Case study
Broad match has a bad name in this trade: it supposedly burns budget. The safe reflex is putting everything on exact. But safe isn't free, because it leaves volume on the table. Instead of debating it, we measured it: broad and exact simultaneously, in the same 132 ad groups.
3x
more volume from broad than from exact match
-17%
lower cost per order on broad (EUR 7.04 vs 8.44)
132
ad groups, broad and exact measured simultaneously
Challenge
Broad match is known as the match type that eats your budget on irrelevant search terms. Many accounts therefore stick to exact and phrase: you only show up on search terms you thought of yourself. Safe, predictable, and limited, because you never reach the variants that weren't on your list.
So the question is not whether broad sometimes wastes, but whether on balance it delivers more than it costs. And that's not a matter of opinion. That's measurable.
Approach
In 132 ad groups, broad and exact match ran side by side at the same time, on the same themes, in the same measurement window of almost two months. That's the core: no before-and-after in which the season or the market can shift, but a simultaneous comparison in which the only difference is the match type. Whatever one does more or less than the other comes down to the match type and nothing else.
Judged on what counts: cost per order and return, not clicks or impressions. And with conversion-based steering plus negatives underneath, so broad can do its job without roaming unchecked.
Result
Broad delivered roughly three times as much volume as exact: more clicks, more orders. And not at a higher price, but at a lower one: cost per order on broad came in 17% below exact (EUR 7.04 versus EUR 8.44), at the same ROAS. So broad opened up reach that exact simply doesn't get to, without hurting profitability.
The win is in the total, not in every single ad group. Where there was enough data per ad group to compare them one on one, the picture was mixed: sometimes broad won, sometimes exact, sometimes a draw. On the whole, broad wins clearly, because it opens up volume that exact doesn't reach at the same profitability. And because broad and exact ran simultaneously in the same ad groups, this is no seasonal or period effect; the difference is the match type. The lesson is not "put everything on broad", but "don't rule out broad on reputation, measure it".
Best practice
Broad's reputation dates from a time without smart conversion-based bidding. Back then, targeting broadly really was gambling. Now bidding steers on your actual conversions, and broad becomes a way to find new, profitable search terms you would never have come up with on a manual keyword list.
The condition is the safety net: reliable conversion tracking and a maintained negative keyword list. Without that, broad does indeed roam. With that net in place, it's often your cheapest source of new volume. Ruling out broad on principle means taking the safe road and leaving growth on the table that can simply be measured.
Self-check
Not a single broad keyword? Then you're choosing certainty over reach. You capture exactly the search terms you already know, and miss the variants you hadn't thought of yet.
Your conversions have stopped growing, not because demand has run out, but because your keyword set is exhausted. Exact match can only harvest what you've already sown.
Often the problem wasn't broad itself, but broad without proper conversion-based steering and without negatives underneath. With that safety net it's a different story.
Decisions made to avoid waste rather than to find growth. Sensible if you're flying blind, a shame if you can simply measure it.
Broad is no miracle cure and exact is not wrong. The point is that the choice should be made on data, not on a reputation from years ago. Measure them side by side and the account will tell you what works.
In a free account scan I check whether your match types and structure are holding back growth, and what there is to gain.
Findings within 5 working days. If there is little to gain, I'll tell you that too.
