Account structure, now that the machine does the targeting

Paid search account structure used to be about control: one campaign per theme, one ad group per keyword, tight match types, manual bids. That structure was rational when a human set every bid.

With automated bidding it actively hurts, and the reason is not aesthetic.

Why fragmentation costs money now

Smart bidding learns from conversions. Split your account into forty campaigns and each gets a fortieth of the conversion data. Most fall below the volume needed to learn anything, and stay in a permanently worse-performing state.

Consolidation is not laziness. It is feeding the mechanism enough data to work.

What should still be separate

Campaigns are where budget and bidding strategy live, so separate when those genuinely differ:

Brand and non-brand. Always. Different economics, different intent, and blending them makes every reported number meaningless.

Distinct budgets that must not bleed. If a product line has a ring-fenced budget, it needs its own campaign.

Genuinely different targets. A campaign targeting £20 cost per lead and one targeting £200 should not share a strategy.

Geographies with different economics, where you would set different targets rather than merely different messages.

What should be together

Everything else. Keyword themes that share intent and economics belong in the same campaign, with ad groups organised for ad relevance rather than for bidding control.

The old instinct to split by match type is largely obsolete. Modern matching overlaps enough that the split creates internal competition without creating control.

The migration problem

Restructuring resets learning. A consolidated account performs worse for two to four weeks before it performs better, and this is the point at which most restructures get reversed.

If you cannot tolerate that period, do it in stages — merge the smallest, lowest-volume campaigns first, where the learning loss is minimal because there was nothing being learned.

The test for any structure

Ask: does every campaign get enough conversions to learn from, and does every campaign have a genuine reason to hold its own budget and target?

If a campaign fails the first test, merge it. If it fails the second, it did not need to be a campaign.