Around 2 years ago we had a prospective client ask for an audit and a proposal detailing our approach to his campaign structure. We ended up losing out on the contract because he didn’t want to use Pmax campaigns because in his words..”well, what are you going to be doing because Pmax does it all for you – and you’re not needed?” 

I suppose it’s a reasonable assumption to assume your PPC agency isn’t doing anything if you’re running Pmax but anyone who runs Pmax in volume knows that these campaigns have wildly different results which can change at any given time. 

Pmax – assets or no assets?

In our previous blog Is your PPC agency policing your automated assets? We looked at a recent update in Google which saw accounts that were previously opted out of Automatic Asset Creation at account level, opted back in. 

This happened on or around the 7th of May this year. 

The result was Pmax campaigns running with unauthorised assets – text and images. 

We run lots of Pmax campaigns, some with assets, some without and there isn’t any definitive answer to which is best – although personally, I prefer no assets to having to avoid policing the shopping report on a weekly basis. 

The summary here is – should you run with assets? Yes and No. 

What do we test in Pmax as a PPC Agency for our clients? 

These are the top tests we run in Pmax and we often  use a combination of one or more to assess the uplift in the desired key metrics. 

Assets

The most commonly tested format in Pmax. Set up is simple but you need to create this in a new campaign as trying to remove assets from an existing asset group won’t work. What you’re looking to see is if channeling more spend through the PLA listings drives an uplift in sales or roas – whatever metric you measure as a PPC agency or retailer 

Audiences and FPD (First Party Owned Data) 

First Party Owned data is the data that you own as a business and is GDPR compliant. This may be email lists or audience lists from analytics as two examples. Audiences can also be added to a Pmax and a common audience is In Market Audiences in Google Ads. I have searched for a taxonomy of all In Market Audiences without success, but you can search in here and find like-for-like audiences like North Face, Barbour and some real beauties. Adding or removing data segments can help you identify any uplift in ads performance 

Inventory Volume Vs Segmentation 

Pmax likes lots of inventory. This is an ideal test for retailers with smaller budgets and smaller inventory. Pmax cycles the same products over and over for easy results on the platform and giving it more inventory does help maintain a good ROAS. But for larger retailers with more SKUs and the need to manage budgets across the inventory, account segmentation is key. We have many Pmax campaigns also segmented by gender for budget control to push between M/F

Most people assume that this cyclical behaviour originates in the ads account, but this is actually learned at Merchant Centre level 

Flip the Merchant Centre 

On the previous point, Pmax likes to cycle the same products over and over so it produces nice easy wins. The downside is that you tend to sell the same products over and over. You may start to assume they’re your best sellers. 

It’s not the Google ads account that cycles these products… It’s the Merchant Centre. What you can test here is adding an additional MC into the ads account and link this to a Pmax to serve ads without a historic performance to drive the inventory serving in the PLAs 

Zombies 🧟‍♀️

Technically this is more of a feed optimsation than a Pmax test. But I mentioned the Merchant centre so this is basically the same thing. The old 80/20 rule says that 20% of your inventory is receiving 80% of the clicks. This would be great if it were true, but the data suggests that it’s even lower than that. Creating a Zombie Custom Label allows you to set up a rule in the feed that if a product receives <less than = 20 clicks in 30 days it is added to the line list called zombies. (example numbers) Then those products will feed into a campaign which will force the inventory to serve. 

Brand exclusions and negative keywords 

Brand exclusions and negative keywords are broadly similar, but you may need to opt for adding negative keywords to the account as sometimes the brand isn’t in the list to exclude. This will allow you to shift away from a known affinity audience (even your own) and test new waters

Optimise for New Customer Acquisition 

Great test to reach new customers. This doesn’t mean you can’t include your FPD- it will be used as an observation audience rather than an acquisition based customer list. Typically we see a lower Roas in these campaigns, but that’s not a bad thing. Typically these campaigns will serve to unknown audiences who do not have an affinity with your brand. 

Product feed optimisation testing 

This is a whole separate entity and is mentioned here as a ‘just because’

The success of the Pmax is heavily dependent on a high quality optimised feed and the variants you can include in a test are fairly extensive. Using a reputable feed optimsation software such as Shoptimised is a smart move.

Bidding strategies test

Target ROAS, Maximise Conversions, Target CPA and so on. It’s a fairly simple application of bidding strategy which you can add as a multiplier into any of the above tests to see if you get any uplift. Target ROAS is like an overblown balloon when you try to scale past your target ROAS simply because Google runs out of the high quality ad inventory for the 9% of your products it likes to serve and you start getting placed in the soggy end of the ad market to spend the budget and that’s where you see your ROAS tanking. If this is the case – it’s time to segment your campaigns and optimise with all of the above. 

Margin based bidding / campaign structure 

Not a difficult concept, but it does require some groundwork and investment into your feed. Again, using a solid platform like Shoptimised is a smart move and will allow you to create these rules at scale. This test means you can allocate both a bidding and budget test to products based on their profit margins, allowing for greater scope in testing your budget, safe in the knowledge that the ROAS should, in theory, be decent!

Here’s a non-exhaustive list of other ideas to push the Pmax barn door: 

  • Geo or location based testing 
  • Asset testing
  • Seasonality testing 
  • Landing page optimisation
  • Campaign goal testing

You can include the test in your naming convention and please do keep a record on a spreadsheet or the back of a fag packet as to what’s running and what the impact has been so you can leverage this for future learning. 

Back to that prospective client who suggested we wouldn’t do any work if we ran Pmax campaigns… 😱🚩

Who wants a really sneaky number 11? 

 

  • For the brave ☠️

Google runs on cues, yes? A major cue for Google to benchmark your performance is conversions. Well, you can test (with your client involved for goodness sake!!) accidentally doubling the conversion tracking. See how reporting double, even treble the conversions lowers your CPCs. Just saying.