Most US contractors don’t lose profit in one big blow. They lose it in small, measurable leaks that start weeks before the cost report looks bad. The firms that protect margin don’t “report KPIs”, they run an early-warning system that ties field execution to cash and profit in real time. You’ll leave with a KPI map across the full job lifecycle, clear formulas, and a practical way to run weekly decisions off trusted data instead of Excel and gut feel.
In this article
Profit Leakage, Quantified
Most jobs don’t lose money in one big event. Margin leaks out through repeatable failures that feel “manageable” on site: work done out of sequence, rework that gets logged as progress, changes done before they are agreed, and work earned but not billed. That matters because general contractors often run at 2% to 5% net margin, according to AGC member data. At that level, you don’t get many chances to be wrong before the job stops paying you.
Schedule slippage is the first leak because the cost lands before the cost report explains it. A two-week slip rarely shows up as “two weeks of extra cost”. It shows up as resequencing, overtime, remobilisation, and trade stacking. The early warnings are operational. Track Percent Plan Complete (PPC), constraint removal, and short-interval plan misses weekly. Archdesk sees the same pattern across live projects: plan reliability drops first, then labour efficiency, then cost-to-complete jumps.
Rework is the leak that hides in plain sight. Foreman diaries still show “productive hours” even when crews are ripping out and redoing. Industry benchmarks put rework at 2% to 5% of total cost. On a $15m job, 4% rework is $600k of direct waste before the delay cost. Treat rework as a cost code and measure it, not as a story. Pair it with RFI cycle time and submittal turnaround. Slow answers create wrong work, and that chain repeats on every floor and every zone.
Change order friction is not just about the scope. It is the lag between doing the work and getting it priced, agreed, and billed. One 2023 industry report put the average commercial change order cycle at 30 to 45 days. That lag forces you to fund labour and materials without the client’s money. Run two numbers side by side: change order ageing (days open) and recovery rate (approved value versus submitted). Archdesk supports this by keeping an audit trail from field event to backup, pricing, and approval, so ageing items don’t turn into write-offs at final account.
Billing delays trap cash even on profitable jobs. CFMA’s survey puts the median AR collection period for general contractors at about 55 days. Add retainage, often 5% to 10%, and you are financing someone else’s project. The control metric is unbilled value in Work in Progress (WIP). If AR looks fine but unbilled WIP is climbing, the issue is inside your own application process: missing backup, slow internal sign-off, or progress that isn’t tied to the schedule of values.
Run leading and lagging controls together. Leading controls (PPC, RFI age, submittal turnaround, productivity) move weeks before lagging results (cost-to-complete, gross margin, cash). Weekly review is what protects margin, not a better month-end pack.
Put a simple rule in place on every job. No weekly ops review without a cash view. No cash review without open change order ageing and unbilled WIP. That one discipline forces site, commercial, and finance to work off the same facts. Archdesk is built to keep those facts tied together, so you can act in week three, not explain it in month three.
Winning the Right Work
Win rate is a vanity metric if it fills your backlog with jobs your team can’t deliver at the price you sold. A 2023 FMI study found contractors winning more than 30% of bids were no more profitable than firms winning 1 in 5. The difference was selection, not volume. The operators who protect margin filter every opportunity through three questions: will we win it, will we price it at the right margin, and does it fit how we actually build.
Run your pipeline as a managed funnel, not a list you react to. Use four gates: lead, qualified bid, shortlist, award. Track conversion by sector, delivery route, and client type. The action is in the gaps. If negotiated work converts at 20% and hard-bid converts at 5%, the fix is not “bid more”. The fix is to stop burning senior estimating hours on low-fit work and push earlier qualification onto the business development and ops leads who will live with the job.
Estimating accuracy is the pre-con KPI that stops “good-looking” bids becoming margin fade. A 2024 CFMA benchmarking survey found the median GC’s estimate landed within 3% of final cost on repeat-client projects, but deviated 9% or more on first-time clients. That gap rarely comes from take-off. It comes from not knowing how that client behaves on scope, sign-off, quality thresholds, and closeout. Track estimating accuracy by client, then write clear bid rules like “first-time client needs tighter clarifications and a higher risk allowance”.
