Methodology, Sources, and Refresh Cadence
Every figure on ReworkCost.com is anchored to a named publisher. This page documents the source set, the formulas the calculator runs, what is in and out of scope, the refresh cadence, and the limitations of the model. If a number on the site is not traceable to one of the sources listed below, treat it as a bug and email corrections.
Sources
| Source | Refresh cadence | How we use it |
|---|---|---|
| NIST Planning Report 02-3 (2002) | Static (foundational study) | The $59.5B economy-level cost-of-inadequate-testing figure, the 20-40% rework-of-effort range, and the calculator's conservative 25% default (near the low-middle of that range). |
| DORA State of DevOps Report 2024 (Accelerate) | Annual (Q4 publication) | Change failure rate tiers (elite 5%, high 20%, medium 10%, low 40%; in 2024 medium and high inverted for the first time). Anchors the calculator's elite-team comparison. |
| IBM Systems Sciences Institute - Relative Costs of Fixing Defects | Static (1995 study) | The 1-10-100 cost-of-change multiplier framing. The site uses 5x as a conservative production multiplier on the calculator and 10-100x for the cost-of-change curve illustration on /case-studies and /sources/poor-testing. |
| Boehm, Software Engineering Economics (1981) | Static (foundational text) | Original cost-of-change curve. Cited on /case-studies and the home 1-10-100 callout. |
| Capers Jones, Applied Software Measurement (3rd ed., 2008) | Per new Capers Jones publication (irregular) | Defect Removal Efficiency tiers, the 45% requirements-origin-defect figure, team-size rework adjustment data on /benchmarks. The four-root-cause framework on /sources is Jones' taxonomy. |
| Capers Jones, Software Engineering Best Practices (2010) | Static | Industry-by-industry DRE benchmarks on /benchmarks. |
| ASQ Cost of Quality framework (Juran 1951, Crosby Quality Is Free) | Static framework; ASQ surveys irregular | The prevention / appraisal / internal failure / external failure four-bucket framing on /formula and /manufacturing. Anchors the COPQ section of /case-studies. |
| ISO 9001 quality management framework | Per ISO revision cycle | Quality cost framework reference on /manufacturing. ISO 9001 nonconformance and corrective-action requirements map to the rework definition used on /what-is-rework. |
| Six Sigma DMAIC methodology | Static | Define-Measure-Analyse-Improve-Control structure underlying the /reduce playbook ordering and the /measure metrics framing. |
| Lean rework reduction (Toyota Production System lineage) | Static principles | Waste taxonomy and the necessary-rework vs preventable-rework distinction on /agile-vs-waterfall and /what-is-rework. |
| Stripe Developer Coefficient (2018, updated 2020) | Irregular (Stripe Press) | The 17.3 hours-per-week figure cited on /sources/technical-debt and the home five-causes callout. |
| McKinsey Developer Velocity Index (2020) | Per McKinsey re-publication | Top-quartile vs bottom-quartile productivity ratios on /benchmarks. Cross-referenced on /case-studies. |
| GitHub Octoverse 2024 | Annual | Repository and engineering activity baselines used to sanity-check benchmark figures on /benchmarks. |
| Standish Group CHAOS Report | Periodic re-publication | Project success / rework rates cited on /agile-vs-waterfall. |
| Construction Industry Institute (CII) field rework research | Per CII research-team publication | The 5-15% well-managed and 10-30% poorly-managed construction rework figures on /construction. Cross-industry parallel for /case-studies. |
| BLS Occupational Employment and Wage Statistics (OEWS) | Annual (May reference period) | Loaded-cost anchoring (US median software developer wage plus 1.3-1.5x benefits and overhead) feeding the calculator's $80k-$400k fully-loaded slider range. |
In scope
- Top-down rework cost models: team_size x loaded_cost x rework_percentage applied to engineering organisations from small teams to enterprise scale.
- Bottom-up ticket analysis: ratio of bug / hotfix / regression story points to total sprint points, multiplied by team cost.
- Defect escape rate math: production defect count x average fix hours x hourly rate x 1-10-100 cost-of-change multiplier (default 5x, conservatively chosen).
- DORA-tier rework percentages (elite, high, medium, low) anchored to the 2024 State of DevOps Report.
- Cost of Poor Quality (COPQ) four-bucket framework (prevention, appraisal, internal failure, external failure) applied to software and manufacturing.
- Loaded-cost ranges anchored to BLS OEWS national medians plus typical benefits-and-overhead multipliers (1.3-1.5x base).
