Project Management for AEC

1. Start from base rates

Every project begins with a number, and the number is usually optimistic. In a database of more than 16,000 projects, under half came in on budget, 8.5% came in on budget and on time, and half of one percent also delivered the benefits promised. Those figures vary by sector and country; some transport programmes overrun far less. The point isn't that projects are doomed. It's that the first estimate is a forecast, and forecasts should start from what usually happens. CM-RSK-01

The discipline that corrects for this is reference-class forecasting: find completed projects like this one, see how they actually finished, and adjust from there. If your last six healthcare jobs finished 11 to 24% over their bid estimates, a new one priced as if it won't is a bet, not an estimate. CM-RSK-02 · FIN-FPA-01

Three practices make the forecast honest:

• Classify every estimate. An estimate made at 10% design isn't "plus or minus 5%", whatever the cover page says. Use estimate classes and let risk analysis set the accuracy range; never pre-set it. CM-RSK-03
• Model systemic risk separately from risk events. Scope definition and complexity drive early overruns, and they rarely appear on a risk register. CM-RSK-04
• Set contingency at a stated confidence level (for example P50 or P80), hold management reserve separately, and draw contingency down only through change control. CM-RSK-06

Finally, check who benefits from the number. Optimism needs a counterweight, and estimators paid a bonus on winning work are, without any dishonesty, paid to be hopeful. Know your bonding and cash capacity before you bid, and rehearse what happens to the company if this job goes like the last bad one. FIN-CON-01 · FIN-TAL-06 · FIN-CON-05

2. Choose the delivery model and the contract deliberately

Delivery methods are often chosen by habit or fashion. The evidence says to choose by the project. Design-build reliably saves schedule; its cost advantage depends on project type, and on uncertain ground traditional design-bid-build can give better final cost predictability, because lump-sum design-build bids carry large risk premiums. Integrated project delivery shows promising results in small samples and needs mature organizations and shared risk pools. CM-DEL-01 · VCM-DEL-01 · VCM-DEL-02

Where scope is genuinely uncertain, progressive design-build lets the owner and design-builder develop the design, budget and schedule together before the price is fixed, with an off-ramp for both sides. Standard forms now exist for it: AIA A141PDB-2024 in the US and CCDC 32 (2026) in Canada. CM-DEL-06 · VCM-FRM-03

Whichever model you choose:

• Start from your jurisdiction's standard form and control your amendments, rather than drafting from scratch. Know who administers the contract under that form: the architect under AIA, the owner and constructor directly under ConsensusDocs. VCM-FRM-01 · VCM-FRM-02
• Under a guaranteed maximum price, make contingency transparent. Agree what it covers, log every drawdown, and run it through change control the owner can see. That's where trust is won or lost. CM-DEL-05
• Resource the owner's side. Delivery method explains only part of the variance in outcomes; the owner's own capability, and selecting the team on qualifications, matter as much. CM-DEL-04 · BM-ROL-02

At mobilization, get every party to state what they'll deliver, who decides what, and what success looks like. Differences resolved before production cost a fraction of differences discovered during it. #62

A short example shows why the choice matters. A municipal owner wants a new transit maintenance facility on a brownfield site with unknown contamination. A lump-sum design-build bid at concept stage would price the ground risk heavily, or ignore it and fight about it later. Progressive design-build lets the team investigate the site, develop the design to a point where the ground risk is understood, and then agree a price, with the owner keeping the right to walk away and tender the remaining work if the price doesn't hold up. The same owner building a standard bus shelter programme should simply tender it. Same owner, same year, two different answers, because the projects differ. CM-DEL-01 · CM-DEL-06 · CM-RSK-04

3. Plan the information as carefully as the work

A building or bridge outlives its project by decades, and so does its information, or doesn't. On ISO 19650 projects, BIM management means running the information management process, not running the modelling software. Contract to the current 2018 editions of Parts 1 and 2; the 2026 revisions are drafts. BM-STD-01

The process starts with requirements and flows downward. The owner's organizational and asset needs define project requirements, which become the exchange requirements issued at tender and passed down to every appointed party. If the facilities team isn't asked what it needs to operate the building, the model will contain what designers find interesting instead. BM-REQ-01 · FAM-SYS-01

Four practices do most of the work:

• Specify level of information need, not "LOD 400 everywhere". Say what geometry, data and documents are needed, for which purpose, at which milestone. It exists partly to prevent delivering too much. BM-REQ-07
• Write the data requirements as a machine-checkable file (IDS) and run it at every stage gate, so acceptance becomes a test both sides can run rather than an afternoon of looking. BM-QA-01 · BM-QA-04
• Name one authoritative common data environment for each kind of information, and make "shared" a logged step after a check. Most projects run several platforms; say which one is the record. BM-CDE-03 · BM-CDE-02
• Make handover information contractual from the start, and have the operator accept it against the requirements, rather than receiving four hundred PDFs at completion. CM-DIG-02 · FAM-DAT-01 · FAM-DAT-04

Treat the BIM execution plan submitted at tender as evidence of capability, and keep the confirmed plan as the working governance document. The most common failure is never touching it again after signing. BM-REQ-02 · BM-REQ-03

The most useful single step is to put the operator in the room early. Ask the facilities manager or plant engineer which assets they maintain, what data their maintenance system needs for each, and what they had to collect by hand on the last project. Their answers usually fit on two pages and save months of rework at handover. The project manager's job is to make sure those two pages turn into contract requirements, and that someone checks the deliveries against them at each stage, not only at the end. FAM-SYS-01 · BM-QA-04 · FAM-DAT-04