Bid margin only matters if delivered margin follows it. The gap between the two is your real risk premium, and it tells you where you are under-pricing complexity or over-trusting the tender information. Make the gap visible by sector and project type, then act fast. If one sector is consistently two points behind what you priced, either add risk pricing, tighten scope and exclusions, or stop chasing that work until delivery has the right team and supply chain.
Archdesk is built for this feedback loop. Your bid funnel, estimate, buyout commitments, and cost-to-complete sit in one dataset, so you can see which sectors and clients deliver what you priced. Practical move for next week: pull your last 20 completed jobs, group them by sector and client, and calculate the bid-to-delivered margin gap. Give your estimators one rule change for the worst segment, or cut that segment from the next quarter’s bid list.
Schedule as a KPI System
Schedule control fails when you treat the programme as a date list instead of a KPI system. Run two schedule lenses together every week. Use Percent Plan Complete (PPC) to test whether crews delivered what they promised. Use Schedule Performance Index (SPI) to test whether what they delivered was enough to stay on plan. These two numbers stop the Monday meeting turning into opinions.
PPC is your weekly reliability score. Calculate it as tasks done divided by tasks promised, times 100. Keep the rule tight. “Done” means installed, checked, and released to the next trade. “Mostly done” is not done. PPC drops first when constraints are not being cleared, even if the master programme still looks fine. Treat PPC below 80% as a planning and constraint problem, not a foreman problem.
SPI is the earned value view of time. Calculate it as Earned Value (EV) divided by Planned Value (PV). EV is the budgeted value of the work you have actually completed. PV is the budgeted value of the work you planned to have completed by now. An SPI of 0.90 means you are earning 90p of progress for every £1 you needed to earn by this point. The key is the earned value method. Lock it down. Use weighted milestones per cost code or measured quantities. Don’t let “percent complete” be a guess.
PPC and SPI often disagree, and that is the point. High PPC with a falling SPI means the team is reliably delivering an underpowered plan. The fix is to replan the remaining work so you earn progress earlier on the critical path. Low PPC with a stable SPI means the team is chasing unplanned work and firefighting. The fix is to remove constraints in the look-ahead and tighten handovers, not to add more overtime.
Production KPIs are where schedule and margin meet. Track labour productivity as installed quantity per man-hour, by crew and workface. Normalise it, or you will argue about numbers instead of fixing the work. The same crew output is not comparable between open access and congested areas above live services. Capture constraint reasons alongside time, so you can separate “we planned badly” from “we were blocked by access, design, approvals, or materials”. This is where a single dataset matters. Archdesk is built to keep time, quantities, cost codes, and constraints tied to the same record, so the KPI discussion starts with facts.
PPC tells you whether the team kept its promises. SPI tells you whether those promises were enough to protect the programme. The fix depends on which one moved.
Practical move for next week: write a one-line definition of done for your top 20 look-ahead tasks, then review PPC and SPI side by side by phase. Raise PPC by clearing constraints before work starts. Raise SPI by replanning the remaining work so earned progress lands earlier on the critical path.
Protecting the Margin
CPI only tells you how efficiently you have spent money that has already hit the ledger. It does not tell you what you have already committed to spend. A job can look “ahead” for weeks, then fall off a cliff at buyout or when late package orders land. Treat commitments as real costs the day you sign the subcontract or PO. Track three lines by cost code: budget, committed, and actual. Add a fourth line for change exposure.
Control margin through control accounts, not one job-level number. Use CSI divisions and the subcodes you estimate and buy out against. That structure lets you separate the three causes of margin fade while there is still time to act. Commitment overrun means procurement or scope gaps in the estimate. Productivity overrun means labour hours or plant are burning faster than planned. Scope growth means you are doing work without a signed change. Blending these into “over budget” leads to the wrong fix and the same mistake on the next job.
| Root cause | What it looks like in the numbers | Fast action that protects margin |
|---|---|---|
| Commitment overrun | Committed cost runs above budget before the work ramps | Re-scope packages, reprice gaps, change means and methods, lock remaining buyout |
| Productivity overrun | Actual cost rises faster than earned value, commitments stay stable | Tighten outputs by gang and area, remove blockers, reset sequence and crew balance |
| Scope growth | Progress and cost move, but the change is not approved or funded | Split the change log into approved, pending, and at-risk. Chase backup and dates weekly |
Change exposure needs board-level visibility because it hides in plain sight. Approved changes are funded. Pending changes are submitted and waiting on sign-off. At-risk changes are identified work with no formal instruction or no priced submission. Track all three. Put “pending plus at-risk” next to your forecast cost at complete. If that number rises on the same cost codes that are slipping, you are funding scope growth with your own working capital.
