Recently I had a consulting call with an Irish jewellery brand selling into the American market. Their average order value sits between $300 and $400, they stock around 2,000 SKUs ranging from silver pieces at roughly $150 up to items well over $1,000, and their gross margins are strong: 55-60% on premium pieces, up to 75% on lower-priced items.
On paper, it is a healthy business. But they had pulled back from Performance Max after a bad experience, were running only Standard Shopping and brand campaigns, and despite having strong Meta video creative, they were not using any of it on Google.
Here is everything I told them, and honestly, most of it applies to any high-AOV ecommerce store.
Test Your Meta Video Creative on Google Demand Gen
This was the quickest win on the table. They had proven video ads performing well on Meta and were not running any Demand Gen campaigns on Google. This is far more common than you would expect.
Here is the thing: the same video creative that works on Meta will often work on Google Demand Gen. The creative is already battle-tested. The formats are compatible. For vertical video, YouTube Shorts is an excellent placement. And the targeting options on Demand Gen are genuinely powerful.
For setup, I recommend staying within Google placements only: Google Discover, Gmail, and YouTube. Avoid extending to the Google Display Network; the quality and performance there are typically much lower, and it will drag your results down.
On targeting, the most reliable starting point is a 5% lookalike audience based on previous purchasers. It is consistently the best-performing cold audience for demand gen campaigns. Layer in relevant in-market and affinity audiences from Google's library on top of that. You should also build custom keyword segments. Duplicate each segment into two versions: one set to "purchase intention" and one to "searched for." The "searched for" version typically delivers better efficiency; "purchase intention" typically delivers more scale but sometimes at a higher cost.
On bidding, start with Maximise Conversions. You cannot use Target ROAS until you have at least 50 conversions, and Demand Gen often performs best staying on Maximise Conversions indefinitely. Do not switch just because the campaign has matured.
Give it time. I told this client to commit to at least 2-3 months before drawing conclusions; two weeks of data is not a meaningful test.
Consider Returning to Performance Max, Feed Only
This client had a bad experience with Performance Max and switched back to Standard Shopping. I completely understand that reaction, and I see it regularly. But the issue is almost always the setup, not the campaign type itself.
My recommendation was to consider returning to PMax, specifically with a feed-only setup. This is a meaningfully different product from what most people run when they say "Performance Max."
Feed-only PMax stays focused almost entirely on Shopping. You attach your product feed, skip all creative assets, and turn off asset optimisation so Google cannot auto-create assets from your website. Add up to 50 of your most relevant keywords as Search Themes to act as a compass for the algorithm. That is the setup.
The reason to choose this over Standard Shopping comes down to bid strategy. Standard Shopping locks you into Maximise Clicks when starting out, which optimises for traffic, not revenue. Feed-only PMax lets you start with Maximise Conversion Value from day one, meaning the algorithm is learning which clicks lead to actual purchases. For a store with a wide product range and a long conversion window, that is a significant advantage in the early stages.
One caveat worth knowing: even with no assets uploaded, Google has started occasionally spending small amounts on non-Shopping channels, possibly by auto-generating assets from your product feed or landing pages. This is typically less than 2% of overall spend. Monitor the channel performance report in the first 30-60 days. Once you have a Target ROAS set and the campaign has exited the learning phase, this issue largely resolves itself.
Brand Segmentation Strategy
When should you separate brand traffic from Performance Max? The answer depends almost entirely on your marketing mix.
If Google Ads is your only or primary advertising channel, keep brand inside PMax. Brand conversions help the algorithm learn, and you need as much signal as possible.
If you are a multichannel advertiser with meaningful Meta spend, and Google is only 10-20% of your overall marketing picture, then segmenting brand out of PMax makes sense. The brand conversions in PMax are not adding much incremental information if the majority of your customer acquisition happens elsewhere.
Here is how to do it properly:
Add brand exclusions in your PMax campaign settings.
Add brand keywords as negatives to the PMax campaign, including misspellings and variations.
Create a dedicated brand Standard Shopping campaign.
Add non-brand keywords as negatives to that brand Shopping campaign.
Do not skip step 3. This is the most commonly missed step, and it is critical. If you stop running Shopping ads on your own brand terms, Google will fill those auction slots with competitor ads. Always run a dedicated brand Shopping campaign when segmenting brand out of PMax.
ROAS Product Segmentation: The Four-Campaign Structure
This is where things become genuinely interesting, especially for a store with 2,000 SKUs and a wide range of price points and margins.
The most effective structure uses four campaigns based on historical ROAS and spend data.
High ROAS Campaign: Your champion products. Set Target ROAS slightly lower than your account average. The goal here is to spend more on your winners, not squeeze every last percentage point of efficiency out of them.
Low ROAS Campaign: Products spending above your cost threshold but performing below your ROAS target. Set Target ROAS higher than average to either improve efficiency or reduce spend.
Low Spend Campaign: Products with some data but not enough to categorise definitively. Run these on Maximise Conversion Value with no Target ROAS, with a daily budget set slightly above historical average spend. The goal is to generate enough data to graduate products into the High or Low ROAS campaigns.
Zombies Campaign: Products with zero spend at all. Also on Maximise Conversion Value with no Target ROAS. If you have hundreds or thousands of zombie products, test in batches of around 100 rather than dumping everything in at once. Manually select the products you most want to test first.
To categorise your products, choose a meaningful date range, typically 90 days with solid conversion data. Set a cost threshold of $50-100; below this, data is statistically unreliable. Then set your ROAS threshold based on your actual breakeven point. To calculate it: divide 1 by your gross margin. For 60% margin, that is 167%. For 75% margin, it is 133%.
Do not over-engineer this. If your margin variation between product groups is less than 30-40%, a single ROAS threshold across all products is fine. Margin segmentation only makes sense when the difference is large enough to materially change your bidding decisions.
When to Kill Products
Almost never. Google's algorithm will naturally stop spending on non-performers over time by reducing CPC bids for those products. Unless a product has accumulated significant spend with ROAS consistently below breakeven, the campaign structure is correct, and Google has not naturally reduced spend to zero, there is usually no need to actively kill products.
Most underperformers will die on their own. Your time is better spent moving winners into the High ROAS campaign and nurturing low-spend products into either category rather than pruning losers manually.
The Underlying Principle
Everything above comes back to the same idea: more signal, better decisions. Demand Gen diversifies your top-of-funnel. Feed-only PMax gives the algorithm revenue-optimised bidding from day one. Brand segmentation keeps brand conversion data clean when you are spending across multiple channels. Product segmentation means bidding decisions are based on meaningful data, not diluted by products with wildly different performance histories.
For a high-AOV store with a large catalogue, clean structure is not a luxury. It is the difference between accounts that scale and accounts that plateau.
Conclusion
Here are the key takeaways from this consulting call:
Test your best Meta video ads in Google Demand Gen campaigns. The same creative often works. Use 5% lookalike audiences based on purchasers as your primary cold audience.
Give Demand Gen at least 2-3 months before evaluating. Stay on Maximise Conversions; there is often no need to switch to Target ROAS.
Consider feed-only PMax over Standard Shopping for new campaigns. It gives you Maximise Conversion Value from day one instead of being limited to Maximise Clicks.
If segmenting brand out of PMax, always run a dedicated brand Shopping campaign to prevent competitors filling your brand auction slots.
Structure products into four campaigns: High ROAS, Low ROAS, Low Spend, and Zombies. Each uses a different bid strategy to match where the product is in its lifecycle.
Rarely kill products manually. Google reduces spend on non-performers automatically over time.