Broad match in an online shop: when it works and what it costs

Broad match is a data question, not a setting

Whether broad match pays off in your shop is decided not by the match type but by the quality of the signals your bids are calculated from. According to Google, broad match is the default option, and ads can appear for searches that are related to the keyword without containing its wording. To match, Google also uses the user's recent search activity, landing pages, assets and the other keywords in the ad group (Google Ads Help). The keyword therefore only roughly determines where you take part in the auction. How much you bid is determined by Smart Bidding, and Smart Bidding is only as good as the conversion value it optimises.

My assessment: with clean, margin-based values, broad match can find reach that pays off. With revenue as the value, incomplete tracking or too few conversions, it buys expensive reach that can still look decent in the report.

What Google promises and what independent figures show

Google says Smart Bidding works best with broad match (Google Ads Help) and, for a switch from phrase to broad, cites on average about 25% more conversions with target CPA and 12% more conversion value with target ROAS. The basis is Google's internal data from 2020, and Google adds that more budget may be needed to reach these results (Google Ads Help). These are manufacturer claims about extra volume, not a statement about your contribution margin.

Independent analyses paint a more mixed picture:

None of these studies measures what matters to a retailer: additional contribution margin. They compare match types by ROAS or CPA. A high ROAS on broad keywords can just as well come from search terms your account would have served anyway through exact, Shopping or Performance Max.

Four prerequisites before broad match runs in a shop

This list is my assessment as someone who works in the account myself. It names no fixed thresholds, because those depend on assortment and basket size.

1. The conversion value is correct. Duplicate or missing purchases, the wrong currency, or gross instead of net distort every bid. How to check is covered in Checking conversion data.
2. The value is close to margin. With revenue as the value, Smart Bidding treats an item with a 10% margin and one with a 50% margin the same. Broad match opens up many new search terms, and that is exactly where this distortion becomes expensive. More in Margin as conversion value.
3. The campaign has enough conversions. Broad match spreads bids across far more searches. With thin data, the bid strategy guesses for longer.
4. Budget is not the limit. If a campaign is already limited by budget today, broad match is more likely to displace existing clicks than to bring additional business.

If one of these points is missing, I would not roll out broad match but fix the missing point first.

The checking pattern: covered or new

In a shop, broad match in a search campaign competes mainly with your own coverage through exact, phrase, Shopping and Performance Max. So the question is not whether broad keywords have a good ROAS, but whether they bring revenue on search terms that earned none in any channel in the account before. This is how I check it:

1. Export the search terms of the broad keywords. In the search terms report, show the "Keyword" and "Match type" columns, as the "Keyword" column is not visible by default (Google Ads Help). Filter on match type "Broad match".
2. Reconcile. Check each search term against three lists: your exact and phrase keywords, your brand terms, and the search terms from Shopping and Performance Max in the same period.
3. Split. Whatever appears in one of the lists is "already covered". The rest is "new".
4. Assess only the new part. Total cost, conversion value and contribution margin for "new" separately and set them against your break-even ROAS.

The pattern has one limit: according to Google, search terms with too little query activity are omitted from the report for data privacy reasons (Google Ads Help). You only see this remainder in aggregate as "other queries", not individually. How to read the report economically beyond that is covered in Analysing the search terms report economically.

Worked example, invented figures: suppose a shop's broad keywords cost €10,000 a month and bring €50,000 in revenue, a ROAS of 5. After the reconciliation, €35,000 of revenue and €6,000 of cost fall on already covered search terms. The new part is €15,000 of revenue for €4,000 of cost, a ROAS of 3.75. At a 25% contribution margin rate before ad spend, the break-even ROAS is 4. The new part therefore loses money: €3,750 contribution margin against €4,000 cost. For the covered part, the example assumes it would have sold anyway. At a 30% contribution margin rate, break-even would be 3.33, and the same new part would be just positive: €4,500 against €4,000.

If the new part is small or below break-even, broad match is mainly buying existing demand at a higher price, however good the overall ROAS looks.

Where phrase and exact stay

Exact and phrase are not redundant once you have broad match. According to Google, an exact match keyword that is identical to the search term takes priority over broad and phrase keywords and over Performance Max. Conversely, a Performance Max search theme identical to the search term beats a phrase or broad keyword that is not identical (Google Ads Help). Exact is therefore your tool for tying search terms of known value firmly to the campaign where you control their bid and ad copy.

From its data for target ROAS, Adalysis derives a sequence: start with exact, add the best keywords and search terms as phrase if the budget is not used up, and only add broad for extra volume (Adalysis). I consider that a sensible starting point in a shop, with one addition: broad only comes in once the four prerequisites are met, and it only stays if the checking pattern shows a positive new part.

For maintaining exclusions: negative keywords do not cover synonyms or singular and plural forms, and a negative broad match keyword only blocks a search if all of its terms appear in it (Google Ads Help). How to build lists without accidentally blocking revenue is covered in Negative keywords for shops. AI Max for Search campaigns goes one step further than broad match and is a topic of its own.

When this rule does not apply

The checking pattern assumes that Shopping or Performance Max already cover a large share of product searches. For a shop without Shopping campaigns, such as one selling services or configurable products without a feed, the covered part is small, and broad match has more room for genuinely new demand. Likewise, with a new range that has no search term history, there is little to reconcile against. In that case I would run broad in an experiment with split traffic and compare the contribution margin of both halves.

At the other end, the limit is firm: if you use revenue as the conversion value and your range varies widely in returns or margins, you should not test broad match until the value is fixed. The experiment would then only measure how well Google optimises on the wrong value.