5 Construction Estimating Inputs You Aren't Checking

Why your estimating software for construction looks accurate when it isn't

Your estimating software for construction is producing accurate-looking bids from inputs that decayed eighteen months ago. Win rate looks steady, the framework looks right, and the margin keeps shrinking— because the software is doing exactly what you told it to with data that no longer reflects the market.

The five categories of inputs feeding RSMeans, Sage, HeavyBid, ProEst, STACK, or whatever platform sits in your preconstruction stack all decay on predictable timelines. Most firms catch the failure only after several margin-compressed quarters. The software isn't lying. Your unit costs are.

Your software is fine; your inputs are rotting. Fixing this is the operator's diagnostic for AI-augmented preconstruction— a quarterly hygiene practice that audits the data feeding the system, not the system itself.

AGC reported in September 2025 that 43% of contractors had projects canceled or delayed² amid material moves the database couldn't track, and labor cost errors¹ are typically where profit drift originates. The pattern is structural.

Three forces compressing the gap between estimate and actual cost

Three forces converged in 2025 to widen the gap between input data and current cost: tariff-driven material moves, post-pandemic productivity volatility, and the retirement of senior estimators who used to calibrate inputs from memory. Each accelerates input drift. Together they make the previous standard of "quarterly cost-database refresh" inadequate.

AGC's September 10, 2025 PPI release² documented aluminum mill shapes up 22.8% year-over-year, steel mill products up 13.1%, and lumber and plywood up 4.8% by August 2025— moves that arrived faster than any quarterly database refreshes. NAHB confirmed in early 2026³ that elevated building material price growth has persisted despite a sluggish market.

BLS construction labor productivity data⁴ tells a similar story. Aggregate productivity rose 1.9% in 2024, but that headline hides what's happening in any individual subsector. Industrial building construction grew 16.0%; multifamily fell 12.8%. Aggregate numbers reassure leadership about a problem they're already inside. You can't read the label from inside the bottle.

The third force is quieter. Senior estimators who used to compensate for stale inputs by feel— pattern-matching against a thousand prior bids— are retiring, and the data they carried in their heads isn't being captured. When that hedge erodes, input data quality becomes the only defense.

Quarterly cost-database updates are standard practice. They cannot keep pace with tariff-driven material moves that happen monthly.

Five specific input categories carry most of the drift. They decay on different timelines and produce different symptoms.

Crew productivity rates: the silent labor margin killer

Crew productivity rates— output per crew-hour by trade and project type— are volatile, sector-specific, and the input most likely to be carried forward unchanged from last year. U.S. construction labor productivity rose 1.9% overall in 2024⁴, but multifamily construction productivity fell 12.8% in the same year and road, highway, and bridge productivity has fallen 22% since 2017.

The aggregate number is misleading. BLS data through 2024 shows productivity rising in 2020–2021, declining in 2022–2023, and recovering in 2024. Estimators who copied last year's rate without sector-specific recalibration carried 2023's contraction into 2025 estimates that needed 2024's recovery.

Source: BLS Construction Labor Productivity Highlights, September 24, 2025⁴

Q4 2025 BLS data⁵ confirms the pattern: nonfarm productivity grew 1.8% while unit labor costs grew 4.4%. Labor costs run ahead of productivity gains, which means a bid book using last year's rate structurally underestimates labor cost on every job.

Aggregate construction productivity rose 1.9% in 2024, but multifamily fell 12.8%. Relying on industry averages for trade-specific bids guarantees drift.

Audit signal: Compare the last three to five awarded jobs in the relevant trade against actual labor hours at job close. If estimated hours track within five percent of actuals, productivity inputs are calibrated. Off by more and the rate carried forward is wrong.

Refresh cadence: Annually, with post-job validation. Drift cost: Margin compression on labor-heavy trades. Lost work, sometimes at unprofitable rates, with no obvious cause.

