# What Is Engineering Cost? Why Bad Data in Your PER Inflates It **By Dan Cumberland** · Published May 27, 2026 · Categories: Business Growth > What is engineering cost? It's the design fee and the estimate engineering produces. Learn why bad data in your PER inflates it, and what AI can't fix. ## What Is Engineering Cost? The Three Meanings Engineering cost is the expense of the engineering and design work on a project— typically 4 to 8% of construction cost\.[2](/blog/blog-what-is-engineering-cost#ref-2) But the term gets used three different ways, and the costliest one is not the design fee\. It's the cost figure engineering produces: the Engineer's Opinion of Probable Cost that every funding decision rides on\.[3](/blog/blog-what-is-engineering-cost#ref-3) When a principal or a funder asks what engineering cost means, they're usually holding one of three definitions: - **The design fee\.** The line item covering drawings, calculations, and engineering labor, with the design phase consuming roughly 5 to 20% of the total budget\.[2](/blog/blog-what-is-engineering-cost#ref-2) - **The cost estimate engineering produces\.** The Engineer's Opinion of Probable Cost \(EOPC\), carried inside a Preliminary Engineering Report\. This is the one that matters most\.[3](/blog/blog-what-is-engineering-cost#ref-3) - **Cost engineering\.** The professional discipline devoted to estimating, controlling, and forecasting project cost across the asset life cycle\.[1](/blog/blog-what-is-engineering-cost#ref-1) Keep this distinction\. The cheapest engineering cost is the design fee\. The most expensive is an inaccurate estimate, because the money hides in the decisions made on a number that turns out to be wrong\. The accuracy of that number is governed by something most explainers skip: the quality of your project data, not the sophistication of your estimating software\. ## What Goes Into Engineering Cost: Direct, Indirect, and Design Fees Engineering cost breaks into two buckets\. Direct costs are traceable to a single project— labor, materials, equipment, and subcontractors\. Indirect costs are shared across many projects: overhead, management, office costs, and the salaries of the estimators and engineers themselves\.[4](/blog/blog-what-is-engineering-cost#ref-4) ```html-table
Direct costs (traceable to one project)Indirect costs (shared across projects)
Labor, materials, equipment, subcontractorsOverhead, project management, office costs, estimator and engineer salaries
``` That second column trips people up: the estimator's own salary is an indirect cost, overhead that gets allocated rather than billed to a single job\. Separating the two cleanly isn't bookkeeping— it's the difference between an estimate you can defend in a funding meeting and one you can't\.[4](/blog/blog-what-is-engineering-cost#ref-4) On top of the direct and indirect split sits the design fee\. For new construction, engineering and design fees commonly run 4 to 8% of construction cost\.[2](/blog/blog-what-is-engineering-cost#ref-2) Estimators build the underlying numbers from [the cost\-data sources estimators should triangulate](/blog/construction-cost-estimator-software/)— bid history, unit\-price databases, and recent project actuals\. One line item matters more than its size suggests: contingency, the cushion an estimate carries to absorb what isn't yet known\. The less mature your project definition, the bigger that cushion has to be\. All of it gets rolled into a single number long before the project is fully defined— a number that lives in a document called the PER\. ## How Engineers State a Project's Cost: the EOPC and the PER Engineers state a project's early cost as an Engineer's Opinion of Probable Cost \(EOPC\)— a construction estimate built on recent bid history, professional judgment, and whatever project information exists at the time\.[3](/blog/blog-what-is-engineering-cost#ref-3) It is a planning tool, not a guaranteed price\. And funding decisions get made on it anyway\. The EOPC rarely travels alone\. It's carried inside a Preliminary Engineering Report \(PER\), the planning document that assesses existing facilities, evaluates options, and recommends the most cost\-effective, safe design\.[5](/blog/blog-what-is-engineering-cost#ref-5) For utility and infrastructure work, a completed PER is required by state and federal funding agencies before they release a dollar\.