Evaluate · FID and stage-gates
Which projects deserve capital, and what is waiting worth?
Evaluate prices every option, size and timing as a range for capex, schedule and NPV, on the same scenarios as the rest of your book. Plants that compete for the same feed, plot or grid connection are solved together, so the recommendation is a portfolio you could actually build.
Output: P10, P50 and P80 capex, schedule and NPV for each option and size; the value of deferring, staging or abandoning; the market shift that would change the decision; and a stage-gate pack routed for approval.
Who it is for
For the people who take a project to sanction
Corporate planning, project development, the CFO office and deal teams preparing an FID. Evaluate turns the single-point case into a range, and the ranked list into a portfolio that respects what the projects share.
- Corporate planning
- Project development
- CFO office
- Deal teams
- Stage-gate reviewers
Questions it answers
- Which of these options should we build, and at what size?
- Is building now worth more than staging, deferring a year or walking away?
- Which projects compete for the same feedstock, plot, interconnect or crews?
- How bad is the bad year for this portfolio, not just for each project?
- Which market move would reverse the case, and how far away is it?
In the product
One unit, three ways to build it, on the same scenarios
The Evaluate screen for the polypropylene unit in the illustrative coastal refiner workspace, the same one the home page walks through. Full build, staged and deferred are priced side by side, and the chosen option goes to the committee with its range.
- P80 next to the plan dateThe finish and capex that hold in eight futures out of ten sit beside the plan, so the committee sizes against what usually happens.
- Options priced, not arguedFull build, staged and deferred run on the same 4,000 scenarios as the rest of the book, so their values can be compared directly.
- Sent where money is decidedThe chosen option joins the capital plan with its range. The approved budget and its phasing go back to the systems that run the project.
What Capibud produces
A sanction case with its range attached
Every number comes from the simulation and optimisation engines. Language models, if you switch them on, only draft text, and any figure they write must match the recorded run.
Ranges per option
P10, P50 and P80 capex, schedule and NPV for every option and size, with overrun priors from a reference class rather than a contingency percentage.
Timing as an option
Build now, stage, defer one or two years or abandon, each priced on the same scenarios, including what waiting reveals about uncertain markets.
A buildable portfolio
Shared resources, exclusions, sequencing and synergies solved together, so the plan never needs more feed, land or grid capacity than exists.
The bad year
The average outcome in the worst 5% of scenarios for the whole portfolio, beside the expected value, so risk appetite is a choice and not an afterthought.
What would change our mind
The size of market shift, in standard deviations and in native units, at which a project's case reverses.
A routed stage-gate pack
The recommendation, the alternatives and the run behind them, routed through your delegation of authority and kept in the record.
Worked example
Three growth options, one interconnect
Illustrative demo — not client dataA synthetic data-centre developer on the US Gulf Coast has $2.6bn and a 600 MW grid interconnect. The candidates: a 200 MW hyperscale campus, 400 MW of battery storage and a 250 MW solar project with a power purchase agreement.
Ranking by NPV or IRR picks the campus and the battery, which needs 660 MW of interconnect that does not exist. Funding everything above the 9% hurdle needs 910 MW and $2.8bn. Capibud recommends the campus and the solar project, and leaves $290M unspent on purpose: the next unit is too lumpy to clear the risk-adjusted bar.
The same use case on a refiner's seven candidate units is in the coastal refiner example, which uses invented figures.
| Method | Selection | E[NPV] | P10 | Bad yearCVaR 95 | Feasible |
|---|---|---|---|---|---|
| Capibudmean–bad-year optimiser | Campus + solar | $394M | $91M | −$91M | Yes |
| Rank by NPV or IRR | Campus + storage | $220M | −$102M | −$304M | 660 / 600 MW |
| Everything above hurdle | All three | n/a | n/a | n/a | 910 MW · $2.8bn |
Illustrative demo — not client data. Synthetic workspace; one seeded engine run. Infeasible methods are scored as their consumers would run under the real constraint.
Bars scale to 10%. The campus's P50 IRR is 9.9% against a 9% hurdle; execution slippage costs about $116M across the recommended set.
How it works
From a feasibility workbook to a routed decision
What it connects to
A sanction is never decided alone
Evaluate shares scenarios with the other decisions, so a new plant and a recovering project exposed to the same contractor or commodity move together.
Pooled with the rest
Options join venture tickets, recovery moves and capex lines on one envelope.
The capital committeeSanction becomes baseline
The approved range becomes the reference that construction actuals are tracked against.
ExecuteTripwires on the case
The thresholds that would reverse the case are watched against live prices.
The brainReplayable sanction
Inputs, parameters, seed and engine version are stored with the approval.
The recordHonest limits
What Evaluate does not do
A range is only as good as the reference class and the parameters behind it. These are the limits we tell every design partner about.
- Your engineering estimate is the base. Capibud puts a range on your estimate and its risks; the Class 3 estimate and the schedule come from your engineering team.
- Pack parameters are estimates until you replace them. Industry packs are sourced and labelled, but your own history should override them wherever it exists.
- Files today, connectors on your credentials. Schedule exports (XER, XML), ERP extracts, Excel and CSV are read today; API connectors switch on with your own credentials.
- Calibration evidence is still thin. Our published forecast tests are a synthetic backtest and a market-data test; the megaproject replay is shown after sign-in. Nothing has yet been back-tested on a customer's own history.
Related
See it worked, by industry
Two case studies use Evaluate end to end. The industry packs below are where most sanction cases start.
A data-centre developer: what to build, then keeping phase 1 on track
Three options on one interconnect, followed into construction.
Read the caseA coastal refiner: one capital book
Seven candidate units, a supply assumption replaced, and a case that now reverses at 0.6 standard deviations.
Read the exampleBring us the sanction you are preparing.
A Decision Sprint runs one real FID or stage-gate on your own files in six weeks.