At sanction, the estimate becomes the control budget. The same spreadsheet now has a different job: not to predict the cost, but to be the fixed reference every change is measured against. Most of the trouble on projects comes from not noticing that the job changed.
Before sanction, an estimate is judged by how close it will land. After sanction, it is judged by how well it explains movement. Those are different virtues, and they are built in different ways. A number can be accurate and useless as a baseline; it can even be slightly wrong and still be an excellent one, provided every part of it is traceable, structured and frozen. The estimator who understands this hands the project a control system. The one who does not hands it a number and a wish.
What a Class 2 has that a Class 3 does not
Under AACE International Recommended Practice 17R-97, Cost Estimate Classification System (RP 17R-97), Class 2 sits at thirty to seventy-five percent definition, and the difference from Class 3 is not polish. It is four structural upgrades.
- Quantities from issued-for-construction or near-IFC drawings for the majority of the direct cost, each with its drawing and revision recorded.
- Firm or near-firm pricing on the major equipment and subcontracts, with validity that survives to award.
- A resourced schedule the indirects are built from, not assumed, so time-driven cost can move when time moves.
- A risk register re-run with construction-stage risks: weather windows, productivity, market response, not the design-stage risks that have already retired.
How the earlier classes mature into this is the story of the whole series: see What actually changes between a Class 5 and a Class 3 estimate.
Built for control, not just accuracy
A control budget has consumers the earlier estimates never had: the cost controller booking commitments against it, the planner claiming progress on it, the change manager pricing variations into it. Serving them is a structural requirement. The estimate rolls up through control accounts the team will actually report against. Every line maps to schedule activities, one to many in both directions, or earned value dies at birth. Commitments can be matched to estimate lines, so the forecast is built from facts instead of feelings. All of that is decided by the structure chosen back in week one, which is why the structure decision outlives everything: see A WBS that survives construction.
Freezing the basis
The control budget needs a frozen basis of estimate, a frozen WBS and a frozen pricing date. Every subsequent change is priced against that frozen trio, in the same structure, so the trend report can say what moved and why. New scope gets new codes in reserved ranges; nothing is repurposed; the pricing date does not drift. The freeze is not bureaucracy. It is what makes the word “change” meaningful: a change is a difference from something, and the something has to hold still.
The reconciliation
We produce a Class 3 to Class 2 reconciliation as a standard deliverable: every movement over a threshold, classified as scope, quantity, rate, method or risk. It is the document that stops the sanction meeting becoming an argument, because the board is not being asked to trust that the new number is right; it is being shown exactly how the old number became this one. The same method that settles a fight between two rival estimates settles the differences between two generations of the same estimate: see Reconciling two estimates that disagree.
The machinery, running
Sanction the illustrative project at 220 million: base estimate 195, escalation 10, contingency 15, each frozen as its own line. Now run nine months of project life through it. The steel index jumps: the escalation line absorbs 3.2 million, attributed to a named index movement, and contingency is untouched because market movement was never its job. The owner adds a building: change control prices it at 4 million against the frozen basis, and the budget grows by an approved change, not by erosion. The ground turns out worse at the river crossing: the risk register’s item lands, and 2.1 million draws down from contingency against that named item, with 12.9 remaining against the risks still open.
The forecast now reads 229.3 million, and every dollar of movement has a name, a cause and a category. Compare the alternative, a blended budget with no frozen lines: the same nine months produce “we are nine over and the team is investigating”, a sentence that has ended careers without ever explaining anything. Same project, same events, same money. The difference is entirely in whether the estimate was built to remember.
After sanction, the estimator’s job is memory.
Where this goes wrong
The unfrozen baseline. The estimate is approved but the basis keeps drifting: quantities quietly updated, the pricing date sliding, exclusions renegotiated in meetings. Six months in, variance reports compare the project to a reference that no longer exists, and every number in them is an opinion.
The team that leaves at sanction. The estimators demobilise the week the number is approved, and the controls team inherits a spreadsheet without its reasons. The first trend meeting discovers that nobody in the room knows what is inside the indirects line. The handover deserves weeks, not an afternoon, and the basis of estimate is the handover’s spine.
The rolling rebaseline. Every quarter the “current budget” is restated to absorb the bad news, so every report shows the project roughly on plan. The trend history is destroyed, the early warnings with it, and the day of reckoning arrives at the end, undiluted, with interest.
Contingency as the change fund. Scope changes get paid from contingency because it is nearer than the change process. By midway, the fund priced against the risk register is gone, the risks are still open, and the project meets its actual exposures unarmed. What contingency is, and is not, is its own subject: see Contingency is not padding.
The estimator’s last deliverable
The Class 2 is the estimating team’s final major product, and its quality is measured for years by people who never met the estimators: in every trend report that either explains itself or does not, every change priced cleanly against a frozen basis or argued about, every forecast built on structure or on hope. Build the baseline so the controls team can run it, reconcile it to its predecessor, write the memory down, and hand it over properly. That is what the project bought.
Emerald Group develops Class 2 and control baseline estimates inside project teams and to their procedures, carries them across phase gates with the audit trail intact, and reconciles them to what came before. If your project is heading to sanction and the number needs to become a baseline, get in touch.