Key data points
- Across 5,823 approved usability findings, 47.9% were flagged as “quick wins” (fixable without a major redesign), almost evenly split with the 52.1% that require deeper structural work.
- Quick-win rate is not driven by severity in the way most teams assume: medium-severity issues have the highest quick-win rate (69.3%), while low-severity issues have the lowest (20.7%), the opposite of the intuitive assumption that smaller problems are always easier to fix.
- 56% of all findings were reported by only one tester. Just 19.4% were confirmed by four or more testers independently.
- Issues reported by only one tester were the least likely to be rated high or severe severity (33.1%), compared to 55.0% for issues reported by 2-3 testers and 48.5% for issues reported by 4 or more.
- Despite that, the quick-win rate barely moves across frequency bands (48.0% for one-off findings, 46.6% for 2-3 testers, 49.1% for 4-plus testers), meaning how many people hit a problem tells you almost nothing about how easy it is to fix.
The two variables teams conflate when triaging a backlog
Most accessibility and usability backlogs get triaged on a single axis: severity. But this dataset, 5,823 synthesized findings from 15 independently tested products, contains two genuinely independent variables that both matter for prioritisation, and conflating them leads to bad sequencing decisions. The first is severity: how badly does this block or degrade the task. The second is frequency: how many independent testers hit it. A backlog sorted only by severity treats a rare, catastrophic edge case the same as a systemic pattern affecting every user with a given need, and a backlog sorted only by frequency risks burying a severe blocker that only came up once because the study design happened to route just one tester through that path.
Severity and “easy to fix” are not the same thing
The instinct to assume smaller problems are easier to fix and bigger problems are harder does not hold up in this data:
| Severity | Approved findings | Quick-win rate |
|---|---|---|
| Low | 1,731 | 20.7% |
| Medium | 1,670 | 69.3% |
| High | 1,898 | 54.9% |
| Severe | 517 | 44.5% |
Severity bands cover the 5,816 findings that carry a severity rating, out of 5,823 findings in total.
Medium-severity findings are the most fixable category, not the least: 69.3% are quick wins, more than three times the rate for low-severity findings (20.7%). A plausible explanation, consistent with what these categories typically contain: low-severity findings are often minor content, copy, or polish issues that require design or content sign-off cycles even though the change itself is small, while medium-severity findings more often involve a discrete, well-scoped technical fix (a missing label, an unclear button, a contrast tweak) that an engineer can implement without further design input. The practical implication is that severity alone is a poor proxy for effort, and any backlog tool that sorts purely by severity will systematically under-prioritise a large, fast-to-clear category of medium findings.
Frequency tells you about severity, but not about effort
Splitting findings by how many testers independently hit the same issue shows a real relationship with severity, but not with fixability:
| Frequency | Findings | % high/severe | Quick-win rate |
|---|---|---|---|
| 1 tester (one-off) | 3,260 (56.0%) | 33.1% | 48.0% |
| 2-3 testers | 1,432 (24.6%) | 55.0% | 46.6% |
| 4+ testers (systemic) | 1,131 (19.4%) | 48.5% | 49.1% |
Issues confirmed by 2 or more testers are meaningfully more likely to be high or severe (roughly 48-55%) than issues reported by only one tester (33.1%), which makes sense, since a problem severe enough to block a task tends to surface for more of the people who encounter it. But the quick-win rate is essentially flat across all three bands, sitting between 46.6% and 49.1% regardless of frequency. This means a team cannot use “only one person hit this” as a reason to deprioritise on the assumption that solo-reported issues are also trivial to fix: in this dataset, they are not meaningfully easier or harder than issues confirmed by many testers.
A four-quadrant approach the data actually supports
Because severity and frequency are independent variables here, both should be tracked as separate axes on a backlog rather than blended into one score:
High severity + high frequency is the smallest and most urgent group in this dataset: findings confirmed by 4+ testers and rated high or severe make up roughly 9-10% of the total approved findings, and represent the closest thing to a “must-fix” consensus signal available from testing data.
High severity + low frequency deserves separate tracking rather than automatic deprioritisation. A third of one-off findings are still high or severe (33.1%), meaning roughly 1 in 3 single-tester reports is describing something that would seriously block a user with the right combination of need and task: these are exactly the findings most likely to be silently dropped by a backlog sorted purely on frequency or “number of duplicates.”
Medium severity, any frequency is the highest-leverage bucket for closing the largest number of issues quickly, given the 69.3% quick-win rate observed at that severity tier, a backlog aiming to clear the most findings per unit of engineering time should draw disproportionately from here, not from low-severity items.
Low severity findings should be batched rather than fixed piecemeal, since their low quick-win rate (20.7%) suggests they typically require a design or content review cycle regardless of how small the individual change looks.
Frequently asked questions
What percentage of accessibility and usability findings are actually easy to fix?
In this dataset, 47.9% of 5,823 approved findings were classified as quick wins, fixable without a major redesign. That rate is highest for medium-severity findings (69.3%) and lowest for low-severity findings (20.7%).
Does an issue reported by only one tester matter less than one reported by many?
Not necessarily for severity: a third (33.1%) of single-tester findings in this dataset were still rated high or severe. It does matter for confidence that the issue is systemic (findings confirmed by 4 or more testers rose from 33.1% to 48.5% high/severe), but frequency had almost no relationship with how easy the issue was to fix.
Should backlog prioritisation use severity or frequency of occurrence?
Both, tracked separately. This dataset shows the two are only loosely correlated: frequency predicts severity somewhat (issues hit by 2+ testers are 1.5-1.7x more likely to be high/severe than one-off issues) but has almost no relationship with how quickly an issue can be fixed, which severity also fails to predict cleanly.
About this analysis
Figures in this article are drawn from an anonymised aggregation of 15 independent usability testing projects conducted by See Me Please between late 2025 and mid-2026, covering government, financial services, healthcare, higher-education, and consumer digital products in Australia and the United Kingdom. “Quick win,” severity, and frequency (affected tester count) classifications were applied during test synthesis by trained reviewers based on task-based sessions. All client and participant identities have been removed.
