Everyone in PPC has a broad match horror story. You tried it, the CPA exploded, and you turned it off after two weeks. Or you ran a campaign experiment the "right" way and the learning phase wrecked performance for a month. So you went back to phrase and exact, where things were at least predictable.
The frustrating part is that broad match can work. You've seen people say it cut their CPL in half. You've read that broad match with tCPA is the best-performing setup at scale. You're not sure if you're missing something or if they're just in different situations than you.
They probably are. And there is a specific reason why.
Broad match does not perform on its own. It performs when the algorithm has enough data to use it well. Without a strong conversion signal telling Google what a valuable user looks like, broad match is just an expensive way to show your ads to people who are vaguely curious about your category.
Here is how to know if your account is actually ready for it, and how to test it without blowing up the campaigns that are working.
The conversion volume threshold
The number that keeps coming up among practitioners: 30 conversions per month minimum before broad match can meaningfully learn.
This is not a hard Google rule, but it reflects how smart bidding actually works. The algorithm needs enough qualified conversion events to build a real model of your buyer. Below 30, it is extrapolating from thin signal. On broad match, that means impressions expand into adjacent queries that look similar on the surface but do not convert at your target rate.
If you are below 30 conversions per month, the fix is not broad match. It is getting more signal into the account. That means tightening your phrase and exact match structure, making sure your offline conversions are flowing back correctly, and not optimizing for cheap form fills if what you actually care about is qualified leads or closed revenue.
Get the conversion volume stable first. Then consider broad.
Why campaign experiments are the wrong tool for this test
A lot of people try to test broad match with a campaign experiment because it feels like the cleaner, more controlled approach. It is not.
When you run a campaign experiment, you have to take the campaign off your portfolio bidding strategy. That change alone can cause a temporary performance drop, even if you keep the exact same tCPA target. On top of that, the experiment campaign enters its own learning phase, which distorts the comparison from day one. You end up looking at two weeks of messy data and concluding that broad match does not work, when what you actually measured is the chaos of a mid-flight bidding strategy change.
The better approach is a separate, isolated campaign.
Keep your existing phrase and exact match campaigns completely untouched. Build a new campaign dedicated to broad match, with its own budget, the same conversion goals, and the same tCPA or tROAS target. Let both run in parallel. You are not splitting traffic from your existing campaigns. You are adding new coverage.
This matters because your existing campaigns stay stable and keep accumulating signal. The broad match campaign has its own learning period without touching what is already working.
How to set it up
Before you launch the broad match campaign, do this:
Check your conversion quality. Are you optimizing for real qualified events, or just form submissions? If Google is learning from cheap, unqualified conversions, broad match will find more of those. Make sure your conversion tracking reflects actual business outcomes, not just activity.
Build your negative keyword list before you go live. You should already have a core negative list from your phrase and exact campaigns. Port that over to the broad match campaign immediately. You are not trying to block all bad traffic on day one, but you should block the obvious categories: competitors, job seekers, informational queries that never convert, geographic terms you do not serve.
Set a realistic budget for the test. The broad match campaign needs enough daily spend to generate conversion volume within the test window. If you set it at $10 a day, it will not have enough data to draw any conclusions in six weeks. Size it so you expect at least 30-40 conversions over the full test.
Run it for at least four to six weeks. This is where most people bail early. Two weeks of data on a broad match test is noise, not signal. The algorithm needs time to figure out which queries actually convert. If you cut it at two weeks because the CPA looks high, you are probably looking at the learning phase, not the steady-state performance.
When to roll it out vs. kill it
After six weeks with meaningful conversion volume, look at a few things:
Does the broad match campaign's CPA come within 20-30% of your phrase and exact campaigns? If yes, that is a reasonable result for incremental volume. If it is 2x your target CPA and trending up, the signal is probably not strong enough to support it yet.
What are the search terms? If you are getting queries that are clearly off-target despite your negatives, add them and give it another two weeks. If you are consistently showing for irrelevant categories no matter how many negatives you add, the match type may not fit this account's niche.
Is the volume additive or cannibalizing? Check whether impressions and clicks in your phrase/exact campaigns held steady. If they dropped significantly while the broad campaign ran, you might be splitting the same traffic, not expanding it.
The short version
Broad match works when the algorithm has enough signal to use it responsibly. That means 30+ real conversions per month, clean conversion data tied to actual qualified outcomes, and a testing structure that does not contaminate your existing campaigns.
If you are below those thresholds, broad match is not the problem to solve right now. Build the signal first, then expand.
If your account is ready and you want to test it properly, the method is straightforward: isolated campaign, no experiments, solid negatives before launch, and six weeks of patience before you make a call.
Have questions about where your account sits? Hit me at admin@freak.marketing or join the list at freakmarketing.com for more Google Ads fixes like this.