How long to give a new campaign before judging it

The most expensive habit in paid search is killing campaigns too early. Someone launches on the first, checks on the fourth, panics on the seventh, changes the bidding strategy on the tenth, and concludes on the fourteenth that Google Ads does not work for their business. What actually happened is that the campaign never got out of the learning phase, and every change reset the clock.

Why the first two weeks tell you almost nothing

Smart Bidding starts with no information about your account and has to buy that information with your money. Early clicks are exploration. The system is testing audiences, times, placements, and query interpretations to find out which ones produce your conversion. Early cost per conversion is therefore the worst it will ever be, by design.

Beyond that, there is a plain statistics problem. If your normal cost per lead is a hundred and you have spent four hundred, you might reasonably have zero leads or four. Neither result means anything. People routinely draw firm conclusions from sample sizes that could not support a coin flip.

And every meaningful edit restarts part of the learning. Changing the bidding strategy, changing the target substantially, changing the conversion action, or changing the budget by a large factor all reset the clock. Ten small optimizations in three weeks do not add up to careful management. They add up to a campaign that never learns anything.

How long is actually enough

The honest answer is that it depends on conversion volume, not on days. Time is just the proxy people can see. What matters is whether the campaign has accumulated enough conversions to say something.

• Days one to fourteen. Look only for breakage. Is the tracking firing, are the ads approved, is the traffic broadly relevant, is spend pacing sensibly. Do not judge cost per conversion. Do not change bidding.
• Days fifteen to thirty. Begin reading the search terms report properly and adding negatives. This is the first useful work. Still do not judge performance.
• Days thirty to sixty. If the campaign has produced a meaningful number of conversions, the trend becomes readable. Compare against your own target cost per lead, the one you calculated from your margins, not against an industry figure.
• Days sixty to ninety. This is where a fair verdict lives for most service businesses, particularly anything with a sales cycle longer than a week.

Businesses with long consideration cycles need longer still. If your average customer takes two months to decide, a ninety day window contains only one cycle of results, and the leads generated in month three have not had time to close.

How to run the first ninety days properly

1. Write down the target before you launch. Your maximum cost per lead, calculated from margin and close rate. Put it in writing, with a date. A target invented after the fact is not a target.
2. Decide the review schedule in advance and hold to it. Day seven for breakage. Day fourteen and weekly after that for search terms. Day thirty for a structural read. Day sixty for a decision on direction.
3. Change one thing at a time, and only after a review point. Simultaneous changes make the result unreadable no matter how long you wait.
4. Keep negatives and creative work separate from bidding changes. Negative keywords and new ad assets can be worked on continuously. Bidding strategy and targets should be nearly untouched for the first month.
5. Track leads through to outcome from day one. Otherwise at day ninety you will know your cost per lead and still have no idea whether those leads were worth anything.
6. Write a short note in the account for every change you make. When you look back at week ten and wonder why performance shifted, that log is the only thing that will tell you.

When to kill a campaign early

Patience is not the same as passivity. Stop early when the evidence is structural rather than statistical.

Kill it if the search terms report shows the traffic is fundamentally wrong and no reasonable negative list fixes it. Kill it if spend is far above your allowable cost per customer with zero conversions after enough clicks that even a mediocre page should have produced something. Kill it if you discover the tracking was never working, then relaunch cleanly rather than trying to interpret corrupted data. Kill it if the landing page is clearly the failure point, fix the page, and start again.

What is not a reason to kill it: a bad first week, a competitor appearing, a cost per lead that is above target on day twenty, or an uncomfortable feeling on a Friday afternoon.

What to do next

Before your next launch, write two things on one page: the cost per lead that would make this campaign worth keeping, and the dates you will look at it. Then do not open the account between those dates except to check the search terms report. That single piece of discipline improves results more than most optimization work, because it stops you from paying for learning twice.

If you have already killed campaigns that you now suspect were killed too early, it is usually worth relaunching one properly rather than concluding the channel does not work. You can book a free consultation if you want a second opinion on which one.