Performance Max split by margin tier: separate campaigns or one campaign with labels?

Splitting PMax into three campaigns by margin tier only pays off if each one gets enough conversions to learn on its own. Bidding is set per campaign, so a 3x target on the 15% electronics tier and a 6x target on the 60% private-label apparel tier is a real difference. But a campaign doing 15 conversions a month will spend most of its time in learning mode, and you end up with three noisy campaigns instead of one stable one.

Below that kind of volume, I'd keep a single campaign and set the target to the break-even ROAS of the weakest tier you're willing to fund. Then use custom_label attributes in the feed to steer budget toward the products you want pushed. What conversion volume per campaign do you need before splits start paying off?

I'm not sure the custom label part holds up. In PMax the budget sits at the campaign level, so labels only split your products into listing groups inside one campaign. They can't push spend toward certain products, and with a single ROAS target the campaign will still favor whatever converts best on revenue, not margin.

Agreed, that's the part people miss. Labels sort products into listing groups, but the budget still sits at campaign level. The only way I know to get a campaign to favor margin is to send a conversion value that already reflects margin, so a 60% item counts for more than a 15% one. Without that, a 3x or 6x ROAS target is still just chasing revenue. Can your purchase tracking send a margin-based value, or is it revenue only right now?

The margin has to come from the purchase event itself, because conversion value rules can only scale by audience, location or device and have no way to see product margin. If you send gross profit as the order value, your old 3x and 6x targets stop meaning anything. You're bidding on profit-ROAS at that point, so you need to recalibrate from scratch. The catch is that a mixed basket gets a single blended value, which means the signal is only as good as the cost data you have at checkout.

Blended value is what I'd worry about most, but I wouldn't hold off on deciding until the cost data is perfect. Rule of thumb: if your highest and lowest margin tiers are within about 20 points of each other, stay on revenue and set a sensible floor. A 40% item and a 50% item don't differ enough to justify rebuilding bidding around a noisy profit number. Once the spread gets wider and you know the margins for most of the catalog, profit as the value starts to matter, because a 60% item really does return about three times the money of a 20% one.

Before switching, log the profit value as a secondary conversion action for a few weeks alongside the revenue one. Secondary actions don't steer bidding, so you can compare which campaigns look worth funding under each. If the two disagree a lot, your cost data is probably the problem, not the bidding.

The 20-point rule doesn't hold up for me. What matters is gross profit per sale as a ratio, not the gap in points. A 20% item and a 40% item differ by 2x in what each sale returns, and that alone is enough to make revenue-based bidding misleading.

Fair, the points were a shortcut, and at the low end they break down. A 20-point gap between 20% and 40% margin means each sale returns twice as much, while 70% against 90% is a much smaller difference in ratio terms. So the ratio is the right lens.

Even so, I don't think the ratio settles it alone. For me it comes down to how much of your revenue sits in the low-margin tier. If the 20% items are 5% of sales, revenue bidding still gets the overall picture mostly right, and a profit signal mostly adds noise on a small slice. If they're a third of your sales, a revenue target is funding the wrong products at scale, and that's when switching pays off. I'd look at revenue share by tier before worrying about the ratio.