An accessories brand ran 29 Google campaigns and none of them could learn.

The headline finding was not an efficiency problem. The account was acquiring customers at a healthy margin and had been for over two years. The problem was that it was acquiring fewer of them every month, and the Google structure was a large part of why.

A fashion accessories brand spending around $75k a month, roughly half of it on Google, had built itself 29 active campaigns on the Google side. On a budget that size, that is far too many. The algorithm learns from conversions inside each campaign, and when you spread the conversions that thin, no individual campaign sees enough of them to build a useful model of who buys. The account was spending like a serious business and learning like a hobby one.

Underneath the fragmentation sat two compounding mistakes. Around 35% of the Google budget was going to cold search before Shopping was anywhere near saturated, and every campaign was running a ROAS target that optimised the business straight into a shrinking pool of customers. Each one made the cohort decline harder to reverse.

The pattern.

Twenty-nine campaigns on roughly $42k a month means the average campaign is working with a budget too small to ever exit learning. Some were seeing a handful of conversions a month. A campaign on five conversions a month cannot find a pattern in anything. A campaign on fifty can build a genuinely accurate picture of who converts and at what cost. The account had chosen the first situation twenty-nine times over.

The commercial cost of that is not abstract. When the signal is fragmented, each campaign underperforms what a consolidated version of it would deliver. The account was spending like a $42k account and producing results closer to what a well-structured $25k account would. The other $17k was being paid to the structure, not to the customer.

Why it was happening.

Over-segmentation usually starts as a reasonable instinct. Someone wants control, or visibility, or a clean read on a product line, so they split a campaign out. Do that enough times and you end up with a structure that looks organised and behaves like noise. The brand had segmented by region and by product category, neither of which was earning its keep. The two adjacent markets had no regulatory or catalogue reason to be separate, and the smaller one would have learned faster sitting on the larger one's conversion data.

The bidding made it worse. Every campaign was running a target ROAS, which sounds disciplined and is, right up until it becomes the thing capping your growth. A ROAS target tells the algorithm to chase the warmest, most efficient buyers it can find. Over time the campaign concentrates on a narrow slice of the market, exhausts it, and plateaus. New customer volume falls while the efficiency number stays flat, because the system is getting very good at re-finding the same people.

An efficiency target will always constrain volume first. The account looked stable because it was efficiently finding the same converters, not because it was finding new ones.

What it was costing.

The single largest efficiency loss in the Google account was the cold search allocation. For a fashion accessories brand, Shopping is the strongest placement by a distance. When someone searches for a product, the Shopping result shows them the image, the price, the discount, the review stars and the delivery estimate before they click. They qualify themselves on price and product first, so the click is worth more and converts harder. Cold search gives you none of that and costs more per click relative to what it returns.

Putting 35% of the budget into cold search while Shopping was still unsaturated meant the account was spending more to capture colder, lower-intent demand while leaving the cheaper, warmer demand on the table. That is the wrong order. You saturate the placement that converts best before you fund the one that converts worst.

The fix.

• Consolidate the campaigns. Collapse the 29 down toward five to seven, so each one carries enough conversions to learn. Merge the two regional splits, the smaller market benefits from the larger one's data and there is no reason to keep them apart.
• Reorder the spend. Saturate Shopping before cold search gets meaningful budget. Cold search expands only once the account is past the $50k a month mark and Shopping is genuinely maxed, neither of which was true yet.
• Rotate the bidding. Move campaigns into an expansive phase with no ROAS target to let the algorithm discover new audiences, then reapply the target once it has a richer dataset to optimise against. Discovery first, efficiency second, repeat as it scales.

Segmentation, where it is warranted at all, should follow margin profile rather than product category. A high-margin line can afford to bid more aggressively and accept a lower apparent ROAS because the margin carries it. Splitting by category alone, when the categories share the same buyer, just fragments the learning again under a tidier label.

What changed.

The brand became a client off the back of this audit. The diagnosis held: this was never a cost problem dressed up as a growth one, it was a structure problem throttling volume while every efficiency metric stayed green. Consolidate the campaigns, fix the order of spend, and rotate the bidding to rebuild audience discovery, and the same budget starts buying the new customers the structure had been quietly suppressing. The efficiency was already there. The job was to stop the account from spending it on itself.