98% Isn't Much
The article provocatively asserts that a 98% success rate is frequently 'not enough,' particularly in software development and critical systems, challenging the common perception of high percentages. It highlights how failing to support a small percentage can still impact millions and underscores the importance of graceful degradation and edge case handling. This sparked a lively Hacker News debate on the practical and ethical trade-offs between engineering perfection, business profitability, and user inclusivity.
The Lowdown
The article '98% Isn't Much' challenges the intuitive assumption that a 98% success rate is always good. It argues that while 98% is excellent for extraordinary achievements like lottery wins or high exam scores, it becomes critically insufficient for basic expectations, safety-critical systems, or when dealing with a large-scale user base.
- The author contrasts scenarios where 98% is a blessing with those where it's unacceptable, such as a restaurant where 2% of meals cause food poisoning or an employer paying employees only 98% of the time.
- In a digital context, 98% browser compatibility, though seemingly high, can mean millions of users are left out, especially when considering a global audience (e.g., 2% of 7.5 billion people is 150 million).
- A key example cited is CSS nesting: while 'widely supported' globally, a client's specific audience might show much lower support (e.g., 70%), leaving a significant portion behind.
- The article emphasizes that truly robust engineering focuses on gracefully handling edge cases rather than solely optimizing for the majority.
- It concludes that if a new feature cannot degrade gracefully for the unsupported 2%, it has failed to meet basic minimums.
The piece serves as a reminder for developers and product managers to critically evaluate statistics in context, advocating for a user-centric approach that considers the real-world impact on all users, not just the statistical majority.
The Gossip
Contextual Criticality
Users largely agree that the meaning of '98%' is entirely context-dependent. They highlight the distinction between scenarios where partial credit is acceptable (e.g., a slightly imperfect website layout) and those where it's a critical failure (e.g., food safety, system uptime, or mission-critical software). The discussion emphasizes that the 'cost of failure' for the remaining 2% dictates whether the percentage is 'enough,' making a universal '98% rule' misleading.
Economic Edge-case Evaluation
A significant portion of the discussion revolves around the economic realities and business trade-offs of supporting the 'last 2%.' Many argue that optimizing for 98% is often a pragmatic, profit-driven decision, as the cost and effort to serve the remaining few can outweigh the benefits, especially for commercial products. Others counter that this mindset can lead to exclusion, particularly for public services or essential platforms, and highlight that some 'bad clients' (e.g., enterprise users with old systems) can still represent significant revenue.
Browser Battlegrounds & Graceful Gestures
The article's web development examples sparked a fervent debate about modern browser features versus supporting older or niche browsers. While some advocate for aggressive adoption of new technologies, pushing users to update and dismissing those on outdated systems, others emphasize the crucial importance of 'graceful degradation'—ensuring core functionality works even if advanced styling or features are unavailable. This also touched upon accessibility concerns, the impact of browser monoculture, and the perennial challenges of web compatibility for developers.
Statistical Stakes Beyond Software
Commenters extended the '98% isn't much' concept to various other fields, including AI accuracy, GPU programming, and government services. They pointed out how a small percentage of failure can lead to significant real-world consequences when scaled up (e.g., AI in self-driving cars, cumulative errors in large datasets). The discussion also drew analogies to physical infrastructure accessibility (like wheelchair ramps) and how profit motives can influence societal access to essential services, illustrating that the principle applies broadly where reliability and inclusivity are paramount.