Think about it this way. Every keyword in the world could theoretically work — but some have a high conversion likelihood, and some have a negligibly low one.

If you're a plumber in Sydney, "emergency plumber Sydney" is worth a high CPC. "Plumber near me" is still worth a decent one. "Plumber" on its own? Who knows what they want. A definition? Photos? Someone to call right now? They haven't told you anything. And the broader you go from there, the closer you get to bidding on "bananas" or "shoes" — where the likelihood of converting into a plumbing job is essentially zero.

It's not quite zero, though. That's why display advertising exists. That's why billboards exist. In theory, if you could bid £0.0001 on the keyword "bananas", it might even be profitable.

This is actually why Google invented Quality Score in 2005. Before it, people were doing exactly that — bidding on low-competition random terms at near-zero CPCs, and even with terrible conversion rates it was still profitable. But users were getting served so many irrelevant ads that they started ignoring them altogether. Quality Score closed that loophole by forcing relevance — between the keyword, the ad, and the landing page.

KEYWORD WHAT IT SIGNALS VERDICT "emergency plumber Sydney" High urgency, location-specific, ready to call now. Bid high "plumber near me" Clear intent, local, likely comparing options. Worth it "plumber" Could be anything. Information, photos, definitions. No signal. Uncertain "how does plumbing work" Researching, not buying. Nowhere near a decision. Avoid

So what makes a keyword good?

Here's where it gets a bit nuanced — because the answer depends on whether you're using smart bidding or manual CPC.

  MANUAL CPC SMART BIDDING Who sets the bid You. You can price the uncertainty yourself — bid high on specific terms, lower on broader ones. Google. Based on a target CPA or ROAS — but it's guessing until it has enough data. Keyword scope Wider pool from day one. You control how much risk you take on each keyword. Must start tight. Only high-intent keywords until the account has enough conversion data. The catch Needs someone experienced at the wheel. But done well, unbeatable control. The learning phase costs real money. And the keyword list has to do the heavy lifting.

With smart bidding, your keyword list has to do the heavy lifting. Start only with keywords that are very likely to convert. Tight, specific, high-intent. Let smart bidding find its feet on those before you go anywhere near the uncertain ones.

What the data actually looks like

We audited a fencing company recently with around a $40k budget. Enough spend to make the pattern obvious.

KEYWORD INTENT LEVEL CPA "fencing contractor" Broad. Could be commercial, residential, comparing prices. ~$250 "fencing company Fairfax" Location-specific. Narrower intent. Lower "vinyl fencing company Fairfax" Material + location + service. Maximum specificity. ~$80

Here's the kicker: smart bidding wasn't prioritising that last keyword. It was spending on the broader terms and underserving the one that was actually working.

This is the pattern. More specific keywords — where someone has typed more words, signalling more intent — almost always convert better. The extra words aren't just descriptive. They're a buying signal.

The extra words aren't just descriptive. They're a buying signal.

Where to go from here

Start with the most specific, highest-intent keywords you can find. Get a CPA that works, then broaden your keyword scope if you need more volume.

What you don't want to do is go broader in your match types instead. That's a different problem entirely — and an expensive one. See our article on why broad match is a bad idea →

If you're feeling brave enough to try manual bidding, you'd have more flexibility from the start — you could add broader relevant keywords and simply bid lower on them to reflect the uncertainty. Not bananas. Just "fencing contractor" at a lower CPC than "vinyl fencing company Fairfax." But you'd want someone who knows what they're doing managing it.

Not sure if your keyword list is working for you?

We review accounts every week. 20 minutes and we'll show you exactly what's pulling its weight — and what isn't.

Get a free review →