Every number we use, and where it comes from.

Vendor pages are full of statistics nobody can trace. This page is the opposite: each figure we cite anywhere, with its primary source, its year, and its limitations. If a number is our own derivation, we say so. If a famous figure did not survive verification, it is not here, and not on our site.

Last reviewed: September 2026.

What a coating scope is built on

  1. Two professionals grading the same steel can disagree by more than 100 percent. Where multiple inspectors assessed identical elements in field trials, the coefficient of variation was "typically greater than 50%, and often greater than 100%." Source: Washer, Hammed, Jensen and Connor, "Quality of Element-Level Bridge Inspection Data," Transportation Research Record, 2674(2), 2020. Scope we state rather than hide: this is bridge element condition data, not coating rust grading. No equivalent study exists for coating grading specifically, and we publish that gap rather than paper over it.
  2. The subjectivity is not disputed inside the discipline either. Vague coating rating criteria mean "there could be considerable variation in the ratings among various bridge inspectors." Source: Bowman, Hagan and Hurdle, Steel Bridge Coating Evaluation and Rating Criteria, FHWA/IN/JTRP-2022/23, Purdue Joint Transportation Research Program for INDOT, 2022.
  3. Maintenance painting costs have risen "200% to 500% over the past decade" on containment and environmental compliance requirements alone. Source: same Purdue JTRP report for INDOT, 2022.
  4. The parties who produce the estimate are not neutral, and the reason is structural. On estimating the extent of remedial coating areas, the paint supplier and the contractor "both have a vested interest to estimate higher rather than lower." Source: IMIA (International Association of Engineering Insurers), Working Group Paper WGP 126 (22), Coating Failures, revised 1 September 2022, page 24. The same paper names the misalignment: the constructor is interested in productivity, the operator in through-life performance. This is a finding about incentives, written by engineering underwriters. It is not an allegation of bad faith, and we do not present it as one.
  5. Coating during construction costs about $10 per square metre, and repairing it in service can reach $160. Verbatim: "During construction the cost of a good paint coating is about $10 per sqm, while the cost of repair in-situ can be up to $160 per sqm." Source: IMIA, same paper. Our English pages print this in square feet, $0.93 and $14.86, rounded to under a dollar and as much as fifteen; the conversion is ours. This compares coating costs to each other. It is not evidence of what replacing a structure costs, and it must never be used as such.
  6. Spot painting, overcoating, and removal and replacement cost 40 percent, 70 percent and 135 percent of the original coating cost. Source: Ebrahimi, Khanlarian, Momeni, Ahmadi and Gallant, "Toward sustainable steel bridge maintenance: anti-corrosion coating systems," Journal of Coatings Technology and Research, 22(6), 2025, pages 2293 to 2312. Open access and peer reviewed. Read the qualifier with the number: the paper’s own wording is "It is assumed that the coating cost for spot painting, overcoating, and removal and replacement are 40%, 70%, and 135% of the original coating cost, respectively", and it cites Helsel and Lanterman (2022) and Brevoort et al. (1996) for it. It is a costing assumption inside a peer-reviewed model, not a measured result of that study, and we label it that way.
  7. Corrosion mainly starts at edges and welds. Verbatim: "corrosion mainly initiates at edges and welds. Between 50% and 90% of the corrosion attacks could be attributed to this, depending on the type of structure. The paint failed due to low film thickness." Source: Knudsen, Hagen, Skilbred, Bruaas and Naerland, "Root causes for corrosion on painted steel structures in marine environments," Materials and Corrosion, 76(6), 2025, SINTEF. Abstract verified. An author copy sits in SINTEF’s open repository and the publisher version is paywalled, so we publish the range and not a per-structure figure, because we have not read the per-structure figures.
  8. The US federal deferred maintenance backlog more than doubled to $370 billion between fiscal 2017 and fiscal 2024, and the auditor’s own conclusion is that deferred assets "need premature replacement, which can be significantly more expensive than if maintenance and repairs were done when originally scheduled." Source: US Government Accountability Office, GAO-25-108400, April 2025. English edition only: this is US federal data and it does not localize to Brazil.
  9. Standards we cite, and what each one is for. ISO 12944-1:2017 for the definition of coating durability. ISO 12944-2 for corrosivity categories. ISO 8501-1, ISO 4628-3 and ASTM D610 for the grading scales. SSPC-PA 2-2022 for dry film thickness sampling. Standards are purchased documents: we cite the clause and the scale, and we do not reproduce them.
  10. A typical inspection lot has its film thickness measured across roughly 0.02 percent of its area. This is our own arithmetic from the sampling frequencies published in SSPC-PA 2-2022, not a figure the standard states, and it carries that label everywhere it appears.

The workforce is walking out the door

  1. Roughly 1 in 4 workers in US heavy industry is already 55 or older. Manufacturing 25.3%, mining and oil and gas 22.6%, construction 22.2%, utilities 21.4%. Source: US Bureau of Labor Statistics, Current Population Survey, 2025 annual averages.
  2. About 2.8 million US manufacturing job openings will come from retirements within the decade, and 1.9 million of the jobs needed are expected to go unfilled by 2033. Source: Deloitte and The Manufacturing Institute, 2024.
  3. Manufacturing loses on average 2.4% of its workforce every month to separations; construction 4.0%. Over a year, that is equivalent to roughly a third of the workforce turning over in manufacturing, and nearly half in construction (annualization is our derivation from the published monthly rates). Source: BLS JOLTS, 2025.

What training actually costs

  1. US companies spent $98 billion on training in 2024, and roughly 1 in 4 of those dollars went to travel, facilities, and equipment rather than learning itself (share derived from the report's published figures). Source: Training Magazine, 2024 Training Industry Report.
  2. Replacing an employee costs one-half to two times their annual salary. Source: Gallup, 2019. A conservative floor: 33% of base salary. Source: Work Institute.

Safety and human error

  1. "It is estimated that up to 80% of accidents may be attributed, at least in part, to the actions or omissions of people." Quoted verbatim. Source: UK Health and Safety Executive, HSG48, Reducing error and influencing behaviour.

What immersive training changes

  1. VR learners trained up to 4x faster than classroom learners on comparable material (about 3x including headset setup time). Source: PwC, The Effectiveness of Virtual Reality Soft Skills Training in the Enterprise. We used only information that is applicable across use cases and industries.
  2. VR learners were 275% more confident to apply what they learned. Source: PwC, same study.
  3. Peer-reviewed research finds VR training comparable to traditional training in aviation, surgery, driving, and assembly tasks, with reduced error rates, while flagging real limits: motion discomfort, haptic gaps, and the risk of negative transfer when a simulation misrepresents the real task. Source: Frontiers in Virtual Reality, review, 2021.
  4. Not a number, but cited: the most significant failure outcome for enterprise immersive technology is the stall between pilot and program, and the barriers are usually organizational, not technical. Source: AR Insider, 2024.

Our rule: primary source or it does not get published.

Before any statistic appears on this site, in our documents, or in our decks: we fetch the primary source and read the sentence the number lives in. Derivations are labeled as ours. Dated figures carry their year. Vendor-blog numbers that cannot be traced die in review. And when a famous number fails verification, we keep a record of it, so it never sneaks back in. It is more work. It is also why you can forward anything on this page without checking it first.

Sources

Every source used on this page, in the order the figures appear, linked wherever a free authoritative copy exists. The page links the source phrase inside each figure to the address here.

Numbers open the conversation. The field settles it.

Talk to a specialist about what these numbers look like in your operation. Real people, real field experience. No chatbots.

Talk to a specialist