When to pause a UA campaign: write the pause rule before launch
Most UA teams kill campaigns on gut feel and week-old data. Then they call it discipline.
Here is how it usually goes. A campaign underperforms. Someone says it needs more time to learn. Someone else says the new creative is about to land. Two weeks later you are still spending, and the only thing that changed is how attached everyone got to it.
The decision is not the problem. The timing of the decision is.
What I write before launch
Every campaign gets a predicted D90 ROAS off its first three days. Not installs, not CTR, not spend pacing. The early signals that predict where a cohort lands:
- Day 1 to day 3 retention
- Day 0 to day 3 ROAS
- Session time
- Tutorial completion
- Cost per early event
Then three tiers, and the action for each one is written down before a single dollar moves.
| Predicted D90 ROAS | Action |
|---|---|
| Under 50% | Paused on day three. Automatically. No meeting. |
| 50% to 100% | Budget capped low, three more days to prove it can climb. |
| Over 100% | Budget up 25% to 30% every three days, for as long as it holds. |
The exact thresholds depend on your margin and payback window. The structure is what matters: three outcomes, decided in advance, and one of them is a kill.
Why it has to be written first
A live campaign has an owner. It has a slide. Someone has already defended it out loud.
A rule you write before launch is a decision about math. The same rule written after launch is a negotiation about someone's judgment. You will lose that negotiation, and you should, because by then nobody in the room is really arguing about ROAS.
Why three tiers and not two
Most frameworks offer two outcomes: keep or kill. That forces a bad choice on every campaign that is not clearly good or clearly bad, and those are most of them.
The middle tier is the salvage lane. A campaign at 70% predicted ROAS is not a loser yet. It gets a small budget and three more days. If it climbs, it moves up. If not, it is paused by the same rule.
The test nobody likes
Your pause rule will handle the campaigns nobody cares about perfectly.
The only test that counts is whether it can kill the one with a champion. If it bends there, it was never a rule. It was a preference with a threshold written next to it.
The move
- Pick your early signals. Use the events that best predict long-term revenue in your own data, not a list from a blog post.
- Fit a predicted D90 ROAS from day 3. Even a rough model beats week-old blended ROAS.
- Write the three tiers and their actions into the media plan before launch.
- Automate the bottom tier. If a pause needs a meeting, it will not happen on day three.
- Review the rule quarterly, not the campaigns daily. Change the thresholds with data, never mid-flight for one campaign.
Bottom line
Write the pause rule before you launch, not after you are attached.
FAQ
How many days should a UA campaign run before you pause it?
With a predicted ROAS model, three days of data is often enough to sort campaigns into tiers. Without one, teams tend to wait one to two weeks and decide on feel. The key is deciding the rule before launch, not the exact number of days.
What signals predict D90 ROAS early?
Day 1 to day 3 retention, day 0 to day 3 ROAS, session time, tutorial completion and cost per early event are common predictors. Which ones work best depends on your product, so test them against your own past cohorts.
Why not just wait for real D90 data?
Because by day 90 the budget decision that mattered has already been made. Waiting means spending three months on campaigns you could have sorted in three days.
What if a campaign in the bottom tier has a strong internal champion?
That is exactly the case the rule exists for. If the rule bends for the campaign someone loves, it is not a rule. Adjust thresholds at the quarterly review, based on data, never for one campaign.