Fully-burdened labor rate: where workers' comp and EMR quietly reset every year

Labor burden in construction adds 30% to 60% on top of base wage⁶, with workers' compensation insurance contributing 10% to 30% or more depending on trade classification and your firm's experience modification rate. Both components reset annually— workers' comp at the policy renewal, EMR based on three years of claim history— and the inputs in your software almost certainly haven't tracked the reset.

Workers' comp is the largest single line item in burden⁷. Procore recommends configuring burden by worker classification and updating quarterly or whenever insurance, benefits, or tax rates change. Most firms set the rate at fiscal-year-start and forget it until the next renewal.

The drift compounds across the bid book. Every labor line in every estimate inherits the error. A burden rate two percentage points too low, across a year of labor-heavy bids, is the difference between gross margin holding and slipping. The slip looks like "tougher market," not "stale data."

The component breakdown matters because the components drift on different cadences:

• Workers' compensation: 10–30%+ of base wage, resets at policy renewal and EMR recalculation
• Payroll taxes (FICA, FUTA, SUTA): Federal stable; state SUTA can move annually
• Paid leave and benefits: Resets with benefits-renewal cycle, often staggered from workers' comp
• Retirement contributions: Resets at plan-year-start; varies if firm changed match policy
• Davis-Bacon prevailing wages: Continuously updated by county per construction type by DOL⁸ for federally funded work

Workers' compensation insurance is often the single largest component of labor burden. An EMR shift of fifteen points moves the bid book by fifteen points on the largest line.

Audit signal: Pull the most recent insurance renewal and reconcile EMR plus workers' comp factor against the rate baked into the estimating software. Mismatch is the diagnostic.

Refresh cadence: Annually plus at every insurance renewal and union contract reset. Drift cost: Structural under-recovery across the bid book; every job is mispriced in the same direction.

Labor inputs drift on annual cycles. Material inputs in 2025 drifted monthly.

Material unit prices: tariff-driven moves your quarterly database can't track

Construction material prices accelerated sharply through 2025 as new tariffs hit specific commodity categories. By August 2025, aluminum mill shapes were up 22.8% year-over-year, steel mill products up 13.1%, and lumber and plywood up 4.8%²— moves that quarterly cost-database updates simply cannot track.

Source: AGC analysis of BLS PPI data, September 10, 2025²

AGC reported the same month that 43% of contractors had projects canceled or delayed due to tariff-driven cost increases, and roughly 40% raised prices in response. AGC Chief Economist Ken Simonson attributed the moves bluntly²:

"Tariff increases enabled domestic producers to push up their selling prices."

The lag is structural. RSMeans⁹— the cost database underlying much of the estimating software for construction— delivers quarterly updates across 85,000+ unit prices and 970+ North American locations. AGC's PPI is monthly. When tariff-exposed commodities move monthly, a quarterly database loads data that's already three months stale by the time a contractor pulls a unit price.

ENR's Construction Cost Index¹⁰ adds a spot-check tool: 25 cwt fabricated structural steel, 1.128 tons Portland cement, and 1,088 board feet of 2x4 lumber priced monthly across 20 U.S. cities. Useful for calibration; not a primary truth.

Steel and aluminum get priced and procured early. A fixed-price contract bid in February using December's database, on a steel-heavy project, locked in a price that lost ten to fifteen percent of its margin before the first delivery.

Aluminum mill shapes rose 22.8% year-over-year by August 2025. Quarterly RSMeans updates were already lagging the market by a full quarter when contractors loaded their fall bids.

Audit signal: Spot-check three to five line items (steel, aluminum, lumber, copper, electrical) against current AGC or BLS PPI before any major bid.

Refresh cadence: Tariff-exposed materials: monthly. All others: quarterly. Drift cost: Direct cost exposure on bid; severe for front-loaded materials in fixed-price contracts.

