How to Plan a Construction Project

What are the stages of planning a construction project?

Each row is a decision gate. The estimate class listed is the best the available information supports — not the best an estimator can produce with enough time.

What belongs in the program, and why it comes first?

The program is the written statement of what the building has to do. Without it, every later estimate is pricing a moving target, and every variance investigation ends in an argument about what was assumed.

State requirements in units, not adjectives. Bed count, classroom count, stall count, clear height, bay spacing, tons of cooling, dock positions, net-to-gross ratio by function. These are the drivers a conceptual estimate multiplies against. "High-quality finishes" is not a driver; a finish standard by room type is.

Name which leg of the triangle is fixed. Budget, schedule and scope cannot all be fixed. Decide before design starts which one is immovable, because that determines what happens when the first estimate comes in high — value engineering, a phased schedule, or more money. Projects that never make this call make it later under deadline pressure, which is where scope gets cut badly.

Price the project, not just the building. Construction cost is a subset of project cost. Sitework and offsite utilities, permits and impact fees, design and consultant fees, testing and inspections, FF&E, moving and commissioning, insurance, financing carry, and owner's contingency all sit outside a contractor's number. Many disputes about an estimate being "wrong" turn out to be a construction-cost estimate compared against a total-project budget.

How do you test feasibility before there is a design?

Feasibility runs on parametric estimating: you take a cost per unit of capacity from projects you have actually built, adjust it for this project's location, size, complexity and construction timing, and add the project costs that sit outside construction. This sits at the lowest-definition end of the estimate classification system published by AACE International, where the expected accuracy range is widest.

The limitation is worth stating plainly, because it causes more damage than any other number on a project. A cost per square foot carries every assumption of the project it came from: its structural system, its foundation conditions, its MEP density, its finish level, its site constraints, its labor market and the month it was bought. Two warehouses at identical square footage can differ by a wide margin on foundations alone if one has poor soils. When you borrow a rate, borrow it from a project you know intimately, and write down what you adjusted and why.

A feasibility number is defensible when it is presented as a range with its assumptions attached, and indefensible when it is presented as a single figure. The single figure is what gets typed into a pro forma, printed in a board packet, and quoted back to you eighteen months later. Preliminary budget estimating done properly produces a range, an assumption log, and an explicit exclusion list — the exclusions being the part everyone skips and everyone later needs.

What level of estimate does each design stage support?

Design stages exist to convert assumptions into decisions. The estimate that matches each stage measures what has been decided and prices the remainder as an allowance.

Schematic design — estimate by element. Areas, massing and major systems are set; sizes and specifications are not. Organize by building element rather than by trade, following UNIFORMAT II (ASTM E1557): substructure; shell, meaning superstructure, exterior enclosure and roofing; interiors; services; equipment and furnishings; special construction and demolition; building sitework. Pick one level of the classification and hold it across every stage estimate, otherwise the estimates will not reconcile against each other. Element-based structure matters here because it maps directly onto design decisions — a designer can see what the enclosure costs and act on it. A MasterFormat estimate at this stage implies a level of definition that does not exist.

Design development — measure what exists, allow for what does not. Structural grid, envelope assemblies and major equipment are selected. Quantities for those become measurable; interiors, specialties and final MEP distribution usually are not. The honest deliverable is a hybrid: real takeoff quantities where the drawings support them, assemblies or allowances elsewhere, and every allowance labeled as such with its basis written down. DD is generally the last stage where redesign still costs less than the change it avoids.

Construction documents — full quantity takeoff. At 100% CDs the drawings are dimensioned, coordinated and specified, and most of the work can be counted or measured directly. The exceptions are the performance-specified and delegated-design scopes — fire protection under a hydraulically calculated NFPA 13 design, connection design, precast and curtain wall engineering, controls, and any system the spec carries as design-build — which are still priced by system, by area or against a subcontractor budget. This is the first estimate built mainly from measured quantities against current unit pricing, and the tightest number available before bids. It is still a forecast of what the market will price, not a hard number; the hard number arrives with bids and becomes real at contract. Interim CD estimates at 50% and 90% are worth running anyway — they catch coordination costs and specification creep while there is still time to respond.

