Same clicks, same orders, three times the price

Gibbon's Canada account sits inside the North America profile. For the thirteen weeks to 10 September, that profile came in at 28% ACoS, and for the category it is in, nothing about 28% asks for attention. Nobody was going to open it that week.

Our automation engine opened it anyway.

Question Asked: Are there any opportunities in the account to improve?

It breaks every profile down by marketplace, campaign and ad format every day. It doesn't care that the roll-up looks fine.

Canada on its own was 51%. The auto campaign inside Canada was 113%. The easy read was sitting right there. Auto targeting doesn't work in Canada, pause the campaign, move on.

Before anybody could do that, the Sellerbridge engine did the bit that normally eats half a day of an analyst's life. It went and found the closest comparable campaign anywhere else in the account: same ad format, same targeting type, same product line, same thirteen weeks.

CPC didn't match

Canada took 3,390 clicks and made 61 orders. The US took 3,223 clicks and made 60. Two campaigns doing the same job at almost exactly the same rate. The only thing that differed was the price of a click. $3.42 in Canada. $1.19 in the US. There was nothing left over for targeting to be blamed for.

It was the bid

And the reason was structural, which the engine also put in front of us. Canada was running one auto campaign across every product line, so it carried one bid. The US was running three, split by product, each with its own. So we took the Canadian bid down toward the US level and split the campaign the same way.

So we took the Canadian bid down toward the US level and split the one campaign into three, one per product line, so each of them carries its own bid now, and on the account's own numbers, if conversion and order value hold where they are, that should put Canada somewhere around 39% which is exactly where the US campaign has been sitting all along.

Sixty-one orders is a thin base and honestly we would have liked another month on it before calling anything, but the click price was measured across 3,390 clicks, and no amount of thin order data makes a click cost three times what it costs next door. And we are not doing that comparison by hand again: every profile now gets broken down by marketplace, every campaign gets matched against its nearest twin somewhere else in the account, and when the gap turns out to be the click price and not the conversion, it reaches us with that already written on it.