Late

The meeting where it is always too late

Week 31 of detailed design. You are looking at a 3D model review and you can see it plainly: the maintenance access to the pump house is blocked by a rack bent, the two spare pumps nobody will ever run are eating 11 m of plot, and the 316L line spec that Process wrote for a "possible future sour case" runs 340 m of 6″ pipe that will never see a chloride.

You raise all three. The project manager says the sentence that this note exists to kill:

"Good catch — we'll optimise it in detailed design."

You are in detailed design. And here is the question: what fraction of this project's total lifecycle cost do you think is still genuinely available to be optimised right now? Guess before reading on. Most engineers say 40–60 %.

The answer is roughly 5–7 %. By the time detailed design opens, about 80 % of lifecycle cost is already committed — while only about 3 % of the money has actually been spent. That gap between committed and spent is the single most useful idea in project engineering, and almost nobody is taught it.

The curve: committed runs far ahead of spent

Two curves live on the same time axis, and confusing them is the whole misconception.

Committed cost   C(t) — money the project is now OBLIGED to spend, because of decisions taken
Expenditure      S(t) — money that has actually LEFT the bank
Influence        I(t) = 100 % − C(t)   (what you can still change without paying to undo)

Typical industry shape (the numbers the interactive uses):

Milestone Committed Spent Influence left
End of concept / FEL-1 40 % 0.5 % 60 %
End of FEED / FEL-3 80 % 3 % 20 %
End of detailed design 93 % 12 % 7 %
Start of construction 97 % 35 % 3 %

Read the FEED row again. The cheque book still looks almost untouched. That is exactly why the illusion survives: everyone confuses "we have barely spent anything" with "we can still change anything." They are not the same statement. The second one stopped being true long before the first.

Watch the two surfaces trade places — the committed landscape is a cliff at the near end, the expenditure landscape is a cliff at the far end, over the same axes: ▶ open the interactive: project cost influence 3d

Why commitment leads expenditure by so much

Commitment is created by decisions that constrain later decisions, and those are all made early because everything downstream needs them as input. Choose 120 % design capacity and you have chosen every vessel diameter, every pump curve, every line size and every kilowatt of utilities that follow — before a single purchase order exists. The PO, months later, is not the decision; it is the settlement of a decision already made.

The practical consequence is that the "cost" of a late change is rarely the new scope. It is the cost of undoing work that was entirely correct against the previous basis: engineering hours already spent, drawings already issued, material already bought to the old spec, foundations already poured, and the schedule float already consumed. A 6″ line that was correct in Rev 0 and correct in Rev 3 is still a full re-run of hydraulics, stress, supports, MTO and isometrics if the basis moves under it.

What actually locks the cost — and when

Not all decisions are equal. These seven commit most of the lifecycle cost, roughly in the order they close:

The 3D landscape plots exactly this: phase × decision area, height = committed. Slide the marker across and you can see that the "cliff" happens at a different place for each row — capacity is already a wall at concept, automation still has a slope at detailed design. That slope is where your remaining leverage lives.

Symbol key — every symbol on this sheet

Where "80 %" comes from, and how much to trust it

The curve is a shape, not a measurement. Its lineage runs through Paul Barshop's and Ed Merrow's work at IPA (Independent Project Analysis) on front-end loading, the CII (Construction Industry Institute) Pre-Project Planning research, and AACE International's cost-estimate classification (Class 5 → Class 1). The consistent empirical finding across those data sets is not the exact number 80 % — it is the correlation: projects with poor front-end definition at sanction overrun cost and schedule far more often than well-defined ones, and the penalty is measured in tens of percent, not single digits.

So quote the shape with confidence and the number with care. "Most of the cost is committed before detailed design starts, and the projects that get the front end right are the ones that land on budget" is defensible everywhere. "Exactly 80 % at exactly the FEED gate" is a teaching figure, and the true value depends on scope, contracting strategy and how much is repeat design.

The 1 : 10 : 100 escalation — worked in money

The classic rule of thumb: a change caught in design costs 1; in procurement/fabrication about 10; in construction/commissioning about 100.

Cost of change  =  base cost at FEED  ×  M(change type, phase)

Worked example — the pump house that was 8 m out of position. Base cost of making this change during FEED: $4,000 (a redrawn plot plan, a re-run of the suction hydraulics, two revised P&ID clouds, 40 engineering hours).

Raised in M Cost Schedule
FEL-3 FEED 1 × $4,000 none
Detailed design 11 × $44,000 ~2.5 days
Construction 105 × $420,000 3 weeks

At construction you are paying for: demolishing two poured foundations, cutting and re-fabricating twelve spools, re-running the stress analysis and the support MTO, re-issuing twelve isometrics, re-routing a cable tray, re-scheduling a crane, and re-testing the system. The change is still 8 m. Everything else on that invoice is undoing work that was correct.

1 : 11 : 105. The rule of thumb is not an exaggeration; if anything it is gentle, because it counts only direct cost and not the delay.

Pick your own change type and phase and watch the multiplier, the schedule hit and the remaining influence move together — the bar chart is on a log scale precisely because a linear one cannot show 0.1 × and 320 × on the same picture: ▶ open the interactive: project cost influence calc

Where a discipline engineer's leverage actually sits

Your leverage is not proportional to your seniority. It is proportional to how early you speak. The same observation, made by the same person, is worth roughly 100 × more in FEED than in construction — and it costs you the same amount of effort either way.

Three things to actually do:

  1. Flag interface risk in FEED, in writing. "Piping cannot route this line without a 4 m loop unless the exchanger moves 2 m" is a $0 sentence in FEED and a $200k sentence in construction. Put it in the FEED deliverable comments, not in a corridor.
  2. Challenge margin and spec class — with a number. "Why 316L on this service?" is ignored. "316L on this service is 340 m of 6″ at roughly 4 × the CS material cost; the sour case it protects against is not in the design basis — can we hold CS with 3 mm CA?" gets a decision. Margins stack multiplicatively: a 10 % process margin on top of a 10 % equipment margin on top of a 10 % line-sizing margin is 33 % of oversized plant nobody chose.
  3. Insist on constructability and maintainability review before IFC. IFC is the point where the cost of change steps from ~10 × to ~100 ×. A two-hour review with a construction superintendent and a maintenance planner before that step is the highest-yield meeting on any project. Ask for lay-down, crane access, bolt clearance, valve reach and tube-bundle pull space explicitly — those are the four things that get discovered on site.

And know the counter-argument so you can answer it: front-end work is not free either. FEED typically costs 2–5 % of TIC, and a gate held open forever has its own cost. The argument is never "spend unlimited time in FEED" — it is "spend the right time there, because the exchange rate is 100 : 1 and it only runs one way."

Common pitfalls

Outcome

Open items

Know why, not just what.

One sheet from EPiC Circle

Interactive engineering lessons across every EPC discipline — calculators, 3D models and misconception-first teaching. Free to read; sign in to track progress, take the quizzes and ask questions.

Open in EPiC Circle →

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis