If you’ve created a Performance Max campaign and resisted the urge to make changes during the learning period, you’re now ready to optimise for conversion volume, value, and quality.
This guide on optimising Performance Max is split into three sections:
- Level 1 optimisations: Your basic, everyday, salt-of-the earth tweaks
- Level 2 optimisations: More advanced changes for dedicated practitioners
- Frequently asked optimisation questions and answers
If you’re not ready to optimise or want help with creating a new campaign the right way, check out my Performance Max setup guide.
I’ve managed over $250 million in ad spend over 13 years through my agency, and I’ve created Google Ads training and communities used by more than 3,500 PPC managers and agencies.
So I know a thing or two about this stuff…
Advanced pros, let’s kick on.
What’s lacking in Performance Max
Aka what Google still won’t let you do in Performance Max:
- Add or exclude keywords yet. I say yet because I hope that will change at some point (you never know).
- Set manual bids. But quite frankly, we don’t need them. You can’t keep up with the machine here.
- Adjust asset group bids. This is irksome. Hopefully, they bring that in because as it stands, we have to split by campaign to set different bidding strategies.
- Exclude audiences. Now it’s getting absolutely ridiculous. Come on, Google; we need that.
Hands-off approach to optimisation
Performance Max doesn’t like being fiddled with. Changes need to be less frequent and significant unless you have loads of data. Plenty of impressions, spend, conversions, clicks, etc. is always the “get out of jail free” card. If you’ve got thousands of conversions a week, you can make frequent changes.
Otherwise, steer with your feet a.k.a. hands off the wheel. And keep asking yourself:
- Is Performance Max still working?
- Is it hitting my targets?
- Is it spending my full budget?
If so, you don’t really need to do anything; don’t feel like you have to do work. One thing you should do is improve poor assets. Again, there’s a right and a wrong way to do that (more on that further down).
Shortcomings in reporting
Reporting is not great in Performance Max. You can’t see cost per asset group (which is a joke) or network splits (also a joke). But there is a workaround.
You can use the Google Analytics attribution report and handball this reporting. So you can see the splits in Performance Max, but it’s not ideal. Let’s face it: The only thing you can really do with that data is split your campaigns out, or make some changes that hopefully guides Performance Max away from some issue, like overspending on Display.
The data is there; Google just isn’t giving it to us inside Google Ads, which I think is a bit of a piss-take. Hopefully, we do get network split reporting and metrics at the asset group level at some point.
Why is this really important? If I knew Video placements were spending really hard, let’s say half my budget, then I’d want to focus on video production. But those videos might not be converting so well, and video production isn’t a 10-minute job. It’s a hell of a lot easier to create text and image-based assets, so I’d want to know that.
Conversely, what happens if Performance Max is not spending on Video while I’m wasting my time and money creating video assets that won’t be useful to my account anyway?
While it’s really frustrating, you’ve got to focus on what you can see. I’ve just told you about what you pretty much can’t.
Level One Optimisations
A higher budget sends Performance Max a signal that it has some wiggle room. So even if it’s not spending it today, you’re giving it headroom. If you’re budget-limited, and below your target CPA or above your target ROAS, really consider adding budget to unlock some volume.
Or you could simply reduce your target CPA by a small portion, say 10%. That way, you don’t keep hitting that budget ceiling. The same goes for target ROAS: Increase it a notch so you get more for your money.
Ideally, you just add a bit more budget to it to scale it up a bit, but I know that that’s not always an option.
You can use Google’s recommendation for budget increases; the only downside is they’re usually multiples and not increments. If you can afford a 300-500% increase with the expectation that ROAS and CPA might not keep up, go for it. Otherwise, use smaller increments and take steps towards your goal without fast budget burn.
Do we care about ad strength? Yeah, we do. Just remember: Good is good enough.
Ad strength is quite arbitrary. You can toss any old stuff in, and as long as you fill out every field, you’ll get a higher score. It’s not that smart, really. Understand how that score is manipulated and what it actually means for you. A higher ad strength does not mean better performance, only that you’ve given the machine more assets.
More assets is good, because it is a machine learning system, so give it more good things to use. But if you can’t say something worthwhile or add a strong creative asset, I’d park this optimisation until you can.
Testing ad text
I’ll keep this one simple. When it comes to ad copy, there are three key elements to test:
- Different call to action
- Different offer, with or without price
- Benefits versus features
If you want in-depth, real world examples and a done-for-you template that you can just follow along with, grab my 23 Pillars of Ad Copy for $5.
Performance Max has a column that shows you asset performance. Don’t replace low assets unless you’re maxed out.
If you filled out every single asset field—you’ve got all your headlines, descriptions, images, and so on—then start replacing the low-performing assets. Once you’re tapped out with everything you can do, then you can add to your asset group.
Replace the low-performing ones first, obviously. But sometimes, low performance is due to an inability to get data into Performance Max. So what you might want to do is copy and paste, then test a different audience. Or when you paste or duplicate the asset group, you can try different headlines or images, and audiences will have less of an impact.
