
An export promises two minutes and sometimes takes forty, and Maya has to choose between making it faster and making it honest
It Says Two Minutes — Volume I, Understanding Humans Before Products · Trust · Dilemma · Foundations. An export promises two minutes and sometimes takes forty, and Maya has to choose between making it faster and making it honest
01 · Opening
An export promises two minutes and sometimes takes forty, and Maya has to choose between making it faster and making it honest
02 · The setup
Dev brings it up the way people bring up things they have mentioned before.
Dev: The export dialog says about two minutes. For the big accounts it is closer to forty.
03 · The setup
Export duration, last 30 days — Median: 1 min 50s; 90th percentile: 9 min; 99th percentile: 38 min; Dialog says: about 2 minutes
04 · The evidence
Maya: The median is honest. It is the tail that is lying.
Dev: The tail is our largest accounts. It is the only tail that matters.
05 · The evidence
The support notes are all the same shape. Nobody complains about waiting. They complain about being told it would be over.
Maya thinks: They stayed at their desks for two minutes and then had to decide, over and over, whether it had broken.
06 · The evidence
Sam: Two accounts asked me last week whether the export is reliable. It is reliable. It just does not look it.
07 · The evidence
Two options, one quarter, and Dev is clear that only one of them is actually available.
Dev: I can maybe halve the tail. Forty becomes twenty and the dialog is still wrong.
08 · The evidence
Maya: Or we spend the same weeks telling them exactly where it is and stop promising a number we cannot keep.
09 · The evidence
They ship the honest version first, because it is testable in a fortnight and the other is not.
Export screen, after — Shows: rows completed of total; Shows: account size band; Promises: nothing; Median duration: unchanged
10 · The evidence
A customer forgives a slow operation. What they do not forgive is being told it would be over and finding it was not, because that teaches them the product does not know what it is doing.
Every unkept small promise spends a little of the credit you need for the large moments — an outage, a price change, a migration. Speed buys satisfaction. Honesty buys the benefit of the doubt, and only one of those is there when something actually goes wrong.
This is why the honest version is the ambitious choice rather than the resigned one. Halving the tail still leaves the dialog wrong. Telling the truth makes it right at every duration, including the ones you have not measured yet.
11 · You make the call
The story does not tell you first.
A. The waits felt shorter once there was something to watch
B. The product stopped making a promise it could not keep
C. Customers were told the export is reliable
12 · What happened
The tail work happens the following quarter and lands quietly, because by then nobody is watching the clock in fear.
Maya thinks: Same forty minutes. Completely different product.
13 · Complete
Speed buys satisfaction and honesty buys the benefit of the doubt, and only one of those is available when something goes badly wrong. Next: what an apology is worth on the third occurrence.