Decision thresholds stop cost meetings turning into debate. Set them at cost-code level, not at job level. Example rules that work in practice: if committed cost runs above budget before you are halfway complete, treat it as buyout failure and reset the forecast the same week. If CPI drops below 0.95 on a division, require a written recovery plan within 48 hours. If pending plus at-risk changes pass a level you would not accept as a write-off, escalate to the commercial lead and client team immediately.
Margin fade is rarely “one big event”. It starts as a commitment miss, a productivity miss, or unfunded scope, then gets hidden inside a single job number until it is too late to fix.
Archdesk is built to keep these controls in one place. Commitments, actual cost, earned value, and change status sit on the same cost codes, so the PM and the finance lead see the same exposure in the same hour. Practical move for next week: pick three live jobs and create a one-page view by CSI division showing budget, committed, actual, forecast to complete, and pending plus at-risk changes. Then set one escalation rule: any cost code that trips two signals in the same week gets an exec review.
Cash Flow Reality
Cash problems start on site, not in the accounts office. A job can show a healthy gross margin and still choke the business if earned value sits in WIP, billing misses the cut-off, or change work is built before it is agreed. CFMA’s 2023 Financial Survey of the Construction Industry found that 1 in 4 contractor failures involved firms with positive gross margins on their books. The work was “profitable”. The timing wasn’t.
Unbilled value is the quickest way a profitable job turns into a cash drain. Keep it simple: unbilled value equals earned revenue to date minus billed to date. Unbilled value usually spikes for three reasons. Your Schedule of Values (SOV), the list you bill against, does not match how the job is actually built. Your team is conservative on percent complete because backup is weak. Your pay app misses the pencil draw because sign-offs land late. None of those are accounting problems. They are delivery controls. Treat them like programme slippage, because payroll and supplier terms do not wait for month-end.
Accounts receivable needs tracking by owner and by project, not as a blended company number. One slow-paying client on a big package can absorb the cash from several well-run jobs. This gets worse under pay-if-paid clauses, because the delay moves straight down the supply chain and shows up as subcontractor performance issues, claims, and damaged relationships. Your commercial team should know, by name, who stretches payment terms and what it costs you in working capital.
Under-billings and over-billings are where billing mechanics create false confidence. Under-billing means you have done more work than you have invoiced. Over-billing means you have invoiced ahead of earned. Over-billing can keep the lights on, but it is temporary oxygen. It disappears fast when retainage steps up, the front-loaded SOV runs out, or change work sits unapproved. Check the WIP position by cost code, not just job total. A project can look fine overall while one division is deeply under-billed and bleeding cash every week.
Cash-on-cash days is the KPI that connects project controls to the bank balance. Cash-on-cash days equals WIP days plus collection days minus payables days. The operational point is blunt. If you choose to pay trade partners faster to protect labour and programme, you must tighten billing and collections to match. Otherwise you have turned “good supply chain relationships” into a working capital loan you did not price.
Most “cash surprises” are predictable. They show up first as rising unbilled value, older AR against one owner, and change work built ahead of approval.
Archdesk helps by keeping earned value, pay apps, WIP, and collections in one dataset, so the PM, QS, and finance lead are looking at the same numbers. Practical move for next week: run one review across live jobs with your PMs and commercial team. Pick the top three items sitting in unbilled value or disputed AR, assign an owner, and set a seven-day deadline to clear backup, submit, or agree the change. That single habit protects more cash than another month of “cost reporting”.