Other times, it can be a good kick in the backside for Performance Max.
Search term extraction
This is probably one of the better things about Performance Max. Take your converting keywords that have volume, and add those to relevant Search ad groups as Exact Match keywords. Don’t waste your time on the non-converters or low-volume keywords.
How do you optimise search terms in Performance Max? If you’re seeing subcategories that aren’t related to the asset group and you’ve got enough search volume, that’s a strong signal that you should be splitting out your asset groups.
This is certainly more information than we got with Smart Shopping, for example, so this data isn’t too bad. I actually like that it’s categorised because it can highlight an optimisation route i.e. splitting out to reduce the cross-pollination of these search categories.
How do we protect our brand?
You can start by building out a branded Search campaign with Exact Match keywords, and then include your brand plus the relevant search term e.g. [brand reviews]. You can use Phrase Match, but only Exact Match will trump Performance Max in an auction, as it forces Google to use your branded Search campaign instead.
I know some people don’t like bidding on branded terms, but it gives you visibility of your brand spend. You should be gunning for 90% impression share, particularly if competitors are bidding on your brand. And in further defence of brand campaigns, new customers are less brand-reliant anyway, so there is a tendency for Performance Max to scoop those up.
Negative keywords and brand exclusions
You can apply two types of query-based exclusions to Performance Max campaigns:
- Account-level exclusions for known bad search terms and n-grams
- Campaign-level brand exclusions. Great for testing another Performance Max campaign to see how much brand spend there is. You could run this in an experiment, but the issue is how much of a budget split to give it. It’s easy enough to duplicate a campaign, exclude brand, and test.
What do you do after this? Performance will be lower without brand, so take a view on that campaign and see if performance is acceptable versus your lifetime value. You could even flip it to new customer acquisition, which is a more aggressive strategy. It requires that you be comfortable with your LTV calculations, as well as spending more at a much lower ROAS.
Performance Max is data-driven attribution, so excluding brand search will have a knock-on effect. By all means, test it. But you might have to actually unwind this. It’s why I always prefer to use a Search campaign to defend the brand terms—at least you have visibility into spend, impressions, and so on. If Performance Max starts to take away from that branded Search, you can at least see it happening in the campaign.
Google recommends the following for images:
- No text
- No logos
- No borders
- No collages
- No filters
These are recommendations, not rules. Your images won’t necessarily get disapproved if you use these elements.
Here are some recommendations from me:
- Make sure they look good on mobile. Performance Max serves a lot of that traffic.
- Give them decent contrast.
- If it’s more zoomed in than normal so people can actually see what it is, all the better.
- You can use text, but only a couple of words. Three to five is not a problem and will get approved.
Basic audience optimisations begin with custom in-market. Lead gen accounts possibly start with “remarketing” to non-converters. It’s not a true remarketing audience because Performance Max uses audiences as signals, but it will give things a kickstart. And then you can scale using the Insights tab.
Use your historic Google Analytics data, which will advise you on the type of in-market and affinity audiences, but start with in-market. Add that, and then you can layer in audiences later—it’s slightly more advanced and entirely optional. Look at level two optimisations for more.
Remember, audiences steer a campaign’s initial trajectory and Google quickly expands past them. But it is worth testing, and if you do a duplication with a new audience, it’s probably the better route.
Audience insights is where the expansion is shown. You might not have added beach-bound travellers, for example, but it will be an optimised audience there. What you can then do is try these audiences with dedicated new asset groups that speak directly to that type of customer currently converting.
And of course, you can take these audience insights and run them against things like YouTube campaigns and dedicated Search, Shopping, or remarketing.
Level Two Optimisations
Not all stores enjoy seasonality or promotion periods, so you need to think ahead. Set your budgets to capture additional headroom during times like Black Friday. Conversely, push your ROAS up to fend off budget and stock exhaustion if you’re not actually in the mood for a sale.
Start making adjustments a week before you think you need to, but change when that is. A lot of people start making adjustments for Black Friday a week before, when they should have done many weeks prior. Don’t start thinking about Christmas in the first week of December. Plan it in September, if not August, and raise budgets in October—just an example.
With new customer bidding, there’s an opportunity to collect new customers in the heightened shopping seasons—pretty much most of Q4, to be honest.
Seasonal bid adjustments
Google supposedly figures out seasonal bid adjustments automatically in the bidding engine, but you can help it along. Here’s an example of Black Friday blowing the doors off seasonal adjustments.
The previous two years have seen a 30-50% conversion rate increase. We’re telling the bidding engine to expect that during the 21-25 November and, therefore, increase the conversion rate 50%. What does this actually do to the bidding system?
If you double your conversion rate in this adjustment, it essentially halves the ROAS for the bidding engine. That’s the way it looks at it. It’s trying to keep the cadence of conversions the same so that it doesn’t underspend or overspend. It’s quite arbitrary in that respect.
