How to write a launch readiness checklist that ends in a decision
Build a launch checklist with owners, evidence tests and a rollback trigger, so it ends in a clear call: ship, delay or launch quietly.

Most launch checklists are long lists of things to do. They tell you what is left. They do not tell you what to decide. So the launch slips on gut feel, or it ships and sales hear about the new plan from a customer. This guide shows how to build a checklist that ends in one of three decisions: ship, delay, or launch quietly.
Who this is for
A product manager two to four weeks from a launch, who has to get sales, support, marketing and engineering to agree that it is ready.
What you will have at the end
A short list of gates, each with one owner and a yes/no evidence test. Three outcomes with rules for which failures lead to which. A rollback trigger. A go/no-go meeting on the calendar.
Separate gates from nice-to-haves
A gate is an item that changes the decision if it fails. A nice-to-have makes the launch better but does not change what you do. Ask one question of every item: if this is not done on launch day, would we hold, shrink or reverse the launch? If the answer is no, it is a nice-to-have. Keep it on a second list and do not review it in the meeting.
Aim for ten to fifteen gates. If every item blocks, then nothing really does, and the team will learn to ignore the list.
Cover five areas
Launches fail at the edges of the product, not only inside it. Write at least one gate for each of these areas, and ask the question beside it.
- Product qualityDoes it work, including when things go wrong?
- Support and opsCan we help and fix people on day one?
- Sales and enablementCan sales explain it and know who it is for?
- Messaging and pricingDoes every public page say the same true thing?
- MeasurementAre events and dashboards live so we can see what happens?
Measurement is the one most often skipped. If you cannot see the result, you cannot use a rollback trigger, which is step five.

Give every gate one owner and an evidence test
The owner is one named person, not a team. A team can share the work, but one person answers for the gate in the meeting. If you cannot name that person, the gate is not ready to be on the list.
The evidence test is a yes/no check that someone else could run. Compare two versions of the same gate. “Help article done” can be ticked by the person who wrote it. “A colleague outside the team follows the article and completes the task” cannot be argued with. Write the test so that a stranger could see it pass or fail.
Check the box on evidence, never on “done”. Done is a feeling. Evidence is something you can show in the room.
Define three outcomes and map failures to them
Decide what each outcome means before any gate fails. Then write down which failures lead where. This is what removes the argument later.
| Outcome | What it means | Which gates failed |
|---|---|---|
| Full launch | Announce and open to everyone. | None. |
| Quiet launch | Limited rollout to a small share of users, no announcement. You learn and fix with a small blast radius. | Gates where small exposure limits the damage, such as support load, sales readiness or part of the measurement. |
| Delayed launch | Hold until the gate passes. Set a new date for the check. | Gates where any exposure does harm, such as money, data, security or a broken core path. |
The test for a quiet launch is simple: with only a few users, can the team still cope and still see what is happening? If yes, a quiet launch is safe. If no, delay.
Set a rollback or kill trigger in advance
Every launch needs a line that says when you stop. Write it before launch, when nobody is under pressure. It has four parts: the metric, the threshold, the window, and the person who decides.
For example: if the error rate on the new flow stays above a set level for a set period, or support tickets about it pass a set number in a day, the named decider pauses the rollout. The numbers should come from your own normal levels, not from a guess. Also check that the pause actually works. A trigger you cannot act on is only a sentence.
Hold a short go/no-go meeting three to five days before
Keep it to thirty minutes. Walk the gates one by one. Each owner states pass or fail and shows the evidence. Do not discuss how people feel about it. Where a gate fails, apply the rules from step four.
Then record the decision: the outcome, the failed gates, who will fix each one, and the date you will check again. Three to five days gives you time to act on what you hear. On launch day, the only choice left is to go or to cause a scramble.
Mistakes that make the checklist useless
- Gates with no owner, so nobody answers for them in the meeting.
- Treating every item as blocking, so the list cannot tell you anything.
- No rollback trigger, so the first bad day becomes a debate.
- Ticking boxes on “done” instead of evidence.
- Holding the go/no-go meeting on launch day, when nothing can be changed.

Worked example: a new self-serve billing plan
Say you are launching a new self-serve billing plan. This is an illustration built to show the method. Here are twelve gates across the five areas, each with an owner and a yes/no test.
| # | Gate | Owner | Evidence | Status |
|---|---|---|---|---|
| 1 | Upgrade and downgrade work end to end | Engineering lead | Ten test accounts complete both paths with a correct invoice | Pass |
| 2 | Failed payments are handled | Engineering lead | A failed test card gives the right email and account state | Pass |
| 3 | No open severity-one bugs | Engineering lead | Bug tracker filter shows zero | Pass |
| 4 | Support macros written and agents trained | Support lead | Five test tickets resolved using the macros | Fail: two of five |
| 5 | Refund and proration process written | Finance lead | One test refund processed by following the steps | Pass |
| 6 | Plan can be switched off | Engineering lead | Flag turned off for a test group; accounts return to the old plan | Pass |
| 7 | Sales know who is eligible | Sales lead | Each rep answers the sample customer questions without help | Pass |
| 8 | Accounts on older contracts identified | Sales lead | List exported and reviewed by account managers | Pass |
| 9 | Pricing page and in-app copy match | Marketing lead | Side-by-side review signed off by finance | Pass |
| 10 | Terms and invoice wording approved | Finance lead | Written approval attached to the checklist | Pass |
| 11 | Purchase events fire on every payment path | Analytics owner | A test purchase on each path appears in the dashboard | Fail: one path missing |
| 12 | Dashboard shows error rate and tickets by plan | Analytics owner | Dashboard open in the meeting with live test data | Pass |
At the go/no-go meeting, ten gates pass and two fail. Gate 4 fails because the support team resolved only two of the five test tickets. Gate 11 fails because one payment path does not send its purchase event.
Apply the rules from step four. Neither failure touches money, data or the core path. Both are limited by small exposure. A small group of accounts creates few tickets, so the support lead and the agents who did pass can handle them. A missing event on one path can be covered for a short time by checking billing records by hand. So the outcome is a quiet launch to 10% of accounts, not a full launch and not a delay.
The record of the decision says: quiet launch to 10% of accounts. Support lead retrains agents and reruns the five tickets. Analytics owner fixes the event on the missing path. Both are checked again at the next review. The rollback trigger is already written: if the payment error rate or billing tickets pass the levels the team set, the product lead pauses the rollout and the engineering lead switches the plan off using gate 6. When gates 4 and 11 pass, the team widens the rollout.
Nobody had to argue about whether the team felt ready. The checklist and the rules did the deciding, and the owners knew what to fix.