Quality and Closeout Speed
Closeout drags because most firms treat it as admin, not production. A 2022 Dodge Data study put the average time from substantial completion to final payment release at 67 days on commercial projects. Those are paid people and hired kit sitting on a job that is no longer moving forward. Treat closeout like a phase with its own plan, owners, and weekly targets. You get final account agreed faster and you free your PM and site team to start the next job.
RFI and submittal logs hide the real problem if you only track “turnaround time”. Split the cycle into touch time and queue time. Touch time is your team doing the work. Queue time is it sitting with the designer, client, or main contractor. A 2023 Construction Industry Institute (CII) study found 80% of the RFI cycle was queue time. That matters because queue time forces guesswork and out-of-sequence work on site. Your labour output drops, then you pay to recover it later.
Submittals are the same story, but with a cleaner early warning KPI. First-pass approval rate tells you whether you are sending “buildable” submittals and whether the spec is stable. CII’s 2023 data put top-quartile teams at 78% first-pass. Teams nearer the middle of the pack sat closer to 55%. Every resubmittal is a new procurement and install risk. It pushes decisions into the period where you are trying to finish, test, and hand over.
Rework is still the fastest way to turn a “nearly finished” job into a long closeout. The metric only helps if everyone counts the same way. Set one rule across all projects: rework is installed work that must be removed, replaced, or repaired to meet contract spec. Do not mix in client changes. Track it at trade or system level, not just as one project number. The Construction Owners Association of Alberta reported that firms tracking rework at trade level cut repeat defects by 37% within 12 months.
Days to closeout improves when you stop treating closeout as one bucket. Break it into three tracked sub-phases: punch list clearance, O&M and as-built pack, then waivers and final account. A 2024 AGC survey of mid-market contractors reported 28 days to final payment for firms tracking those sub-phases, versus 54 days for firms tracking closeout as a single line. Archdesk helps by keeping RFIs, submittals, inspections, punch, and commercial status in one place, so closeout blockers show up early and don’t wait for a month-end meeting.
Closeout speed is won during the build. If you wait until substantial completion to chase O&M manuals, approvals, and defect trends, you have already accepted a slow finish and a late final account.
Practical move for next week: publish a one-page “closeout flow” for every live job. Show RFI queue days, first-pass submittal rate, open QA checks tied to payment, and closeout days by the three sub-phases. Then agree one escalation rule, for example any RFI over 10 days in queue gets chased at project director level within 24 hours.
Safety Metrics That Lead
Safety is not a compliance report. It’s a production control. A 2023 CPWR (Center for Construction Research and Training) analysis of OSHA 300 logs found projects with above-average incident rates lost 3.6 times more scheduled working days to disruption than safer projects. Those days don’t just disappear. They turn into broken handovers, re-sequencing, and lost labour output that hits your programme and prelims.
TRIR and DART are still the quickest way to compare sites and packages. TRIR (Total Recordable Incident Rate) is (OSHA recordable cases × 200,000) ÷ hours worked. DART (Days Away, Restricted, or Transfer) is (DART cases × 200,000) ÷ hours worked. The 200,000 figure normalises the rate to 100 full-time workers. The useful insight is the gap between them. High TRIR with low DART points to lots of minor harm, usually housekeeping, access, PPE, or supervision drift. Low TRIR with high DART points to fewer events but higher-energy exposure like falls, struck-by, lifts, and electrical.
Rates only help if your inputs are consistent. Use OSHA 300/300A/301 as the source of truth, and stick to the OSHA recordkeeping rules (29 CFR 1904). The fastest way to wreck your KPI is to log “first aid only” as recordable on one job, then exclude it on another. Decide whether your denominator includes subcontractor hours and keep both views if you need them. Main contractors often need a total-site view. Specialist contractors need a company-hours view for control and pre-qualification.
