Low-margin SKUs in Shopping: separate campaign per margin tier, or custom labels inside one?

Take a store with 300 SKUs, about 40 of them low-margin accessories. Those can look fine on ROAS, but the target that works for the apparel line would lose money on them. The usual fix is a margin custom label in the feed, then a priority setting or a separate campaign per tier. I think separate campaigns beat labels inside one campaign. Shopping bids toward a single ROAS target, and the cheap, high-converting items tend to soak up impressions and budget, so the higher-margin products get squeezed. Labels only help if each tier also gets its own budget and target. Has anyone split by margin tier and seen the high-margin campaign lose impression share?

Depends on whether your high-margin campaign is budget-limited or target-limited. If it's running out of budget, cheap items soaking up the impressions is a real problem, and separate campaigns with their own budgets would fix it. If it's sitting under budget with weak impression share, the more likely culprit is a tROAS target that's too tight for those products. Lost IS (budget) and Lost IS (rank) point to different fixes, so look at those before you restructure anything.

Agreed, check the Lost IS split first. Splitting only pays off if each tier gets enough conversions to learn on its own, and a 40-SKU accessory group may be too thin for that. How many conversions does it get in a month?

Thin data is a real problem here, but I'd check it at campaign level, not per SKU. The rough guidance for stable tROAS is around 30 conversions in 30 days per campaign, and a 40-SKU accessory tier can land well under that even in a busy store. If it does, keep both tiers in one campaign, use the margin label to split ad groups, and give the accessory group a looser target. Two split campaigns would both starve. The trade-off is that the budget stays shared, so this only works if Lost IS (budget) on the high-margin products is low.

A looser target inside one campaign won't work the way you'd hope. In standard Shopping, tROAS is set per campaign, not per ad group, so the accessory group can't get its own number. You're left with two real options. One is a shared target, which means accessories have to clear the same ROAS bar and may simply stop serving. The other is manual CPC on the accessory product groups, which lets you bid lower but puts you back to bidding by hand. Which one fits depends on whether you're willing to manage manual bids on those 40 SKUs.

Right, tROAS is set per campaign, so the accessory group can't get its own number. Whether the shared target is a problem comes down to one figure: the accessories' break-even ROAS. That's 1 divided by the gross margin on those items. A 25% margin means you need 4.0 ROAS just to cover product cost before ads, while a 50% margin breaks even at 2.0.

If break-even sits below your shared target, the accessories not serving is the system doing its job. You're not losing sales you can afford, and manual CPC on those 40 SKUs would only buy traffic you don't want. If break-even is above the target, the accessories lose money at any bid, so the real issue is price or margin, not bidding. Work out that number before you choose between the two options.

Before you do that division, keep in mind that gross margin only covers product cost. On cheap accessories, shipping, payment fees and returns can eat a big chunk of what's left. Say an item has a 25% gross margin, then 8% goes to fees and 10% to returns. You're really working with about 7%, which puts break-even ROAS near 14, not 4. That changes the answer a lot, and the accessories may be losing money even when they look fine on gross margin.

So my rule of thumb is to calculate break-even on contribution margin, and only keep a low-margin tier in the shared campaign if its break-even is within roughly 20% of the shared target. Anything further out gets excluded or handled separately. A shared tROAS can't protect margin on its own, and that small tolerance stops you from paying for sales that only break even on paper.

First thing I'd look at is what value Google is actually bidding on. If your purchase conversion sends gross order value, tROAS is optimizing on revenue that includes orders that come back weeks later. Returns only correct the numbers if you send conversion adjustments for them, so the 7% contribution you calculated isn't what the bidder sees. Conversion value rules based on the margin label, or a margin-adjusted value, would get the shared target much closer to break-even. How are returns getting into your conversion data right now, if they are at all?

Agreed on the value problem, but double-check the fix before relying on it. As far as I know, conversion value rules key off audience, location and device, not product custom labels, so a margin-based rule probably won't split by tier. The margin adjustment has to go into the value you send with the purchase event, so margin or contribution value per order instead of revenue.

The part people miss is that the tROAS target changes meaning once you do that. If you send profit as value, a 4.0 target on revenue at a 25% margin becomes roughly 1.0 on profit, so you have to rescale the number or the bidder will look like it's underperforming when it's fine. Returns still need conversion adjustments on top, and since they arrive weeks later, the target will drift until the lag settles.

Agreed on the rescaling, and there's a reporting problem on top of it. Once the purchase action sends profit, the ROAS column stops meaning revenue, and the apparel team will read it as a collapse. The fix is a second conversion action that records revenue and is excluded from bidding, so you still see true revenue ROAS while the bidder optimizes on margin. Also look at discounted and free-shipping orders, because they can push profit to zero or below and skew the values the bidder learns from.