Patrick McKenzie switched Bingo Card Creator from a Java Swing shareware app to a web app, doubling trial-to-purchase conversion and halving AdWords CPA.
Key Takeaways
Web app trial-to-purchase conversion hit 2.32% vs 1.35% desktop; AdWords CPA dropped from $20 to $9 on the same $28 product price.
Support tickets per 50 customers: 15 desktop, 3 web – no install failures, no lost registration keys, no stale versions on third-party download sites.
The shareware funnel has ~17 distinct failure points (download, JRE upgrade, installer, re-finding the app); web apps eliminate all of them.
Web analytics surfaced that “baby” was the most-used word in cards, revealing suppressed Baby Shower demand that desktop telemetry never exposed.
More features correlated with worse sales; 80%+ of purchases happen within two hours of first use – first-run experience is the only lever that matters.
Hacker News Comment Review
Core pushback: the argument is structurally circular – web apps outperform for audiences (non-technical consumers buying bingo cards) where desktop friction is highest; the conclusion doesn’t generalize to developer tools, open-source software, or tech-savvy users.
The 17-step funnel critique drew direct challenges: steps 1-4 are identical for web apps, modern browsers surface executables inline, and a native binary often loads faster than a heavy JS bundle.
Commenters flagged the 2009 timestamp as load-bearing: Electron collapsed install friction, and the current B2C performance leader is mobile apps, not web – the era-specific context makes the numbers hard to apply today.
Notable Comments
@mwkaufma: “Over a decade of circular ‘web apps are better for the subset of problems webapps are good at’ tautologies.”
@sudb: Asks what the numbers say now that Electron exists and mobile dominates B2C – the 2009 framing may invert entirely on a modern platform map.