Company-wide rates hide where you’re actually bleeding time. Segment incidents per 10,000 hours by activity, not by location. Activities travel from job to job, so the fix travels too. “Overhead MEP”, “hot works”, “manual handling”, “MEWP work”, “core drilling”, and “temporary works” give you something you can plan around. Add one site-condition flag such as night shift, live-occupied, weather, or tight access. Repeat spikes in one activity under one condition are a planning and supervision problem, not bad luck.
| Metric | What it tells you | How to run it | Cadence |
|---|---|---|---|
| TRIR | Overall recordable harm rate | Trend by project and trade. Treat an increase like productivity drift. | Monthly |
| DART | Severe disruption risk | Any case triggers a same-day review of access, method, and sequencing. | Monthly, escalate immediately |
| Incidents per 10,000 hours by activity | Where exposure is coming from | Track task type plus one condition flag to spot repeatable hotspots. | Monthly |
| Observation-to-incident ratio | Whether supervisors are controlling risk | NSC (2022) found firms above 50:1 reduced TRIR by 40% over three years. | Weekly |
Leading indicators only work if you manage them like operations. Track a small set weekly: observations per supervisor, near-miss rate per 10,000 hours, and corrective action close-out time. Don’t celebrate “more near-misses” on its own. Watch for repeat themes by activity and anything left open past seven days. That’s where exposure turns into a recordable, then into a programme problem.
Archdesk makes this practical because your labour hours, trade breakdown, and site records sit in one dataset. That means the denominator is real, and your safety rates can sit next to programme and cost on the same dashboard. Practical move for next week: pick one live job and run a 30-minute review with four lines only. Check TRIR trend, DART trend, top three incident activities per 10,000 hours, and actions open over seven days. Give each hotspot an owner and agree one change to method or access before the next weekly look-ahead.
Dashboards: Archdesk Advantage
Dashboards only change job outcomes when the team trusts the numbers. Trust breaks down when cost, progress, and commercial records sit in different places and get patched together at month end. That is when KPI meetings turn into debates about whose spreadsheet is right. Archdesk avoids that by running bidding, buyout, delivery progress, and commercial control in one dataset. The dashboard becomes the output of how the job is run, not a report someone rebuilds after the fact.
Spreadsheets are still doing too much heavy lifting. The 2024 JBKnowledge ConTech Report found that 61% of contractors still use spreadsheets as their primary tool for tracking project financials. Manual reporting always comes with a time lag. That lag costs margin because it delays the only actions that protect a job: fixing productivity, removing constraints, and stopping change work before you have written instruction. Weekly visibility is where you still have options. Monthly visibility is where you mostly have explanations.
Most operators get caught by the “under budget” trap. Low spend often means low production, not good control. Earned value helps because it forces you to put progress and cost on the same scale. A job can look healthy on cost alone, then flip when the missing work lands late as overtime, stacked trades, out-of-sequence installs, and premium freight. Archdesk links progress rules to the budget, so SPI and CPI are calculated from live inputs rather than negotiated in a meeting.
“Single source of truth” only works if each KPI has a clear system of record and a named owner. Precon owns the live pipeline and tender position. Procurement and the PM own commitments, so the committed position updates when the subcontract or PO is issued, not when the invoice lands. Site owns installed progress, based on agreed rules per work package. Finance owns WIP and billing, so unbilled value and retention are visible before cash gets tight. Archdesk works best when those owners update little and often, then review together every week.
| Dashboard signal | What it usually means on site | What to do this week |
|---|---|---|
| SPI < 0.90 with CPI > 1.05 | “Profitable but in trouble”. You are spending less because you are installing less. | Find the top 3 blocked work packages. Remove constraints. Resequence by area. Don’t accelerate disputed scope. |
| Commitments high but actual cost low | You have bought the job, but you can’t build it. Often access, design, or predecessor delays. | Run a short readiness check. Confirm permits, access, materials, inspections, and latest drawings before you add labour. |
| Unbilled value rising | Work is going in, but cash is not coming back. You are funding the client. | Submit the application, sort sign-off blockers, and tighten the rule for what counts as billable progress. |
| Open change exposure climbing while programme slips | You are building change on hope. Time and money get argued at closeout. | Get written instruction or priced agreement before accelerating. Confirm time impact notice is issued. |
Practical move for next week: pick one live job and agree three rules in writing. Keep one cost code structure from estimate through buyout and invoices. Set one definition of percent complete per work package that site and finance both accept. Hold one weekly dashboard review where the PM turns the top three variances into named actions with dates. Archdesk then acts as the referee, so your meeting stays on decisions, not data clean-up.