Equipment ownership and operating rates: diesel, depreciation, and the rental-vs-owned drift

Diesel fuel is more than 35% of construction equipment operational cost¹¹, so equipment unit rates that don't reflect current diesel pricing produce systematically low estimates. Ownership cost per hour also drifts with utilization rates and depreciation schedules— neither of which most estimating software updates automatically.

Equipment cost lives in six components. Each drifts on its own clock:

• Depreciation: Resets when the firm changes its depreciation policy or when the asset's residual value moves with the used-equipment market
• Fuel: Tracks diesel pricing— quarterly at minimum, monthly during commodity volatility
• Maintenance and repairs: Drifts with parts and labor inflation
• Insurance: Annual reset
• Storage and yard cost: Stable until lease renewal
• Rental rates: Quarterly minimum; rental market shifts faster than owned-asset cost

Owned versus rented cost structures differ materially. Owned equipment cost per hour is a function of utilization— a low-utilization year drives per-hour cost up because fixed costs spread across fewer hours. When utilization drops twenty percent, the rate the software is using is twenty percent too low. AnTerra's guidance¹¹ is to update equipment cost data quarterly or after major fuel, materials, or labor changes. Most firms don't.

Diesel fuel is more than 35% of construction equipment operational cost. An equipment line item that doesn't track fuel doesn't track the bid.

Audit signal: Recompute ownership cost-per-hour against actual twelve-month utilization. Spot-check fuel against current pricing for major equipment classes.

Refresh cadence: Quarterly for fuel and operating cost; annually for ownership cost-per-hour. Drift cost: Bid-level mispricing on equipment-heavy site work. Lost work to competitors who priced fuel correctly.

Subcontractor markup: stale quotes, capacity-based pricing, and the broker GC's exposure

Subcontractor markup drifts faster than most GCs realize, especially where sub capacity is constrained. A markup correct on a similar project six months ago may understate current sub pricing by ten to fifteen percent— particularly when subs are pricing capacity, not just cost.

The drift triggers worth tracking:

• Trade mix shifts: Mechanical and electrical capacity tightens faster than general carpentry
• Local market rates: Sub rates are hyper-local; a national average misses your bidding market
• Schedule risk pricing: When subs see compressed schedules, they price the risk; historical markups didn't
• Capacity-based pricing: When books are full, subs price at a premium, and historical data understates the bid
• Out-of-market subs: Travel and per-diem add to bids that historical local-only data won't anticipate

GCs that self-perform major trades carry less exposure— internal labor and equipment data is the cost truth. Broker GCs face the opposite: sub pricing IS the cost data, and stale markup assumptions expose the bid book structurally.

A quote from six months ago is already a "historical" price in volatile markets. Some AI tools target this gap— Bidi Contracting trains its estimating engine on real subcontractor bids from a network of 2,000+ subs nationwide¹². Whether that helps depends on the trades and markets where your subs operate.

Subcontractor markup is the input most likely to come from outside your firm. And the one your software can't refresh on its own.

Audit signal: At every major bid, request fresh quotes from at least two trades whose historical data is more than three months old.

Refresh cadence: Per-bid for major trades; annually for minor trades. Drift cost: Bid-level mispricing. Broker GCs face systemic margin exposure when capacity tightens.

A 30-day audit of your estimating software's inputs (no new software required)

A 30-day input audit can be run by a chief estimator and a project controller without buying any new software for construction estimating. The structure is one input per week, with a one-page reconciliation document for each: current value in the software, current market value or actuals, gap, and proposed adjustment.

The whole exercise should take a chief estimator no more than ten hours over four weeks. The output is the diagnostic, not a software change.

1. Week 1— Crew productivity. Pull the last three to five awarded jobs per major trade. Reconcile estimated labor hours against actual at job close. Per CrewCost's guidance¹³, historical data only works when the project management team validated all costs at job close; if validation didn't happen, the audit signal is to fix that capture process first.

1. Week 2— Labor burden. Reconcile current EMR, workers' comp factor, payroll tax, and benefits against rates baked into the software. Procore⁷ recommends quarterly updates at minimum. If the last update predates the last insurance renewal, the gap is the audit finding.

