Artificial intelligence (AI) has been taking over our news feeds for some time now, and in recent years automation on ad platforms like Google Ads has advanced rapidly.
Predictive AI and machine learning in paid media ads is not a new thing; in fact, we recognised the shift in this direction years ago and have been ensuring our clients are future-ready ever since.
That's why we’re launching a playbook outlining the available AI-assisted tools in Google Ads. It will cover what these tools are, our top tips for success, and how Croud has been using predictive AI and machine learning tactics in our campaigns to give our clients a competitive edge.
To give you a taste of what’s to come, this article will explore Google's Performance Max (Pmax) - a tool that represents modern AI development.
What is Performance Max?
Pmax is a single and objective-focused campaign type which accesses a wide variety of Google’s ad inventory, including YouTube, Display, Search, Discover, Gmail and Maps, in a highly automated way to effectively capture the right audience across different stages of the user journey.
How is AI used?
As mentioned, Pmax exemplifies modern AI development and is promoted as a fully AI-powered campaign type.
AI is used by the tool in the following ways:
- It optimises where and when your ads show up across a wide variety of Google surfaces and inventory, using AI to show ads on YouTube or Search or Display when it thinks performance will benefit
- Creative assets are automated to find the most effective combinations
- It continues to learn while capturing more data signals to self-improve towards business objectives
Overall, Pmax takes the attribution, budget and bidding conditions all into consideration to a level of complexity that surpasses human ability. Although this optimisation is on autopilot, it's important to note that human guidance is essential to ensure it hones in on the target faster, and more accurately.
How do we set up for success?
Starting with accurate human intervention is essential. To guide the algorithm, start by providing the right set up, clear business objectives and audience signals, including search themes and first-party data. It’s important we do this to ensure the machine focuses on the most relevant signals in finding the best-performing combinations effectively and efficiently, saving precious ad budget.
Pmax can be seen as a blackbox, and cannibalisation is often a challenge. There are ways we can manage this, however, such as brand exclusions set up and clever reporting/scripts, to make the best and the right use of Pmax.
Become a Performance Max pro
Our PPC team is always on the lookout to maximise the efficiency and performance of our Pmax campaigns, so from our learnings, we’ve gathered our best strategies and top tips to share. From optimising URL expansion to refining bidding strategies, implementing the below will bring your Pmax campaigns to the next level.
Keep an eye on your landing page reports
This is particularly important if your site doesn't have different domains for different markets. You might find your Pmax campaign is pushing pages for the wrong markets (or even the wrong languages).
Don't enable URL expansion in the following campaign types:
Zombie shopping - Spend on ‘zombie’ products should ideally be pushed through standard shopping campaigns. However, if Pmax needs to be used, you might find a lot of the spend through that campaign goes toward Dynamic Search Ad-type targeting against non-zombie product pages and category pages that other Pmax campaigns should be picking up. (You can learn more about zombie shopping in our guide here).
Even if you disable URL expansion, spend may go against other inventory that you may rather have picked up by your core PMax or other campaigns (e.g. YouTube, Google Display Network, discovery placements) if you decide to use Pmax over standard shopping.
Catchall Pmax campaigns - For the same reason as above, URL expansion shouldn’t be used for those campaigns which target a small number of miscellaneous/uncategorised products that don't currently fit anywhere in your wider structure.
Make sure your key brand search queries are covered with keywords that exactly match them:
This differs case by case, but you should be able to get a decent share of your branded search going through keyword targeting exclusively by making sure you have broad keywords that exactly match your top volume brand queries.
How aggressive your keyword bidding is against branded search will change from period to period, and the same can be said for Pmax campaigns bidding against the same placements.
With this, comes the risk that you won't be able to tell whether, for example, a sudden increase in ROAS was the result of a genuine improvement in performance, or just a case of much higher brand search cannibalisation. This will make it difficult to make well-informed adjustments to your targets.
It's worth noting that exactly matching keywords shouldn’t always take precedence, but this is your bet for keeping brand and generic somewhat segmented outside of doing something like using brand exclusions on Pmax in conjunction with a 'brand-only' shopping campaign.
Reliably bid towards margin:
Breaking out your Pmax campaigns by grouping products into different margin levels and bidding more aggressively against high margin products (e.g., targeting high margin products and low margin products in two separate campaigns) can help improve your overall margin. However, this isn’t the most reliable method of doing so as:
- this can lead to an unnecessarily segmented or siloed campaign structure, particularly if you need to layer it with other campaign segmentation (e.g., if you also need to keep campaigns separated by product category)
- high margin product purchases may not account for the entirety - or even the majority - of your high margin campaign's revenue, since the product ad that a user clicks on won't always be the product they purchase. Even if it is, it may be purchased alongside a number of other products that don't share a similar margin.
The most reliable way of pushing Pmax campaigns to prioritise profit/margin (or any other additional product-dependent value not reflected in the conversion data that's fed back to your bid automation) is to incorporate it directly in your tracking so that you’re reporting margin back to your bid strategy, rather than total revenue.
If you’re using SA360, this can be done by using the cost of goods sold attribute in your product feed, otherwise you can achieve the same through any other custom tracking setup, provided it can securely push sensitive profit data back to Google Ads (for example, Google’s Soteria solution).
A quick way of ensuring full product coverage:
- Go into the product report at account-level and apply the highlighted filters in the image below.
- You should then see a list of all products that aren't running in any of your campaigns.
- A note on the third filter - for accounts that have multiple linked Merchant Centre (MC) IDs, make sure you're excluding MC accounts that you're not using; the same goes for feed labels.
- You can then just take a look at the product type column - or custom labels/whatever segmentation you're using to break out your listing groups - to reference against your listing group set up and figure out what's blocking them.
Croud top tips for success
Get started on your journey to Pmax proficiency with these quick tips below:
- Reduce brand cannibalisation and overlap with other campaigns by relying on the brand exclusions feature, and funnel your brand searches through an additional standard shopping campaign.
- Increase quality control by regularly reviewing search themes, expanded URLs, placement reports, and AI-generated assets, and negate whichever you find irrelevant.
- Stop Pmax from serving irrelevant landing pages by using Page Feed to provide a list of URLs for Pmax to choose from, instead of opting in URL expansion.
- Use seasonality adjustments during major promotions/sales periods to ensure automated bidding adjusts as quickly as possible where sudden conversion rate changes are expected.
- Be smart with the campaign structure when it comes to performance-focused structure versus a product-focused one. For example, if profitability is key for clients, we can separate campaigns to target low and high margin products with more aggressive return on ad spend (ROAS) targets on the latter.
To conclude, Google's Performance Max is a tool that certainly showcases AI's prowess. Whilst this automation is impressive, human guidance remains crucial, demonstrated by a few examples of our best practice tips.
Have a go at implementing these strategies - we hope you find them useful! Reach out to our team for assistance or further discussions on Pmax, and stay tuned for more AI-powered PPC best practice in our upcoming playbook…
