Exact and phrase match in one ad group, or split them out?

Honestly, I'm not sure separate exact and phrase campaigns earn their keep anymore. Google has been loosening how close a match has to be for years, so the old argument for splitting (tight control over which queries trigger your ad) is weaker than it used to be.

For accounts under a few hundred clicks a month, one ad group with phrase match is simpler, and the search term report tells you most of what you need. Once a head term has real volume, though, pulling exact out into its own campaign still pays off, because you can bid it separately and see its CPA without the noise from looser variants.

That's also where I'd push back on the "just use broad" crowd. On high-value terms you want exact control over the auction. What volume do people actually use as the point to split?

Splitting head terms makes sense, but I'd decide it on conversions, not clicks. In e-commerce a term can pull plenty of clicks and still convert only a handful of times a month, and a bid strategy trained on that is mostly guessing.

The part people skip is the overlap. If exact and phrase sit in the same ad group, add your exact terms as negative exact keywords on the phrase side. Otherwise the two keep competing for the same queries and your exact CPA ends up blended with phrase traffic.

Agreed on conversions, and I'd boil it down to one test: can the exact term get roughly 30 conversions a month on its own? If yes, split it into its own campaign or ad group with its own bid target. If not, leave it with phrase and add the exact-as-negative setup.

The reason is that smart bidding needs data to learn. Split off a term that converts five times a month and you starve both sides. The exact campaign never leaves learning, and phrase loses the queries that were carrying it. Keeping them together with the negatives gives you the separation you want without throwing away the data.

One catch: the negative only stops phrase from showing on exact queries. It doesn't stop exact from picking up close variants, so check the search terms report for those before you trust the CPA split. And the 30 is a rough guide, not a rule. With thin margins, you'll want more volume before you split.

I'd push back on 30 as a blanket threshold, because it's really a smart bidding rule. If an exact term only converts five times a month, splitting it out can still make sense when that campaign runs on manual CPC or maximize clicks, since neither needs a learning history to work. What you give up is confidence in the numbers, because five conversions is thin for judging CPA. That's a different problem from starving the algorithm, and it's usually one you can live with while you collect more data.

I'd push back on the manual CPC part. Sure, it doesn't need a learning history, but it also gets no conversion signal, so you're setting bids by hand on a term that converts five times a month. At that volume, a couple of good or bad weeks can look like a real CPA gap and you can't tell which it is. Maximize clicks is worse, because it ignores conversions entirely and you end up judging the split on clicks, which is the trap Iris pointed at. I'd keep it in phrase with the exact negatives until the monthly count is high enough to read.

Fair, five conversions is too thin to judge CPA on. But I'd still split it out if the goal is seeing what it costs to hold position on the head term, because impression share and CPC are readable at low volume even when CPA isn't. Pooling 90 days of data also gives you more to work with on the CPA side. What's the target CPA you'd judge it against?