Why do you reconcile estimates instead of just producing them?

Every new estimate should be reconciled against the previous one, line by line, with each movement explained by a cause: a design change, a quantity correction, a pricing update, a scope addition, or an error. A stack of unreconciled estimates tells you nothing except that the number changed.

Reconciliation is what converts an estimate from a price into a control document. It exposes scope that entered the design without a decision — the classic case being a system upgraded on a drawing sheet that nobody flagged as a cost event. It also exposes the opposite problem: an estimate that came in flat because two large errors canceled each other.

Run the comparison in a structure that survives the design stage change. That usually means coding every estimate line with both an element reference and a work-result reference from schematic design onward, so either view can be produced without re-mapping. The relationship is many-to-many — cast-in-place concrete lands in substructure, superstructure and sitework, and one subcontractor's scope crosses several elements — so a single conversion at design development both loses the element view while designers still need it and fails to reconcile cleanly in either direction. Where a line genuinely splits, split it in the estimate rather than in the crosswalk. Teams that skip this end up unable to compare their own estimates and default to arguing about totals.

Where do contingency, escalation and general conditions fit?

These lines are not padding. Each one covers a specific, nameable exposure, and each behaves differently as the project develops. Collapsing them into one percentage is how projects lose track of what is actually covered.

Design contingency. Covers the work that will exist in the finished documents but is not drawn yet — the hangers, blocking, flashing, access panels and coordination that appear between DD and CD. It should shrink at every stage and reach effectively nothing at 100% CDs. If it has not shrunk, the design has not actually developed.

Construction contingency. Covers what happens after the drawings are complete and inside the contractor's own control: field coordination conflicts, means-and-methods problems, minor quantity growth, subcontractor default, small scope gaps between packages. It is the contractor's own risk allowance and does not shrink with design — it draws down as work is completed and exposures are retired. It does not cover differing or concealed site conditions, and it does not cover unusually severe weather: under standard contract language those generate a change order or a time extension, so the money belongs in owner's contingency. Read the actual contract clause before assuming either way, because that allocation is what determines which contingency gets hit.

Owner's contingency. Covers owner-driven change: added scope, revised requirements, program changes. It sits in the project budget, not the construction estimate, and it should never be visible to the contractor as available money.

Escalation to the construction midpoint. Pricing is current as of the day it is assembled. A project bidding a year out and building over two years is exposed to material and labor movement across that whole window, so escalation is carried to the midpoint of construction, not to the bid date. On long-lead packages — switchgear, elevators, structural steel, curtain wall — the exposure is concentrated and worth pricing separately rather than as a blanket factor.

General conditions and general requirements. Supervision, project management, temporary facilities and utilities, hoisting, cleanup, safety and small tools. These scale with schedule duration more than with contract value, which is why a schedule extension costs money even when no work is added — price them from a staffing plan against a real duration. Bonds and the insurance program behave the opposite way: payment and performance bonds are rated per thousand of contract value on a sliding scale, and GL, excess and builder's risk are rated on value. Carry those below the line with fee and sales tax rather than inside general conditions. Keep the time-driven and the value-driven lines separate, or a schedule change and a scope change will both move the wrong one. Estimates built from a per-unit-of-value assumption rather than a staffing plan and a duration tend to come in low here.

How does the delivery method change the sequence?

Design-bid-build keeps the stages clean and sequential: complete documents, competitive bids, one contract. You get the clearest apples-to-apples comparison, and the sharpest pricing when there is real bidder interest — check coverage before you rely on competition, because in a busy market complete documents buy you fewer bids and more qualifications rather than a lower number. And you get the price late, after the design is finished and expensive to change. This is also the delivery method least forgiving of an incomplete set: where DBB documents have gaps, each bidder prices them differently and the spread becomes unreadable.

Design-build and construction manager at risk pull pricing forward. A CM at risk is on board during design, produces estimates at each stage, and converts to a guaranteed maximum price at a point where the documents are usually incomplete. That trade-off is the whole point and also the whole risk: the GMP is only as good as the qualifications and allowances attached to it, and a GMP set on partial documents with a thin clarification list transfers far less risk than it appears to.