- Cross-industry context for manufacturing (Six Sigma, ASQ, ISO 9001) and construction (CII, CIOB, Love et al. civil-engineering research).
Out of scope
- Organisation-specific rework predictions. Calculator outputs are working bands, not commitments.
- Non-software vertical deep-dives beyond the context pages. /construction and /manufacturing exist to anchor the cross-industry parallel, not to compete with CII or ASQ reference works.
- Vendor-funded research with undisclosed methodology. Where a tool vendor publishes a quality-cost figure without sample size or method, the figure is not used.
- Single-anecdote stories without aggregation. A specific organisation's rework cost is described as a case study only where multiple sources align.
- Individual engineer productivity claims. The calculator reports team-level economics; nothing on the site evaluates a named engineer.
- Sub-regional wage variance below the national median. BLS state-level data exists but is outside the scope of the home calculator's loaded-cost slider.
Calculation framework
Top-down formula
annual_rework_cost = team_size x fully_loaded_cost x rework_pct. At the 25% default (near the low-middle of Capers Jones' 20-40% range) a 20-engineer team at $200k loaded cost spends $1M / year on rework. This is the headline figure on the home calculator.
Bottom-up ticket analysis
(bug_points + hotfix_points + regression_points) / total_sprint_points x team_cost. Requires consistent Jira / Linear labelling discipline. Worked example on /formula.
Defect escape rate model
escaped_defects x avg_fix_hours x hourly_rate x cost_of_change_multiplier. The multiplier defaults to 5x (conservative). IBM SSI / Boehm research supports 10-100x for late-phase fixes; the calculator's advanced section exposes this slider.
DORA elite-vs-current gap math
current_rework_cost - (team_size x fully_loaded_cost x 10%). The calculator displays this as the annual savings achievable by reaching elite performance. 10% is more conservative than DORA's sub-5% elite floor; chosen because the elite cohort is small and the median achievable target is closer to 10%.
Capers Jones DRE-to-escape conversion
escape_rate = 1 - DRE. A team at 85% Defect Removal Efficiency has a 15% escape rate. Elite teams reach 95%+ DRE through layered detection (unit, integration, review, QA). Driver tables on /benchmarks.
Cost-of-change curve (1-10-100)
A defect costing $1 to fix in design costs $10 in development and $100 in production. The principle holds directionally across IBM SSI, Boehm, Capers Jones, and modern CI/CD-era research. The /case-studies page covers the original studies and their caveats.
Refresh cadence
A monthly first-business-week pass re-verifies vendor pricing on /tools and re-reads the named-source landing pages on every other content page. The visible "Updated" stamps (footer and inline tokens) and the Article schema dateModified are the day each page last changed, read from the site's version history when it is built, so the footer and the body cannot drift apart. The "Data verified" date comes from a single LAST_VERIFIED_DATE constant in src/lib/schema.ts.
Out-of-cycle updates trigger when:
- New DORA State of DevOps Report (annual Q4 publication).
- New Capers Jones publication or dataset update.
- Major NIST or SEI rework-economics revision.
- ASQ Cost of Quality survey re-publication.
- Construction Industry Institute new field-rework benchmark publication.
- Material vendor-pricing drift on the tools listed on /tools (Linear, Jira, Notion, Productboard, etc.).
- Calculator math improvement (slider range, multiplier default, or formula correction surfaced by a reader).
Limitations
- Ranges are not predictions. The calculator surfaces a discussion-starter number, not an audit-grade budget.
- Survey response bias on benchmarks. Teams with the highest rework rates are the least likely to complete industry surveys; published benchmark medians systematically understate the real distribution by 5-10 percentage points.
- IBM SSI multipliers vary by organisation maturity. The 100x production multiplier holds in regulated and safety-critical contexts; modern teams with strong CI/CD and feature flags often see compressed ratios closer to 5-10x.
- Modern CI/CD may compress the cost-of-change curve. Where deployment frequency is high and rollback is trivial, the marginal cost of a late defect approaches the cost of a development-phase fix. Treat the 1-10-100 ratios as directional, not literal.
Editorial position
ReworkCost.com is operated by Digital Signet, an independent studio. We do not sell quality-management software, do not run a Six Sigma certification practice, and do not act as a process-improvement consultancy. Editorial direction is set by Oliver Wakefield-Smith. Drafts are produced via Digital Signet's autonomous AI development methodology and reviewed against the editorial framework before publication.
See /about for the operator, the sister-site portfolio, and the editorial principles.
Corrections
For source additions, methodology questions, or figures that look wrong, email oliver@digitalsignet.com. Corrections typically land within five business days and are reflected in the next monthly refresh.