In short
Use it when three conditions hold: your purchase tracking is accurate, you are bidding on conversion value rather than clicks, and you have enough conversions for the bidding to steer with. Under those conditions broad match finds demand that exact and phrase keywords never reach, because it reads the whole context of the account rather than the string. Without them it spends the budget on searches that look related and buy nothing. Keep it in its own campaign, keep brand out of it, and review search terms on a schedule rather than when something looks wrong.
Broad match now matches on context, not just on words
The match type stopped being a loose string comparison some years ago. What it matches on today is a reading of intent that draws on the landing page, the other keywords in the ad group, the products in your account and the search history of the person typing.
That changes the decision. The old objection to broad match was that a keyword for running shoes would collect searches about running clubs. The current risk is different and subtler: the system finds searches that are genuinely adjacent to your products but are earlier in the buying process, or aimed at a price band you do not serve, and those clicks cost the same as good ones.
Which is why the bidding matters more than the keyword. On value based bidding, the campaign is allowed to spend more on the queries that produce revenue and less on the ones that do not, and the breadth becomes a source of new demand. On manual bidding or maximise clicks, the same breadth is just a wider net at a flat price.
The practical test is whether the conversion signal reaching Google is the one you want optimised. If purchases are tracked cleanly with a value attached, broad match has something to aim at.
A structure that lets you use it without losing the account
Keep broad match separate. Give it its own campaign with its own budget, so that when it explores, it explores with money you allocated for exploration rather than money that was funding your proven terms.
Inside that campaign, keep the keyword set small and the ad groups tight to a theme, because the other keywords in the group are part of what the matching reads. A group with fifteen unrelated broad keywords gives the system a muddled picture of what the page is for.
Attach the account level negative list and the brand exclusions before it goes live. Brand queries are the fastest way for a broad campaign to look successful while adding nothing, and they will be picked up otherwise.
Then set the review rhythm: search terms weekly for the first month, monthly afterwards. Add negatives for the patterns rather than the individual queries, because broad match will generate a long tail of near identical searches and excluding them one by one never catches up. When a broad term proves itself repeatedly, promote it into the exact match campaign where it can be bid deliberately.
When phrase and exact are still the right choice
Stay tighter when conversions are scarce. A campaign producing a handful of orders a month cannot support a bidding strategy that needs to learn which of thousands of query variants are worth buying, and the exploration cost falls on a budget that cannot absorb it.
Stay tighter on regulated or specification-driven catalogues. Parts that fit a particular vehicle, sizes that fit a particular machine, ingredients that matter clinically: in those categories an adjacent query is not a softer version of the right query, it is the wrong product, and the return costs you more than the click did.
Stay tighter where margin varies sharply across the catalogue and the campaign cannot see it. If the value you send is order revenue rather than contribution, broad match will happily find volume on your worst margin products and report it as success.
And stay tighter during a period you need to be predictable, such as a peak trading window. Exploration is worth funding in a quiet month and expensive in the four weeks that carry the year.
Send better signals and broad match behaves better
The quality of what you feed the bidding decides most of the outcome. Send conversion values that reflect what an order is actually worth to you, not a flat number for every purchase, so that the system can tell a high margin order from a low one.
Where your margins differ a lot by product, consider sending a value based on contribution rather than revenue. It changes what the campaign optimises towards, and it is usually the single largest improvement available to an account that has already got the basics right.
Feed first party data in as well. Customer lists let the campaign distinguish existing customers from new ones, which matters if your goal is acquisition rather than total orders. Enhanced conversions improve the share of orders that can be matched back to a click, which makes the signal denser without changing what you spend.
Then give it time. A broad campaign judged after ten days will look wasteful, because the exploration happens first and the efficiency arrives later. Judge it over a period long enough to include the orders it started.
How the three match types divide the work
- Exact match campaign
- Proven queries, highest bids, tightest ad copy
- Phrase match campaign
- Known intent patterns, moderate bids
- Broad match campaign
- Discovery, separate budget, value bidding only
- Brand
- Its own campaign, excluded everywhere else
- Promotion rule
- A broad query that converts repeatedly moves to exact
- Negative rule
- Patterns excluded at account level, not per campaign
An illustrative division rather than a template. What matters is that discovery has its own budget and that proven demand is not bought at exploration prices.
Related questions
Does broad match work with manual bidding?
It works far less well. Manual bids apply the same price to a query you know converts and a query the system is testing, so the exploration is funded at the price of your best traffic. If you are not on value based bidding, keep the keyword set tighter until you are.
How many negative keywords does a broad campaign need?
Fewer than you would expect if you exclude patterns rather than individual searches. Build lists around the recurring themes you do not serve, such as free, repair, wholesale or a competitor's product line, and apply them at account level so every campaign inherits them.
Will broad match cannibalise my exact match keywords?
The auction generally prefers the closest matching keyword you have, so an exact match term you own will usually take the query. Problems arise when the broad campaign has a much higher budget or a looser target, so keep the budgets and targets deliberate rather than equal.