Are quality issues stable?
№ 204 · Defect / rework rate
Definition
MetricShare of units that fail first pass and require rework, replacement, credit or return
Unit% of units
Single line of defects detected per units produced over time. Should trend down with process maturity. Pull from quality control logs. Spikes after vendor changes or new SKU launches are common. Example: 2.1% baseline, spike to 4.8% after vendor change (investigate vendor).
- SKU
- Stock keeping unit. One distinct sellable variant. Counted at the variant level, not the product level.
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| above 25% of units | 15% to 17% of units | under 8% of units returned or reworked |
US retail and ecommerce first-pass failure measured as the return rate, NRF and Happy Returns 2025 Retail Returns Landscape. Online-only operations sit structurally higher at 19.3%, and apparel higher again. The software delivery analogue, change failure rate, runs on a completely different scale and is covered in the note. · The bands are anchored on retail returns because that is the only large-sample, annually refreshed measure of first-pass failure available for commerce operations. Two adjustments before use. First, channel: 15.8% is the all-channel figure and online alone is 19.3%, so a pure ecommerce operation should read against the higher number. Second, category: apparel returns run far above the blend, with estimates of 25 to 40% of gross sales, while food and beverage sits at 1 to 3%. The software analogue is not comparable in level and should never be plotted on the same axis: DORA's elite cluster runs change failure under 5% against low performers at roughly eight times that, and 19% of surveyed organisations reach the elite cluster. Also note that roughly 9% of retail returns are fraudulent, which is a loss prevention problem wearing a quality problem's clothes.
Category split omitted: Only one Tier 1-2 publisher (NRF with Happy Returns) reports the retail channel and category split; the vertical breakdowns available elsewhere come from single-publisher operator syntheses, so the two-source test is not met.
When it looks bad
The headline rate is flat month to month while the mix underneath rotates, so one defect code grows into the majority of the volume without the total ever moving enough to trigger a review.
Defect rate steady at 8.4% for six months while wrong size sent grows from 12% to 41% of all defects. That is a pick accuracy failure being averaged into a quality number.
What to do about it
- Split the rate by defect code and chart the mix alongside the total. A stable total with a rotating mix is the standard way a single failing process stays invisible, and the mix chart is what makes it visible.
- Reserve and report returns at booking using a per-category rate rather than a company average. Apparel returns run 25 to 40% of gross sales against 5 to 10% for beauty, so a blended reserve systematically understates the categories that need attention (Eightx, 2026).
- Capture a structured return reason at the point of return and route the leading code to buying or product weekly. 15.8% of US retail sales came back in 2025 and 19.3% of online sales, and roughly 9% of returns were fraudulent, so the same reason data serves quality and loss prevention (NRF and Happy Returns, 2025).
- Before introducing return fees as the fix, price the demand cost. Among retailers that began charging, 47% report more complaints, 37% lost customers and 34% saw average order value fall, so the policy lever trades a quality number against a revenue number (eMarketer reporting NRF, 2026).
Sources
- National Retail Federation and Happy Returns Retailers estimate 15.8% of 2025 annual sales will be returned, totalling 849.9 billion USD. An estimated 19.3% of online sales will be returned. Retailers expect 17% of holiday sales to come back. Consumers aged 18 to 30 averaged 7.7 online returns in twelve months. nrf.com ↗
- National Retail Federation and Happy Returns Retailers estimated 16.9% of 2024 annual sales would be returned, totalling 890 billion USD, up from about 15% and 743 billion USD in 2023. 93% of retailers said returns fraud and exploitative behaviour is a significant issue, and 51% of Gen Z consumers reported bracketing. nrf.com ↗
- Retail TouchPoints, reporting NRF and Happy Returns 9% of all returns are fraudulent. Retailers tracking fraud report overstated return quantity (71%), empty box returns (65%) and decoy or counterfeit returns (64%). 85% of retailers are deploying AI to detect it. retailtouchpoints.com ↗
- eMarketer, reporting NRF and Happy Returns 72% of retailers now charge for at least some returns, up from 66%. Among those that started charging, 47% report more customer complaints, 37% lost customers, 34% saw average order value fall and 24% saw sales decrease. The cure has a measurable cost. emarketer.com ↗
- DORA (Google Cloud), reported by Optimal Elite performers hold change failure rate under 5%, deploy on demand, keep lead time for changes under one day and restore service in under one hour. Low performers deploy less than once per month. getoptimal.ai ↗
- Octopus Deploy, reporting DORA Elite performers show 8 times lower change failure rates than low performers, 127 times faster change lead times and 2,293 times faster recovery from failed deployment. Elite performance represents under 20% of organisations. octopus.com ↗
- Libertify, reporting DORA Elite performers are 19% of respondents, high 22%, medium 35%, low 25%. Industry does not meaningfully affect performance level. libertify.com ↗
- EASE, reported by Manufacturing Lead Generation Scrap and rework alone cost the average manufacturer up to 2.2% of annual revenue. There were 2,454 US product recalls through 2024, on pace for a six-year high. manufacturingleadgeneration.com ↗
- Eightx Apparel typically runs 30 to 50% returns by industry estimates, which doubles the outbound plus inbound parcel cost per net sale. Beauty typically runs 5 to 10%. eightx.co ↗