[5](/blog/blog-what-is-engineering-cost#ref-5) That makes the cost estimate it carries one of the highest\-stakes numbers your firm produces\. A PER's total project cost estimate itemizes far more than construction: - Construction - Land and right\-of\-way - Legal - Engineering - Construction management - Administration - Interest during construction - Equipment - Contingency Each line is its own small estimate, and each inherits the data quality underneath it\. When you tighten [your PER process](/blog/construction-engineer-technology/), you're tightening the inputs to nine estimates at once, not one\. Here's the uncomfortable part\. The EOPC inside an early PER often sits at the rough end of the accuracy scale— frequently an AACE Class 3 estimate carrying a range of roughly −15% to \+20%, sometimes wider\.[3](/blog/blog-what-is-engineering-cost#ref-3) That's the number a funder is about to commit millions against\. ## How Accurate Is an Engineering Cost Estimate? The AACE Classes An early engineering cost estimate can be wildly inaccurate by design\. Under AACE International's Cost Estimate Classification System \(Recommended Practice 18R\-97\), a Class 5 estimate— the least defined— can range from roughly −50% to \+100%, while a fully defined Class 1 estimate tightens to about −3 to −15% low and \+5 to \+20% high\.[6](/blog/blog-what-is-engineering-cost#ref-6) AACE estimate classes run from Class 5 \(lowest scope definition, least accurate\) to Class 1 \(highest scope definition, most accurate\); increasing project definition is what moves an estimate toward Class 1\.[6](/blog/blog-what-is-engineering-cost#ref-6) In plain terms: the estimate gets sharper as the project gets clearer, not as the estimator works harder\. ```html-table
AACE ClassScope definitionTypical accuracy rangeWhen it's used
Class 5Lowest (concept, feasibility)−50% to +100%Earliest planning, screening
Class 3Moderate (budget authorization)−15% to +20%Where many PER estimates land
Class 1Highest (near-complete definition)−3% to −15% / +5% to +20%Bid, tender, final check
``` A −50% to \+100% band isn't a mistake— it's what an estimate built on an immature project definition is honestly worth\. Notice what determines the class\. Not the software\. Not the hours logged\. The maturity of the project definition\. ## The Real Driver of Accuracy: Project\-Definition Maturity, Not Your Software The accuracy of a cost estimate is primarily determined by the maturity of the project definition available to the estimator— not by the effort expended, the time taken, or the software used\.[6](/blog/blog-what-is-engineering-cost#ref-6) That single principle, straight from AACE's own standard, reframes the entire problem\. > The quality of a cost estimate is primarily determined by the maturity level of the project definition available to the estimator\.[6](/blog/blog-what-is-engineering-cost#ref-6) Sit with what that rules out\. Not the new estimating suite, not the extra weekend the team puts in\. The lever that actually moves the number is the quality and completeness of the data going in— scope, site information, historical actuals, and the discipline that keeps them consistent\. Firms keep buying better estimating software to fix a problem the standard says software can't fix: thin, inconsistent project data\. Now the honest caveat\. Data quality is necessary, not sufficient\. Estimators will rightly point out that scope changes, optimism bias, and funding pressure also drive overruns— and they're correct\. Bad data doesn't cause every miss\. What makes it worth your attention is that it's the controllable, internal lever your firm actually owns\. You can't legislate away a client's mid\-project scope change\. You can fix the fact that your last forty bids live in forty different spreadsheets\. You can't read the label from inside the bottle— most firms are too close to their own data to notice how inconsistent it has become\. And the cost of that thin data isn't abstract\. The industry has put a number on it\. ## The Cost of Bad Data in a PER Bad data may have cost the global construction industry $1\.85 trillion in 2020, according to a 2021 survey of more than 3,900 construction professionals by FMI and Autodesk\.[7](/blog/blog-what-is-engineering-cost#ref-7) The same study estimated that decisions made on bad data drove $88\.69 billion in avoidable rework that year— 14% of all rework performed in 2020\.[8](/blog/blog-what-is-engineering-cost#ref-8) These are survey\-based figures from a vendor\-sponsored study, so treat the exact dollars as directional\. The direction is what matters\. How bad is the data underneath these estimates? In that same survey, 30% of respondents said more than half their project data is "bad" and leads to poor decisions more than half the time\.