Demand Gen vs Performance Max for ecommerce

Demand Gen and Performance Max both use Google AI and can both show your products, so the obvious comparison is channels and creative. For a shop with hundreds or thousands of products, the more useful question is which products each campaign type should carry, and which return counts as good for them.

If you already run Performance Max and are weighing Demand Gen, start from Google Ads Help’s own descriptions, then read both against each product’s break-even ROAS, which you can check with the break-even ROAS calculator. Which campaign type sells more? Google’s pages don’t say, and we can’t either.

1. What is the difference between Demand Gen and Performance Max?

Performance Max is built around a conversion goal and serves across all of Google’s inventory. Demand Gen is built around image and video assets and serves on Google’s visual surfaces. Google’s Demand Gen page puts the contrast this way: “Performance Max helps you find more converting customers across all of Google’s channels (for example, Search, Maps, Shopping). Demand Gen captures engagement and action across Google’s most impactful surfaces.”

Sources: Google Ads Help, About Performance Max campaigns, About Demand Gen campaigns, Use a product feed in Demand Gen and Google Merchant Center Help, Using data sources in Performance Max.

Two rows matter for a shop. Both can use your Merchant Center feed, so both can show the same products. And both take a goal and a bidding setting that you choose, so the return you ask of each is your decision. Nothing in the table tells you which products should go where.

2. Where a Search campaign fits

A keyword-based Search campaign is the third option people compare. Google says Performance Max can complement keyword-based Search campaigns, and that a Search campaign with an exact match keyword is prioritised over Performance Max when a query hits that keyword. For standard Shopping, see how to optimise Google Shopping ads.

Which of the three shows a product depends on keywords, your feed and your assets. How much return that product needs does not change with the campaign type: it follows from the price and the margin. So compare campaign types by product, and leave the channel list to Google.

3. Margin decides profit, whichever campaign shows the product

Neither Google page tells you whether a product makes money, because that depends on its margin. Break-even ROAS is 1 ÷ contribution margin, so the margin sets the return a product needs, whichever campaign type shows it. Performance Max lets you set an optional ROAS target, and that number should come from the products’ break-even rather than the account average.

The profit equivalent is POAS, which is ROAS × contribution margin. A POAS of 1.0 is break-even for every product (POAS vs ROAS). What is a good ROAS works through the numbers in more detail.

4. Judge both against each product’s break-even

One account target treats every product alike. A target ROAS of 300% looks sensible until you set it beside three products with different margins.

The same 3 return is a profit on one product and a loss on another. Per €100 of ad spend, it brings €300 of revenue, and what is left depends on the margin.

This applies to Performance Max and to Demand Gen alike. If you set a target, set it per group of products from their break-even, and see how to set a target ROAS from your margin for the method. The product-level view of Performance Max is in Performance Max for ecommerce.

5. Which products go where

Google’s pages don’t say which products suit which campaign type, so the rules below are our working view, not Google’s. Start from what you can read. Performance Max already spends on products that get clicks, so products with a clear return are easy to judge there. Demand Gen is a place to test products you can’t read yet.

Decide what you will judge the test on before it starts. Google’s Demand Gen page says value-based and Maximize Clicks bidding “optimize conversion across the full funnel to drive consideration goals such as site visits”. A test that bids for clicks gives you clicks and site visits to read. To read it against break-even, you need a conversion value goal and conversions that are recorded.

Keep the test small enough that you can read it, but not tiny: Google’s Demand Gen product feed page recommends at least 4 products, and 50 as best practice, to cover its ad placements. The result is a verdict on those products against their own break-even, and says little about Demand Gen itself.

6. Split them with a custom label

A Merchant Center custom label is how you make the split. Google’s Performance Max help on listing groups says listing groups are defined by attributes in your product data, including custom labels, and recommends using custom labels to group items. For Demand Gen product feeds, Google’s page says you can filter by brand, product ID, condition, product type, or create a custom filter. Check which filter your account offers for your label. See custom labels.

Plan the label before you build the campaign. Google’s Merchant Center page says a Performance Max campaign’s data source can’t be added, changed or removed after creation, so the feed you choose at the start is the one it keeps.

For the split above, Product Segmentation writes each product’s segment, Ghosts and Dogs included, to a custom label through a supplemental feed, only for the labels you approve. Filtering the Performance Max listing groups and the Demand Gen product filter on that label is your step, as shown on the Performance Max page.

7. Measure before you judge Demand Gen

Google’s Demand Gen page asks you to set up conversion tracking on your website and make sure the Google tag is working properly. Check that first.

An unmeasured conversion is a sale the campaign made and never got credit for. If tracking records only some of what happened, reported ROAS falls below the real return, and a profitable product reads as a loss.

That is arithmetic and applies to any campaign type, but it matters most for a test campaign you are about to switch off. Check what you record before you judge: Full Signal Tracking sends conversions back to Google Ads server-side. The ad blocker conversion tracking test runs in your own browser, and the SDK debugger decodes a request.

To measure whether the test campaign added sales at all, set up an incrementality test before you start.

When the test group is measured and still sits below its break-even, lower the priority of those products or take them back out. When it sits above, move more products in with the same label, a few at a time, and read the result again after the conversion delay.

Your first test

Work out break-even ROAS for your five biggest products and see which segment each sits in; the free analysis in Product Segmentation does that on your own products. Then pick one small group, such as your Ghosts, for a Demand Gen test, and judge it against its own break-even once its conversions are measured.

Keep a dated note of the change, so that later you can tell whether the product mix or your own change moved the return.