4. Run the work

Production control is where schedules are kept or lost. The Last Planner System replaces "push" schedules with commitments made by the people doing the work: phase pull-planning, a six-week lookahead with a constraint log, and weekly commitments only on tasks whose constraints are cleared. CM-LEAN-01

Track Percent Plan Complete, the share of weekly commitments kept. The industry unknowingly plans at about 50%; mature users aim for 70 to 80%, and projects that finished on schedule showed higher, steadier PPC from early on. Feed the reasons commitments were missed into the risk register: materials, design, prerequisite work and weather are risk signals, not excuses. CM-LEAN-02 · CM-LEAN-03

Coordination needs decision-makers in the room. Weekly or fortnightly federated reviews with people who can approve changes resolve in a meeting what otherwise waits a week. Report clashes as clusters by system and level, give every issue an owner and a date, and record which RFIs and change orders each resolved clash prevented. Where a contractor traced clashes to actual RFIs and change orders, savings were modest, about 0.1% of project cost, but still a 127% return. BM-COO-04 · BM-COO-03 · BM-COO-05 · BM-COO-06

Quality needs controls where mistakes can't be undone. Write an inspection and test plan for each work package, put hold points before grout, concrete or cladding hides the evidence, and define hold, witness and review points in the contract. When something fails, fix the immediate problem and, separately, the cause; and close the corrective action only when you've shown it worked, not when it was implemented. QMS-ITP-01 · QMS-ITP-02 · QMS-ITP-03 · QMS-CAP-01 · QMS-CAP-03

Safety rides on the same signals. Schedule pressure drives risk-taking, so read missed commitments and acceleration requests as safety information. Track leading indicators, such as pre-task plan quality, near-miss reports and how quickly corrective actions close, rather than only recordable injuries. CM-SAF-06 · CM-SAF-02

5. Money, notices and disputes

Cash is a project's bloodstream, and payment law sets its pulse. Under Ontario's Construction Act, for example, an owner must pay a proper invoice within 28 days, must give a notice of non-payment with reasons within 14 days or owe the invoice in full, and contractors must pay subcontractors within 7 days of being paid; adjudication decisions bind in the interim. Other jurisdictions have their own clocks, and the UK has a bill to phase out cash retentions. Know yours. VCM-PAY-01 · FIN-CON-03 · VCM-PAY-02

Most disputes are controllable. In consultancy data, the leading causes are errors and omissions in contract documents and missed contractual notices, both of which come down to discipline. So:

• Keep a live register of every notice deadline: non-payment notices, deficiency notices, claim time bars, adjudication timelines, with an owner and an alert for each. VCM-ADM-01 · VCM-ADM-02
• Use the early-warning machinery your contract gives you. NEC4's early warnings and compensation events, and FIDIC's dispute-avoidance boards, only work if someone runs them every week. VCM-FRM-05
• Don't count claims as revenue until they're enforceable. FIN-CON-02
• Index volatile materials with stated trigger thresholds, and allocate tariff risk in writing. Fixed-price supply contracts leave new tariffs with the seller unless they say otherwise. VCM-PRC-02 · VCM-PRC-01

Measure contract value loss from your own ledger (unrecovered changes, late-payment penalties, dispute costs) rather than quoting an industry "leakage" percentage that its own publisher says can't be validated. VCM-ADM-03 · VCM-EVD-01

A notice register doesn't need software. A shared spreadsheet works if it has five columns: the clause, the trigger event, the deadline in days, the person responsible and the date the clock started. What makes it work is the habit: a ten-minute review at every weekly project meeting, and a rule that anyone who sees a trigger event, such as a late drawing, an instruction on site or an invoice not paid, adds a line the same day. Projects that lose money on notices rarely lack the right; they lack the date. VCM-ADM-01 · VCM-ADM-02 · VCM-PAY-01

6. Decisions under pressure

The decisions that define a project are usually made late in the day, with incomplete information and someone senior wanting a particular answer. A few habits change how those decisions go.

• Diagnose before treating symptoms. When a schedule slips or a defect repeats, find out why before choosing the fix. PM-1
• Run a pre-mortem before committing: imagine it's two years from now and the project went badly, and have everyone write down why. It surfaces risks people were reluctant to raise. CS-03
• Pool what each person knows before anyone argues for a choice. Get the superintendent's, the estimator's and the contracts manager's facts on the table before the most senior person speaks. CS-08
• Build dissent into go/no-go decisions, so someone's job is to make the case against. PM-22
• Don't repeat a statistic you can't source. AEC is full of famous numbers that fail that test. CM-EVD-01
• Give the gate a real option to say no. Write the criteria before the pitch, show the cost as a range, and agree the stop rules at approval, before anyone needs them. OGV-GATE-01 · OGV-BIZ-01 · OGV-ESC-01
• Check the schedule and the percent complete before you report them. Measure progress by what's built, not hours spent, and give dates with a confidence level. PCF-SCH-02 · PCF-EVM-01 · PCF-SCH-04

Picture a familiar Friday. The concrete supplier has warned that Monday's pour may be short of trucks. The superintendent wants to go ahead; the project manager has a client meeting promising the slab; the quality manager hasn't seen the final rebar inspection. A good decision process takes fifteen minutes: each person states the facts they hold before anyone recommends; someone is asked to make the case for postponing; the hold point is checked, not assumed; and the decision and its reasons go in the log. Whichever way it goes, the team can explain it on Monday. CS-08 · PM-22 · QMS-ITP-02

None of these slows a good team down. They're the difference between a decision you can defend afterwards and one you have to explain.

The chain. Forecast from base rates. Classify the estimate and know your capacity. Choose delivery by the project and administer the form you chose. Plan the information from the operator's needs. Control production with commitments and PPC. Put quality controls where mistakes can't be undone. Run a notice register. Pool facts, invite dissent, then decide.