1. Week 3— Material unit prices. Spot-check five commodity line items (steel, aluminum, lumber, copper, electrical components) against current AGC or BLS PPI. Tariff-exposed materials get extra scrutiny.

1. Week 4— Equipment and sub markup. Recompute ownership cost-per-hour against actual twelve-month utilization. Request fresh sub quotes for two major trades and compare to historical benchmarks. AnTerra¹¹ recommends quarterly equipment-cost refreshes.

The output is a one-page reconciliation per input, signed by the chief estimator. This is how AEC leaders sequence diagnostic decisions: simple, auditable artifacts that survive turnover.

A fifth artifact is the cadence document itself. This becomes a quarterly hygiene practice, not a one-time project. Commercial databases like RSMeans don't account for how a specific firm builds, the skill of its workforce, or its local market¹³. That gap is permanent and your audit— not the database— closes it.

Does AI fix outdated estimating inputs?

AI helps for two of the five inputs and not for the other three. Tools that pull live market data can refresh material unit prices and sub pricing automatically. Tools optimized for historical data still depend on someone curating that data; AI doesn't fix inputs that aren't being captured.

The defensible numbers on AI-assisted estimating software for construction are conservative. A peer-reviewed study cited by Autodesk¹⁴ documented a 20.4% improvement in estimate accuracy and 51.3% faster completion— and was explicit that the gains came when underlying input data was current. McKinsey's analysis¹⁵ put the upper bound near 20% productivity, 15% cost reduction, and 30% delivery improvement.

Vendor claims of "97% accuracy" or "310% bid win-rate improvement" do not survive a peer-reviewed standard. They show up in marketing, not in the literature. Treat them as one of the hidden costs of AI projects— the cost of believing a number that wasn't measured the way it was reported.

AI accelerates the audit. It doesn't replace the audit.

The honest framing: AI here is intellectual augmentation. It runs the audit faster against the inputs you curate. The audit logic is still human, and the data quality still depends on whether your project management team validated job-close costs.

From one-time audit to quarterly hygiene

Run this audit once and you'll find the drift. Run it quarterly and you'll prevent it. The firms that hold margin in volatile markets aren't the ones with the best software. They're the ones whose inputs reflect the market they're bidding into right now.

Most AEC firms can run this internally. Some benefit from an outside framework when the practice is being institutionalized— particularly when senior estimators are retiring and tribal knowledge needs to be documented before it leaves. That's where deciding between an AI consultant and in-house build becomes a real decision rather than a vendor pitch.

If your firm is institutionalizing the input-audit practice, or integrating AI tooling where it adds value rather than where vendors claim it does, Dan Cumberland Labs designs the framework with AEC leadership and helps separate the AI capabilities that matter from the ones that don't.

FAQ: Auditing your estimating software's inputs

How often should construction estimating cost data be updated?

Tariff-exposed materials monthly; other materials quarterly; labor burden annually plus at insurance renewal; crew productivity annually with post-job validation; subcontractor pricing per-bid for major trades. Quarterly is the right cadence for the database overall, but volatile inputs need tighter cycles. AGC², Procore⁷, and AnTerra¹¹ converge on this layered cadence for estimating software for construction.

What is the typical labor burden percentage in construction?

Labor burden adds 30% to 60% on top of base wage for most contractors⁶, with workers' compensation insurance contributing 10% to 30% or more depending on trade classification and EMR. Procore's guidance⁷ is to configure burden by classification and update quarterly.

Why are construction material prices changing so fast in 2025–2026?

New tariffs on imported aluminum, steel, lumber, and electrical components, combined with tighter inventories, drove aluminum mill shapes up 22.8% and steel mill products up 13.1% year-over-year by August 2025². AGC reported 43% of contractors had projects canceled or delayed as a result. NAHB confirmed in early 2026³ that the elevated price growth has persisted.