Delivery method also determines when procurement can start. With an at-risk contractor or a design-builder, early packages — foundations, structural steel, long-lead equipment — can be bought while the rest of the design continues, which compresses schedule and locks pricing earlier. Each has a prerequisite: foundations need a completed geotechnical report, final structural loads and a foundation or partial permit from the AHJ, and that permit is often the binding constraint on the schedule saving. Early steel is usually a mill order against estimated tonnage with a reconciliation clause rather than a fixed price, since final tonnage and delegated connection design are still open. And every design change that later touches an awarded package is priced as a change order without competition, so buying early trades market exposure for change-order exposure rather than removing risk. It also means those packages are priced against incomplete documents, so the scope splits between early and later packages have to be written explicitly. Gaps between packages are among the most expensive estimating errors in a phased procurement, because there is no bidder who owns the missing scope and no contingency labeled for it.

What does preconstruction actually produce?

Preconstruction is where the estimate becomes a purchasing plan. The deliverables are specific.

Bid packages and scope splits. The work divided into biddable scopes, usually along MasterFormat lines, with a written scope sheet for each that states what is included, what is excluded and where the boundary sits with adjacent packages. The scope sheet is the part that prevents gaps; the division number alone does not.

Scope gap analysis. A deliberate pass across package boundaries looking for work that no package claims — flashing between trades, backing and blocking, equipment connections, patching, final cleaning, temporary protection. Do this before bids go out, not while leveling them.

Bid leveling. Normalizing returned bids to a common scope so they can be compared. A low bid with three exclusions is not a low bid. Leveling sheets are also the cheapest quality check on your own estimate: if every bidder is above you in one division, your quantity or your unit price is wrong, and the quantity is the first thing to re-check.

Buyout log and allowance conversion. Tracking each awarded scope against its estimate line, and converting allowances to actual values as they are bought. This is what tells you mid-project whether your contingency is healthy or already spent.

Schedule and logistics integration. Sequencing, site access, laydown, crane positions and phasing tied back to the general conditions duration. A schedule developed independently of the estimate produces a general conditions number that does not match the job being built.

Where does project planning most often break down?

Common failure modes, listed roughly in order of impact:

• A feasibility range gets quoted as a single number, then becomes the budget of record before any design exists
• Construction cost is compared against total project budget, with soft costs, FF&E and fees never separately funded
• Estimates are produced at each stage but never reconciled against each other, so nobody can explain what moved or why
• Design contingency stays flat through CDs, meaning the design never actually resolved the work it was covering
• Escalation is carried to the bid date instead of the construction midpoint on a long-duration project
• General conditions are estimated as a factor of contract value rather than from a staffing plan and a real duration
• A GMP is set on partial documents with a short clarification list, so the qualifications do not cover what the drawings left open
• Early procurement packages are issued without written scope boundaries, and the gaps surface after every package is awarded
• Geotechnical work, utility capacity confirmation and permit timelines are deferred past the point where a bad answer can still change the design

Do you need an outside estimator for any of this?

For the early stages, often not. If you build the same building type repeatedly and keep clean cost history, your own completed-project data will usually produce a more reliable conceptual number than an outsider working from published cost data, because it already carries your crews, your subs, your market and your overhead structure — context an outside estimator does not have. Guard it, and keep it current — the value decays fast in a moving market.

The stage where outside help genuinely earns its cost is the one where volume and deadline collide: full quantity takeoff on complete documents, on several projects at once, against a bid date. That work is measurable, checkable and time-bound, which makes it the part that outsources cleanly. Conceptual work depends on judgment about a project nobody has seen yet, which does not.

If you do bring in an outside estimator or takeoff service, the quality of what you get back depends on what you send with the plan set. A complete drawing set with specifications, your scope boundaries, your exclusions, and your own recent unit costs where you have them will produce something you can bid from. A partial set with no specification and no direction produces a quantity list with assumptions embedded in it that you will have to find yourself. Ask what the pricing basis is and what date it reflects, and expect the answer in writing — an estimate that does not state its basis cannot be checked, and one that cannot be checked is not worth carrying into a bid.