[7](/blog/blog-what-is-engineering-cost#ref-7) When a third of firms are working from data that's wrong more often than not, the EOPC in a PER is being built on something close to a coin flip\. The downstream pattern is well documented\. McKinsey Global Institute found in 2017 that 98% of megaprojects overrun by more than 30%, and 77% run at least 40% late\.[9](/blog/blog-what-is-engineering-cost#ref-9) Zoom out and bad data costs the U\.S\. economy an estimated $3 trillion a year, per Harvard Business Review in 2016[10](/blog/blog-what-is-engineering-cost#ref-10)— construction is one slice of a systemic problem\. **The numbers worth remembering:** $1\.85 trillion in bad\-data cost \(2020\) · $88\.69 billion in rework · 30% of firms with majority\-bad data · 98% of megaprojects over budget\. This is where the cost originates: a wide\-band EOPC built on inconsistent inputs is the moment a future overrun gets locked in\. Some firms break the cycle with [a six\-month data cleanup that unlocked their ERP](/blog/erp-construction/) before touching a new estimating tool\. Faced with numbers like these, the reflex in 2026 is to reach for AI\. That reflex is half\-right\. ## Can AI Fix Inaccurate Cost Estimates? \(You Can't AI Your Way Out\) AI cannot fix an inaccurate cost estimate built on bad data\. Under the "garbage in, garbage out" \(GIGO\) principle, output quality is bounded by input quality— so an AI model fed inconsistent project data produces faster, more confident, equally wrong numbers\.[11](/blog/blog-what-is-engineering-cost#ref-11) The defects are baked in before the model ever runs\. Connect that to the AACE principle and the conclusion is unavoidable\. AI doesn't create project\-definition maturity— it can't visit the site, reconcile the scope, or invent historical actuals you never captured\. So on its own it can't move the accuracy band; it just gets you to the wrong answer sooner\. This is what "AI Can Make Words, But Not Meaning" looks like in estimating: a model can produce a number, but not the project definition that makes it mean something\. That's not an argument against AI\. It's an argument for sequence\. On good data, AI earns its keep: - **What AI does well in service of good data:** flags anomalies in historical bids, helps reconcile records into a consistent set, speeds up takeoff comparison, and surfaces missing fields a human would skim past\. - **What AI can't do on bad data:** manufacture scope clarity, fix inconsistent inputs, or replace the engineer's judgment about what a project actually requires\. AI amplifies the data discipline you already have; it doesn't substitute for it\. The pattern holds outside construction, too\. Fielding Jezreel, a federal grant consultant we worked with, said it plainly about his own field: AI doesn't replace the expert, and neither the human nor the AI is as strong alone\. The work is in the pairing, and the pairing only holds when the inputs are sound\. The takeaway for estimating: [the input layer is the strategy](/blog/3-layer-architecture/)\. Data discipline first, then AI on top\. ## Frequently Asked Questions ### Is engineering cost the same as cost engineering? No\. Engineering cost is the expense of the design work or the cost estimate engineering produces\. Cost engineering is the professional discipline that produces and manages it— estimating, controlling, and forecasting project cost across the life cycle\.[1](/blog/blog-what-is-engineering-cost#ref-1) ### What percentage of construction cost is engineering and design? Design and engineering fees commonly run 4 to 8% of construction cost for new construction, and the design phase as a whole consumes roughly 5 to 20% of the total project budget\.[2](/blog/blog-what-is-engineering-cost#ref-2) Discipline\-specific shares vary by project type and complexity\. ### What is an Engineer's Opinion of Probable Cost? An Engineer's Opinion of Probable Cost \(EOPC\) is a construction cost estimate based on bid history, the engineer's professional judgment, and the project information available at the time\.[3](/blog/blog-what-is-engineering-cost#ref-3) It's a planning tool, not a guaranteed price, and it's typically carried inside a Preliminary Engineering Report\. ### Why are early cost estimates so inaccurate? Early estimates have low project\-definition maturity, so their AACE accuracy bands are wide— a Class 5 estimate can range from −50% to \+100%\.