The B2B match type framework I use across $650K/month

If you mix exact, phrase, and broad match in the same ad group, you've surrendered control of your account.

I'm not being dramatic. The mechanics of how Google handles overlapping match types means that mixing them creates unpredictable auction behavior, cannibalizes your own bids, and makes it impossible to know which match type is actually driving performance.

At $650K/month across two MCCs and five brands, ambiguous match type behavior is expensive.

What match types actually do now

The definitions have changed significantly. Google has quietly expanded what each match type captures:

Broad match (not used in any of my accounts): Can match to queries that share a vague semantic relationship with your keyword. "Safety labels" might match to "how to label things safely at home." It will. And you'll pay for it.

Phrase match: Once strict. Now functions closer to the old "broad match modifier." It respects word order somewhat but allows significant expansion. "Arc flash labels" in phrase match will match to "OSHA arc flash warning label requirements" and "buy arc flash labels online" but also starts matching to variants you didn't intend.

Exact match: Used to mean the query had to match exactly. Now allows "close variants" that Google considers semantically equivalent. "Dot-peen marker" in exact match might serve for "dotpeen marker" or "dot peen marking." Usually fine, occasionally problematic.

The framework

Tier 1: Exact match, uncapped budget Proven keywords with documented Salesforce pipeline contribution. These are your core performers. You know they drive closed deals. Give them all the budget they can productively spend.

These live in tightly controlled SKAGs or near-SKAGs where the ad copy is specifically written for the query.

Tier 2: Phrase match, capped budget, strict negative control Query mining territory. You're using phrase match to discover what real buyers are searching. These campaigns have significant negative keyword coverage and a monthly budget cap so they can't cannibalize Tier 1 spend.

Every query that proves itself in Tier 2 gets promoted to Tier 1 as an exact match with its own ad group.

No Tier 3: Zero broad match Broad match is off in all of my accounts. Has been for years. The performance degradation in niche B2B is severe enough that the claimed benefits (discovering new queries, more volume) don't come close to compensating for the waste.

If I want to discover new queries, I use phrase match with an aggressive negative structure. I get the discovery benefit without the chaos.

The cross-campaign negative structure

Here's the critical piece most frameworks miss: if you're running both exact and phrase match campaigns, you need to add all your exact match keywords as negatives in your phrase match campaigns.

Why: if you bid "arc flash labels" as both exact and phrase match, Google might serve either campaign for that exact query. If exact match has a lower bid (because you're managing them separately), phrase match might win the auction despite having weaker ad relevance.

Adding your exact terms as negatives in phrase campaigns forces all exact-query traffic to your Tier 1 campaigns. Clean data. Clean attribution. No cross-campaign cannibalizing.

Set this up once. Then automate it with a script that adds new exact match terms to the phrase campaign negative list whenever you add them to Tier 1.

Why this beats "let the algorithm sort it out"

Google's recommendation is to mix match types, use broad match with Smart Bidding, and let the algorithm figure out which queries to serve.

The algorithm will figure it out. Over time. While burning significant budget learning things you already know from 12 years of experience.

In niche B2B with limited conversion volume, the algorithm doesn't have enough data to learn your vertical. It will make mistakes for months. The framework above shortcircuits that learning period by applying human expertise directly.

At $650K/month, months of suboptimal learning cost hundreds of thousands of dollars. The manual framework is worth the maintenance overhead.