98% isn't much

While 98% support may sound impressive, it represents failure for critical services and can exclude millions of people—for a website, that's potentially 150 million users left with broken functionality. The article argues that 98% is only acceptable for exceptional improvements to quality of life, not basic expectations like safety or accessibility, and that true robust engineering requires graceful degradation for edge cases rather than breaking for 2% of users. Context matters: a feature "widely supported" by 98% of the general population may only work for 70% of a specific website's actual audience, making blanket percentage claims misleading.