[6](/blog/blog-what-is-engineering-cost#ref-6) Accuracy improves only as scope and data mature, not as the team works harder or buys better software\. ### Can AI make cost estimates more accurate? Only on top of good data\. AI can flag errors and speed up reconciliation, but it can't substitute for missing or inconsistent project data\.[11](/blog/blog-what-is-engineering-cost#ref-11) Garbage in, garbage out applies— a model fed bad inputs produces faster but equally unreliable numbers\.[11](/blog/blog-what-is-engineering-cost#ref-11) ## The Most Expensive Engineering Cost Isn't the Fee The most expensive engineering cost on any project isn't the design fee— it's an inaccurate estimate, and the lever you actually control is the quality of the data behind it\. Engineering cost is three things: the fee, the estimate, and the discipline\. The money hides in the estimate\. AACE's standard is clear that accuracy follows project\-definition maturity, not tooling\. So the work that makes AI useful comes before AI: getting your project data consistent, complete, and trustworthy\. Fix the data discipline first, and AI makes good estimators faster and sharper instead of making bad estimates quicker\. Firms that want to get AI\-ready usually want better estimates\. What they need first is data they can trust\. If that's the gap, it's worth mapping [your data and AI\-readiness before you buy another tool](/service/)— the work before the tool is where the accuracy actually comes from\. ## References 1. AACE International, "What Are Cost Engineering & Total Cost Management" \(2024\) — [https://web\.aacei\.org/about/about\-aace/what\-is\-cost\-engineering](https://web.aacei.org/about/about-aace/what-is-cost-engineering) 2. Monograph, "Structural Engineer Fees: Complete Cost Breakdown" \(2024\) — [https://monograph\.com/blog/structural\-engineer\-fees\-cost\-breakdown](https://monograph.com/blog/structural-engineer-fees-cost-breakdown) 3. KTA\-Tator, "Developing Opinions of Probable Cost" \(2023\) — [https://kta\.com/developing\-opinions\-of\-probable\-cost/](https://kta.com/developing-opinions-of-probable-cost/) 4. Procore, "Indirect Costs in Construction: An Essential Guide" \(2024\) — [https://www\.procore\.com/library/indirect\-costs\-in\-construction](https://www.procore.com/library/indirect-costs-in-construction) 5. USDA Rural Development \(Rural Utilities Service\), "Guidance on Preliminary Engineering Reports \(RUS Bulletin\)" \(2014\) — [https://ndep\.nv\.gov/uploads/water\-financing\-srf\-drinkingwater\-docs/RUS\_Bulletin\_Guidance\_for\_PERs\.pdf](https://ndep.nv.gov/uploads/water-financing-srf-drinkingwater-docs/RUS_Bulletin_Guidance_for_PERs.pdf) 6. AACE International, "Cost Estimate Classification System \(Recommended Practice No\. 18R\-97\)" \(1997\) — [https://services\.austintexas\.gov/edims/document\.cfm?id=280770](https://services.austintexas.gov/edims/document.cfm?id=280770) 7. FMI Corporation & Autodesk, "Study from Autodesk and FMI Finds Better Data Strategies Could Save the Global Construction Industry $1\.85 Trillion" \(2021\) — [https://www\.prnewswire\.com/news\-releases/study\-from\-autodesk\-and\-fmi\-finds\-better\-data\-strategies\-could\-save\-the\-global\-construction\-industry\-1\-85\-trillion\-301376278\.html](https://www.prnewswire.com/news-releases/study-from-autodesk-and-fmi-finds-better-data-strategies-could-save-the-global-construction-industry-1-85-trillion-301376278.html) 8. Autodesk Construction Cloud, "Study from Autodesk and FMI Finds Better Data Strategies Could Save the Global Construction Industry $1\.85 Trillion" \(2021\) — [https://www\.autodesk\.com/blogs/construction/autodesk\-fmi\-study\-global\-construction\-industry\-data\-strategies/](https://www.autodesk.com/blogs/construction/autodesk-fmi-study-global-construction-industry-data-strategies/) 9. McKinsey Global Institute, "Reinventing Construction: A Route to Higher Productivity" \(2017\) — [https://www\.mckinsey\.com/capabilities/operations/our\-insights/reinventing\-construction\-through\-a\-productivity\-revolution](https://www.mckinsey.com/capabilities/operations/our-insights/reinventing-construction-through-a-productivity-revolution) 10. Harvard Business Review \(Thomas C\. Redman\), "Bad Data Costs the U\.S\. $3 Trillion Per Year" \(2016\) — [https://hbr\.org/2016/09/bad\-data\-costs\-the\-u\-s\-3\-trillion\-per\-year](https://hbr.org/2016/09/bad-data-costs-the-u-s-3-trillion-per-year) 11. TechTarget, "What is garbage in, garbage out \(GIGO\)?" \(2023\) — [https://www\.techtarget\.com/searchsoftwarequality/definition/garbage\-in\-garbage\-out](https://www.techtarget.com/searchsoftwarequality/definition/garbage-in-garbage-out) --- Source: https://dancumberlandlabs.com/blog/what-is-engineering-cost/