What's the difference between RSMeans and your firm's own cost database?

RSMeans⁹ provides location-adjusted national average data, refreshed quarterly across 85,000+ unit prices and 970+ locations. Your firm's historical bid data is specific to your crews, methods, and local market— but only useful when your project management team has validated all costs at job close¹³.

Does AI fix outdated estimating inputs?

Partially. AI tools that pull live market data can refresh material and sub pricing automatically. They don't refresh firm-specific inputs (crew productivity, labor burden, equipment utilization), which still require manual audit. A peer-reviewed study cited by Autodesk¹⁴ documented a 20.4% improvement in estimate accuracy and 51.3% faster completion when firms adopted AI-assisted workflows— but only when underlying inputs were current. McKinsey's analysis¹⁵ suggests up to 20% productivity gains. Bidi Contracting¹² is one example of an AI tool that fixes a specific input by pulling live sub bids; it doesn't fix inputs your team isn't capturing.

References

1. McCormick Systems, "Most Common Construction Estimating Mistakes (and How to Fix Them)" (2024)— https://www.mccormicksys.com/blog/most-common-construction-estimating-mistakes-and-how-to-fix-them/
2. Associated General Contractors of America, "Construction Material Costs Continue to Accelerate in August Amid 'Extreme' Price Hikes for Steel, Aluminum and Lumber After New Tariffs" (September 10, 2025)— https://www.agc.org/news/2025/09/10/construction-material-costs-continue-accelerate-august-amid-extreme-price-hikes-steel-aluminum-and
3. National Association of Home Builders, "Building Material Price Growth Remains Elevated Despite a Sluggish Market" (January 2026)— https://www.nahb.org/blog/2026/01/building-material-price-growth
4. U.S. Bureau of Labor Statistics, "Construction Labor Productivity Highlights" (September 24, 2025)— https://www.bls.gov/productivity/highlights/construction-labor-productivity.htm
5. U.S. Bureau of Labor Statistics, "Productivity and Costs, Fourth Quarter and Annual Averages 2025" (February 6, 2026)— https://www.bls.gov/news.release/pdf/prod2.pdf
6. SmartBarrel, "Labor Burden in Construction: What It Is & How to Calculate It" (2025)— https://smartbarrel.io/blog/labor-burden-in-construction/
7. Procore Technologies, "How to Determine Your Fully Burdened Labor Rate in Construction" (2024)— https://www.procore.com/library/fully-burdened-labor-rate
8. U.S. Department of Labor, Wage and Hour Division, "Davis-Bacon Wage Determinations" (2024)— https://www.dol.gov/agencies/whd/government-contracts/prevailing-wage-resource-book/db-wage-determinations
9. Gordian / RSMeans, "RSMeans Data Services" (2026)— https://www.gordian.com/products/rsmeans-data-services/
10. Engineering News-Record, "How To Use ENR's Cost Indexes" (2024)— https://www.enr.com/economics/faq
11. AnTerra Technology, "Estimating Construction Equipment Costs: Key Factors & Tips" (2024)— https://anterratech.com/blog/estimate-construction-equipment-costs/
12. Bidi Contracting, "AI Construction Estimating: How It Works and Why GCs Are Switching (2026)" (2026)— https://www.bidicontracting.com/blog/ai-construction-estimating-guide
13. CrewCost, "Leveraging Historical Data: The Basics of Construction Cost Databases" (May 22, 2024)— https://crewcost.com/blog/leveraging-historical-datathe-basics-of-construction-cost-databases
14. Autodesk Construction Cloud, "How AI and Automation Are Supercharging Construction Estimating" (2026)— https://www.autodesk.com/blogs/construction/ai-estimating/
15. McKinsey & Company, "Artificial Intelligence: Construction Technology's Next Frontier" (2018)— https://www.mckinsey.com/capabilities/operations/our-insights/artificial-intelligence-construction-technologys-next-frontier