Sam & @OpenAI ran an accidental* and wildly interesting experiment that's likely to help their impending IPO.
What: They launched a $500 plan on 9/29. Some upgraded accounts, like mine, did NOT have their limits increase... but we thought we did
The Experiment: How does customer behavior change when you placebo them into believing they have +150% more of something vs reality?
> Do customers realize?
> Do customers care?
> Does churn, usage, revenue, support tickets, happiness - go up⬆️ or down⬇️?
Global OpenAI token usage went up +28% post plan launch (incl. real upgraded users). I can't confidently guess what they learned.
Value of a Token
OpenAI can now internally re-price the "value" of their token - via their customers willingness to pay. Ex: if [prices tripled & usage maintains] vs if [prices go up slightly and usage plummets] signal different price elasticities.
Frequent pricing experiments are more important for heavily subsidized ai products like Codex or Claude Code. See fmr. openai COO video say - "pick a number, 20 sounds good".
During IPO season, pricing & margins suddenly start to matter... and subsidies disappear :/