Together, GA4 and Google Ads are invaluable tools. But how can you use them together to power your smart bidding?
This article will be split into three acts. But will it be a tragedy or a comedy? Well, hopefully, you’ll be able to determine that shortly.
This is a write-up of the talk that Michael Patten gave at GA4ward. You can find a link to Michael’s Slides here and you can also find a recording of his great talk here:
Smart bidding has been the norm in Google Ads for about five or six years now. But as you can see in the image above, not everyone is thrilled about that. It’s easy to sympathize with this opinion. Google has shared results but not techniques or philosophies behind the features they bring out.
Here’s our philosophy for using smart bidding to its maximum potential. If you treat smart bidding like a robot, it will behave like a robot. Your data will go in and out and you won’t get any special results. But instead, if you think of smart bidding like an animal, you are likely going to get a better result.
For example, if you kept a dog locked in a garage with only junk food, and then asked it to fetch you a ball, it probably wouldn’t be that responsive. By contrast, if your dog is allowed to roam free and have fun, you’ll get a very different result. You’ll have a happy dog and be a happy owner.
In this metaphor the ball represents conversions. The food and toys represent good, clean, robust data. But whilst all that sounds very simple, it’s often not applied in practice.
The smart bidding solutions of Google Ads are like a dog with a sniper rifle. When you combine these solutions with the machine learning of GA4, it’s like adding a second helper dog spotting targets with a telescope. This metaphor represents the power of combining these two Google forces.
But how do we achieve this?
Go to product links in the GA4 property menu, and Google Ads links.
Choose the account that you wish to link.
Have a look at the settings that follow, and then hit save.
And there we have it! If only that were the case. Because GA4 is new, there are a lot of good and bad guides out there. Many won’t go beyond the steps that we just looked at.
The actual strategy goes much deeper than simply sharing data between two platforms. The approach and the results generated, actually come as more of a trifecta:
To do this you need to lay some foundations. In other words, you need to empower your GA4 property. This means using some of the following features:
Important: Be sure to enable Google Signals. This is something that Google is really focusing on at the moment. Just remember before enabling, make sure you meet the policy requirements. Otherwise, you might find yourself in an awkward legal situation.
Google Ads is full of powerful machine-learning features and audiences within GA4. But these can’t be passed to Google Ads without enabling Google Signals and ads personalization.
When it comes to policy requirements, there’s a lot of text on the page. It’s better to conduct due diligence and check that your clients are happy with the terms. If in doubt, reach out for legal advice.
Once GA4 is all set up and ready, you have some work to do on Google Ads. This includes:
*If you can group campaign themes, types, and behaviors into a portfolio, you’ll get better results. On this subject, what you DON’T want is one shared budget with two portfolio strategies. Much like two dogs fighting over toys, there might be some conflict. Instead, have a shared budget for each portfolio strategy.
With Universal Analytics, there were various different options for tracking conversions. The same goes for GA4 and Google Ads. If you’re tracking the same option across multiple platforms, you need to make sure that it’s firing at the same point.
The discrepancies between the platforms are notable as it is. You don’t want to add further discrepancies to make your results a wonky mess.
For those of us that like to house everything in Google Tag Manager (GTM), this could be simple. All we need to do is change the triggers within. Any platforms that represent the same action use the same trigger logic to fire.
BUT this could go wrong. If two of the three platforms are using the same trigger, but the third platform is using something different. Remember, you will need to consider deduplication features too. For instance, for Google Ads conversion involving an e-commerce purchase.
If you add a transaction ID to a tag that fires twice by mistake, it will know it’s the same interaction. The second instance will be removed. If you don’t do this, however, you’ll see inflated data, a situation that you don’t want.
Now it’s time to get your strategy off the ground. You’ll need to begin the process of importing any learnings from GA4 into Google Ads.
The first thing that you want to look at is audiences. The image above represents the older days of Universal Analytics in conjunction with Google Ads.
