The Rule of 100 Clicks: Why I Won't Let You Judge a Campaign Too Early

A client emails me. "This campaign isn't working." Tone of someone who's already decided to pull the plug.

I open the Google account. The campaign has been live for nine days. It's gotten seven, maybe ten clicks a day. It hasn't even cracked a hundred clicks total.

This happens constantly. It is, without exaggeration, one of the most reliable ways I've landed a client: not by showing them something broken, but by showing them something that was never actually given the chance to work.

You Can't Read a Verdict Out of Noise

The surface story is always the same. The campaign isn't working. The agency is wasting money. The algorithm doesn't get my business. Something is broken and someone should fix it now.

But ask the obvious question: how do you know it isn't working? Usually the answer is a feeling, backed by a number too small to mean anything.

Here's the bedrock truth underneath all of it: a sample size too small to be meaningful isn't a verdict, it's just noise dressed up as one. Seven clicks a day for nine days is not a failed Google Ads campaign. It's an experiment that hasn't run long enough to produce a result yet, in either direction.

I have a video I send clients in exactly this situation, and the analogy is always the same: if seven people walked into your physical store today and none of them bought anything, would you conclude your store doesn't work?

Would you fire your staff, repaint the walls, and rewrite your signage based on seven browsers? Of course not. You'd say the same thing I say about a Google Ads campaign with seven clicks a day: you don't have a failing store, you have a slow Tuesday.

The number that actually means something is 100 clicks.

Not because it's magic, but because it's the rough point where you stop gambling on a feeling and start looking at an actual sample. Below it, your Click-Through Rate and Cost Per Conversion are just statistical noise wearing a costume. Above it, you can finally start asking real questions about what's happening.

What I Used to Do By Hand, the Algorithm Now Does By Force

Before Smart Bidding existed, this is how I decided whether something was working: I waited. I pulled reports, I tracked clicks manually against conversions, and I refused to let a client's panic about a quiet week override a sample size that wasn't there yet. That discipline wasn't a nice-to-have. It was the entire job.

What's funny is that the machine learning models running Maximize Conversions, Target CPA, and Performance Max today need exactly the same thing I needed back then: enough real data to tell signal from noise. Google's algorithms aren't matching intent so much as matching predictions, and a prediction engine fed seven data points a day for nine days isn't being lazy or broken when it underperforms. It's doing exactly what a human strategist would do with that little information: guessing.

When you pause a keyword or rewrite a bid strategy before it's hit a real sample size, you're not fixing the algorithm's mistake. You're recreating, with a machine, the exact same impatience that used to make me refuse to trust a client's gut over the data.

The Challenge

The next time a campaign feels like it isn't working, don't open the campaign looking for someone to blame even if it is tempting. Open it and count the data. If you're under 100 clicks, you don't have a failing campaign and you don't have a broken algorithm. It is a data density issue. You have a store (online) that hasn't had enough visitors yet to tell you anything.

You have two real options at that point, and only two: give it more time, or give it more money. Everything else, every keyword pause, every bid change, every panicked ad copy rewrite, is just you trying to soothe your own impatience and calling it optimization.

The Bottom Line

A campaign with too little data hasn't failed. It hasn't been tested yet and the data isn’t flushed out fully.

The discipline that used to separate a good account manager from a bad one wasn't talent or intuition, it was the willingness to wait for a real sample before drawing a conclusion.

That hasn't changed just because a machine is the one waiting now.