But you can do these things for freak conditions. The obvious ones are sales and promotions, but you can also do stock and availability. If you have a stock or supply issue, and your conversion rate is going to drop through the floor, this is a way to somewhat counteract that.
Do use this sparingly and in very short bursts—one to two weeks at the very most. It mitigates overspending and underspending, and helps balance your ROAS out through a potentially choppy period.
You need more data, not assumptions. But you can layer things like custom segments, your own data, in-market, demographics, the whole shebang. Then you have a layered audience. Again, this is a signal. It does work quite well on bigger accounts.
This is an example for a mortgage advertiser—obviously quite a big market, so there’s a lot of data in this account and it can work with this. But I wouldn’t use this setup if you don’t have sufficient data.
If you’ve seen the templated videos that Performance Max creates for you, it’s probably scared you into wanting to create your own assets; they’re that awful.
I’m not going to walk you through how to create a video, but here are some things to think about:
- Does it grab attention in the first five seconds?
- Does it call out the target audience in the next 5-10 seconds?
- Are you qualifying who they are e.g. with a call to action in the middle or end?
- Does it demonstrate your offer and its value?
What can you test in a video?
Many people get stuck with what to put in the video, so here are loads of examples:
- Call to action
- Talking head
- Screen recordings
One thing I will say is try and split your video into the hook (that initial five seconds) and then the body and call to action. Then, record a few variations of each component—three or four hooks and bodies, and two or three calls to action. You can then turn all these sections into a multitude of videos, allowing you to test a new hook or call to action each time.
It sounds like a lot of work, but it’s far easier than trying to record 50 unique video concepts.
Performance Max does suggest trends i.e. things you may be missing out on. This is far more advanced because it’s not always applicable. Should you be advertising on these strategies? Can you even do it?
We’ve got pyjamas and sleepwear here. You might be selling shirts, and pyjamas are completely inappropriate, but it might give you impetus to expand into a certain area based on what you do now. Google will typically show related category trends, so this is more of an expansion plan or thought experiment.
Performance Max optimisation FAQ
When should I split asset groups?
- When they start to compete against one another. Look at these four asset groups that are triggering the same Search categories. They need to be split out because clearly, they’re in competition.
- When subcategories are not related.
- When you struggle to get impressions. You might even want to put them in a different campaign, alter the budget and bidding strategy e.g. Maximise Conversions.
- When reporting simply isn’t good enough for you. If you need to see asset-level spend and all the click/impression/conversion metrics, the only way to do that is run one asset group in one campaign for “dedicated reporting”.
I wouldn’t do that last one without enough data. Splitting out Performance Max into small chunks of data will hurt. With something like 100 conversions a month per asset group, you can get away with it. If your entire account does 100 conversions a month, you run the risk of crushing that or throttling performance.
When should I replace asset groups?
You asking this assumes that you’ve done all the level one optimisations before thinking about nuking your asset groups. You might have audiences that are showing mixed results, or you might want to duplicate and keep the winning audiences or add them as signals. But essentially, when none of the level one optimisations work is when you consider this.
What’s the purpose of placement reports?
Google has listened and promised placement reports, but they’re not here yet.
As they exist today, they’re a disgrace. You only see impressions (which is ridiculous), so we can’t see a placement’s spend, conversions, or anything else. You can export this data, filter it, play around with it, look at the domains in it, and then add it to the exclusions in content suitability. Or better yet, you can use the Google Ads Editor. But it doesn’t support placement exclusion lists, which I find to be a bit ridiculous.
What you will see is “Google-owned and operated”—that includes Gmail, Play (apps), Discover, and YouTube networks. That one line item in your placement report includes everything, and you only get impression data against it. It’s atrocious, but at least you can exclude some placements. Not from inside the report, of course; you have to do that manually.
If you’re losing your marbles over ad copy and competitor reporting, grab this free training, template, and report.
What are the Insights within Performance Max?
These are helpful, and can trigger intelligent bid and budget changes. It gives you a bit of a heads-up of where you need to be, and if you should be bumping your budgets, or updating and splitting assets out into new customer campaigns. If there’s an opportunity in the insights to do that, go after it.
They might also give you an opportunity to catch some expansion. And there’s a bit of a spend prediction ahead as well to give you insight into how much you should be allocating.
Next steps: Optimise beyond Google Ads
This includes things like landing page and product page content, which has considerable influence on conversions (micro and macro). If your pages aren’t converting, you need to look at why.
- Study what else is in the market; see what you are or are not saying.
- Add keywords to headlines, body copy, titles tags, and meta descriptions—no stuffing.
- Don’t ignore page speed.
- Produce that proper video if it’s appropriate and viable. A lot of accounts get away without video.
- For lead generation, look at form optimisation and lead qualification.
- For ecommerce, test bundles, multi-buys, and subscriptions. They’re often boosters for lifetime value and attract return customers.
These are paths to quick wins that increase revenue and customer retention, but that doesn’t mean they’re easy. Take them seriously.