Generally, you could make audiences within the platform of Google Ads with simple logic applying. We could then create a segment in Universal Analytics and send it to Google Ads for use.
But there’s no use dwelling in the past. Reassuringly, however, Google Ads remarketing audiences still have the same logic, as does GA4. But what properly moves the strategy along is the additional machine learning-powered magic.
This is the star of the show when it comes to linking GA4 with Google Ads. Primarily, it’s Ecommerce focused but there are benefits of doing a lead generation campaign as well.
But why use predictive audiences? Well, as the name suggests, for their predictive qualities. It can give you visibility over the users most likely to purchase soon. Or even the customers that are likely to spend the most out of all other users. This can be powerful when applied to a Google Ads smart bidding strategy.
Similarly, it will identify users that have previously visited but are not likely to visit again. Or, if someone has purchased before and is not likely to purchase again in the near future.
You may want to run a separate re-engagement campaign for these users. For instance, let’s imagine that somebody has bought a product in a previous Black Friday sale. You want them to return in a future sale. You might target them with very specific messaging about this year’s Black Friday offers.
Predictive audiences can also be combined with other factors such as locations and traffic sources.
Lifetime audiences (LTV) give you more valuable visibility over several key areas. This includes when users completed their first key interactions, their last known activity, or their behaviors since they became a user. This doesn’t just mean visibility for yourself, but also for your Google Ads machine learning system.
Audiences become much more accurate after you have integrated user ID with GA4. Depending on your site platform or your CRM, this could be an easy process.
Shopify has a global object for customer information. This can be called upon for logged-in users. Not only can we call the user ID, but also lifetime data variables that can be populated. This includes the number of orders made, and the total that a user has spent. This information is usually highly accurate and provides an alternative source for logging LTV data.
This information can be added to GA4, meaning you have two measures working together. Why not send these as user properties on the login event?
Did you know that you can use audiences in Google Ads to influence the overall conversion value? The tool uses information that it can already gain, such as demographics, to provide valuable insights. For example, by looking at data, Google Ads can identify your most valuable users, as well as the least valuable users.
But if there is any information that Google Ads doesn’t have, you can provide additional information. Let’s imagine you are a business running car washes. People can go on your website, enter the make and model of their car, and gain a price. They can then go on to book an appointment slot and get their car washed.
You can now analyze customer data from these users. You find out that people who own certain models of luxury cars are more likely to buy additional extras on the day of the wash.
These instances are valuable in a way that is not instantly detectable through online activity. We could create an audience using the make and model of the car used to book an appointment. Google Ads will then recognize certain users as being more valuable. Having identified this, the tool will actively try to find more of these users.
Google calls value-based bidding smart bidding 2.0 because it adds additional context. In other words, it won’t bid too high or too low depending on the anticipated value of the customer. It also allows the smart bidding solutions to learn more. This is by providing additional context to create smarter solutions in real time.
This is relatively easy for e-commerce businesses, as you only need to take the transactional value.
But it is also fairly simple for those doing lead generation. It is simply a case of creating proxy values based on internal sales data and behavioral values.
Whether you are an e-commerce or lead generation business, its value relates to revenue. But there is also another way.
We have a choice in our value-based bidding options. We have ROAS, this is revenue over ad spend, sometimes referred to as return on ad spend. Or, we have POAS, this is profit over (or on) ad spend, sometimes known as profit on ad spend.
Here’s why you might want to use POAS, rather than ROAS. Let’s envision a business in the consumer electronics industry. This particular business sells laptops as well as laptop cases. Looking at these products, which would you like to sell more of? Of course, the answer is laptops, as these generate the most revenue.
But, Let’s dig a little deeper. Looking at profit margins, we see that the business actually makes more money on laptop bags. When you consider the cost per click, the picture becomes even less rosy. If you’re spending £12 per click in order to sell laptops with a very small profit margin, you won’t have a good return on your investment.
You can see why growing a business in the above example was so hard to achieve. As you can see, simply measuring ROAS can give you a misleading picture of success. In this case, it would have been better to use POAS.
