It is March. It failed. Why?
Ask a room what could go wrong and you get manners. Tell it the project already failed and you get a list.
Same room, two questions
Same eight people, same kickoff. Only the question changes.
Move card processing to a new payments provider. Kickoff on 13 January, cutover on the first weekend of March.
“Before we start: what could go wrong?”
- Dana, payments engineer
Their API docs are good. Should be fine.
- Marco, finance
Nothing from my side.
- Priya, tech lead
Testing, maybe? But we have seven weeks.
The read
Forty seconds of silence, three polite answers, nothing to log. Doubting a plan out loud, in front of the person who wrote it, costs something. So nobody does.
“It is 31 March. The migration failed. Take three minutes and write down why, one reason per note.”
- 1Saved cards never moved. The token export took six weeks, not two.
- 2Refunds on old charges had nowhere to go after cutover.
- 3The old contract auto-renewed for a year on 1 March.
- 4We switched routing and found there was no way back.
- 53-D Secure challenges tripled. Checkout conversion fell 9%.
- 6Webhooks arrived out of order. Orders marked paid twice.
- 7Fraud rules did not carry over. Chargebacks spiked.
- 8Finance could not reconcile: new payout report, new format.
- 9Only Priya knew the old integration. She was on leave in February.
- 10Sandbox behaved nothing like production.
- 11Cutover weekend landed on the spring sale.
- 12PCI sign-off for the new checkout came in late.
The read
Twelve notes in ten minutes, from the same eight people. The failure is already in the past, so naming a cause is not doubting the plan. It is explaining history.
Twelve notes, three risks
Twelve notes get read once and forgotten. Cut them before you leave the room.
-
Cluster
Stick the notes on a wall and let the duplicates find each other. Four piles.
- Money in flight1, 2, 4
- Checkout5, 6, 7, 10
- Finance and contract3, 8
- People and calendar9, 11, 12
-
Pick three
Only the ones that are both likely and expensive. The sandbox gap is likely but cheap: you find it in week one. A late PCI sign-off is expensive, but the audit is already booked.
-
Log each one
Each goes into the RAID log with an owner, a trigger and a date. A risk with no date is a worry.
Saved cards do not reach the new provider before cutover.
- Owner
- Dana
- Trigger
- Old provider has not confirmed an export date in writing.
- By
- 6 February
The old contract renews for twelve months on 1 March.
- Owner
- Marco
- Trigger
- No signed acknowledgement of our 30-day notice.
- By
- 30 January
Cutover cannot be reversed once card traffic moves.
- Owner
- Priya
- Trigger
- Routing back to the old provider not rehearsed in staging.
- By
- 20 February
Running one
Read the prompt word for word. The past tense is the whole trick.
“It is [the date after launch]. [The project] failed, badly. Take three minutes, on your own, and write down every reason why. One per note.”
- When
- Kickoff, after the plan is presented and before anyone commits to it.
- Who
- The whole team, plus one person who lives with the result: support, finance.
- How long
- 45 minutes. Three of them silent.
- Then
- One note each, round the room, until none are left.
Nobody will doubt your plan to your face. Everybody will explain its failure.