The good news is that you can get external apps to help you understand and analyze three profit margins on your website. You can also have a go at handling this yourself. Again, this will all depend on factors including
If you can deal with a surplus in profit margins yourself, you’ll want to run revenue and profit together. If you do this on all eligible e-commerce events, you’ll be able to track specific conversions. For example, people that add objects to their basket but have not returned to make a purchase.
But what about the laptop business referred to above? This approach would give them a much better picture of their advertising spend vs revenue and profit on laptops and bags.
If you can set this up, and pull this into GA4 as a custom metric, you can then plot revenue and profits together. You can also see the campaigns that are driving the most revenue vs profit. Are they the same or are they different? These factors need to be determined before you inform your smart bidding of what you should chase.
In the above image, we have created a custom metric. Interestingly, when creating custom metrics in GA4, you’ll be asked whether a unit of measurement is involved. You can also specify whether this relates to revenue or cost data. Reassuringly, if you set a custom metric based on revenue data, it won’t affect the revenue count seen in monetization reports.
Let’s say that yesterday we were doing ROAS bidding. But over the weeks, months, and years, the smart bidding solution has become much better at sniffing out high-value users. Now let’s say that we replace conversion value from revenue to profit. This leads to bad performance.
If you’re going to make such a significant switch in bidding, it’s best to start from scratch. This means that you need to take into account the re-learning phase of each. Even if you’ve run smart bidding for years, resetting to day zero will need time to build up those learnings again. You’ll start to see a negative impact on your performance.
We’re reaching our conclusion. But before we finish, there are a few things to be aware of that we’ve spoken about today. You’ll notice that we haven’t recommended the GA4 conversion vein import for linking two platforms together.
Whilst the GA4 version is a lot more accurate than Universal Analytics goal import (we’d see significant underreporting). It still doesn’t beat the dedicated Google Ads conversion tag. Remember, it’s not just you that’s doing the learning. It’s also your machine learning system. If you reduce visibility, they will not be as smart as they could be with the full picture.
Custom variables for conversion seem set to be a game-changer. You’ll be able to pull in variables tailored to your business, directly into the Google Ads conversion tag. You can do this if you use gtag for reporting purposes. It’s also set to arrive with GTM very soon.
Earlier, we covered lifetime value audiences. Quite quietly, Google Ads has added a lifetime value dedicated field, along with new customer bidding options. If you have a source of lifetime value outside of GA4, you can provide this directly to Google Ads. This can bolster the information being passed from GA4.
Are you using custom dimensions and metrics in an exploration report? Things can go quite wonky if you don’t pay attention. If you’re to make solid decisions based on data such as this, make sure that you’re seeing an accurate picture.
Even though it’s GA4 if you’re importing audiences from GA4 to Google Ads. They’ll still be subject to the same minimum audience sizes as Google Ads. It’s all very well using GA4 for niche targeting. But if there are only 9 people every 30 days in the audience, it’s too niche. Keep the numbers above in mind, depending on the activity that you’re running.
If you want to explicitly target an audience, then you’ll have to use the targeting option. If you use observation audiences, even if you set a big modifier, you won’t influence bidding. At most, it will be used for attribution purposes. At worst, it will be used for absolutely nothing.
Of course, it is handy to add these audiences as observations. This gives visibility to your smart bidding solution. Have them as observations regardless. But if you want to target, you will need to create a new ad group.
Have you got a hot brain from reading all this? Let’s remind ourselves of some of the key takeaways.
So, get creative with your smart bidding, but always carry out testing (and test responsibly)!
Want more? Take a look at our blog for more about Google Analytics, and other Google packages.
After working 6 years as the digital marketing strategist at Redweb, Michael now works as the data strategist at Launch Online, a paid media agency based in the UK.
Michael is able to utilise the power of data to advance paid media strategies and inform high-level business decision-making.
Find him on LinkedIn.