Find out in this new study how to get the most out of Google Ads Smart Display campaigns. What are the pitfalls to avoid? Which strategies work? What a performance compared to classic Display campaigns and more. All illustrated with many examples, as always.
This study is based on the analysis of nearly 300 million impressions representing more than € 280k in advertising spending with 13 customers in various and varied fields of activity.
Attach this study to previous studies
What is a Smart Display campaign?
According to Google
You can find the full definition here.
They allow you to reach Internet users at all stages of the purchasing cycle.
They are based on 3 pillars to overcome the difficulties and approximations related to targeting, auctions, as well as the creation of ads:
- Automatic auctions;
- Automatic targeting;
- Automatic announcement creation.
Last important point, according to Google, these campaigns generate an average of 20% more conversions compared to classic Display campaigns.
These are full automatic campaigns where the only thing we give are images, messages and a target (cost per conversion or target return on advertising investment).
Everything else is a black box. ?
Performance level, when they work well, they are war machines.
But this is not always the case (we will come back to this in detail) and your settings are crucial for success or failure (we will come back to this too).
How does it work in practice?
Here I summarize the promises made by Google:
- Smart Display campaigns analyze the targeting and campaigns already present in the account and then integrate into “fill in the blanks” mode;
- In addition, they bring new targeting (mainly custom intention audiences created automatically and specifically for you);
- As a reminder, custom intention audiences are audiences present on the market and personalized and therefore rather in the middle / at the bottom of the tunnel (ie people close to the act of purchase).
They are therefore complementary to campaigns targeting higher up the tunnel (i.e. people not yet ready to buy) such as affinity targeting.
If, however, other campaigns (video or display) are active, the Smart Display campaign will go “Sting” them with the people who have the strongest intention.
Finally, note that SDC campaigns are both prospective BUT ALSO remarketing. It’s all-in-one.
Good setup practices
Bet and invest enough
This means 50% more than the cost per conversion of all the search terms in your account that do not contain your brand / company name.
Budget level, Google recommends putting a daily budget 5 to 10 times higher than your CPA. ?
Wait at least 3 to 4 weeks
This is the time it takes for the algorithm to learn and be at “cruising speed”.
Google therefore recommends 2 to 3 weeks before making minimal changes and 4 weeks before analyzing the results.
Use a multi-key attribution model
Ideally use the data-driven attribution model.
But it requires 15,000 clicks in Google search and at least 600 conversions for the conversion action within 30 days.
Suffice to say that it is not for everyone. ?
Otherwise, it is recommended to use either the position-based model or the time depreciation model.
Set up display impression tracking
Why ? Only 16% of people who see a banner will click on it.
Consequently, we regularly see that conversions after display (i.e. without clicking on the banner) represent as much if not more in volume than conversions (normal) after clicks.
The implementation involves activating Display impressions in Google Analytics.
You cannot do it by yourself, you will need to ask your Google contact.
Finally, note that for it to work (and it makes sense), the Analytics and Ads accounts must be linked.
Have enough historical conversions in the account
If not, you will not have access to the functionality anyway:
At least according to Google 50 Display conversions or 100 Search conversions in the last 30 days.
Note that these minimums tend to drop steadily, as Google wants to maximize the adoption of these “smart” campaigns.
Anyway, when you have enough historical conversions, the option appears:
My opinion and my experience with these recommendations
My experience shows that in principle, Google’s recommendations are fair enough.
I have tested the CPA and budgets a lot and indeed, if you don’t put enough, it rarely works. The campaign ends early on its own (examples to come later).
And indeed, the campaign needs quite a bit of time to perform stably.
She already has 7 days of learning.
It is then necessary to count 1 to 2 weeks more.
And from experience, it’s independent of the number of clicks, it’s really a time story.
Some examples :
Benefits of SDC campaigns
I see 2 major advantages: simplicity and time saving, both in terms of configuration and optimization.
Indeed, the configuration is very simple.
We give the pictures, the video (optional), the messages, the budget, the goal and that’s it!
Since targeting is 100% automatic, there is no need to search for placements, themes, audiences, etc.
It also means not having to create multiple ad groups with different targeting.
Generally, we make a group of announcements and the deal is done.
In addition, since Smart Display campaigns are prospecting + remarketing, well it also makes one less campaign to create. ?
Finally, in terms of optimization, the only things we can do are:
- Optimize images and messages replacing the poor performance elements with new challengers;
- Determine and exclude bad locations (possible only at account level, not campaign level);
- Exclude certain content categories. But it’s not really about optimization, it is set when the campaign is created;
- And then of course (even if it’s not in Google Ads), test and optimize landing pages some announcements.
Disadvantages / Constraints of SDC campaigns
The advantages are also disadvantages, it depends on the point of view where we stand. ?
For now, whatever we say, these Smart Display campaigns are not intended for small advertisers.
On the one hand because of the minimum conversions (but it will go down again and again) and the minimum amounts to invest (there too it will go down, I am sure).
On the other hand because it is recommended to let the algorithm learn before deciding anything.
And it is expensive (a daily budget of 5 to 10 times your CPA, for 3 to 4 weeks).
Note however that to counter this drawback (obvious !!), Google recently offers payment on conversion.
This means that if the campaign does not perform, it does not cost. We will come back to this in detail a little later. ?
Smart bidding mandatory
In principle, I hate not having a choice.
Smart bidding is good. But not always.
He has undeniable strengths:
- Save time compared to manual management;
- Superior performance in certain specific contexts.
But also weaknesses:
- Often an unfortunate tendency to raise CPCs (no big deal as long as the volume of conversions rises to equivalent CPA or the volume remains stable and the CPA decreases);
- Sometimes an annoying tendency to perform incredibly for 15-20 days and then perform less (always monitor, smart bidding is not equal to not taking care of it anymore);
- A loss of control and finesse.
Hence my preference to have the choice, just to use one or the other as the case may be.
Prospective display + remarketing = black box
A big time saver but also a big loss of control.
Unfortunately, it seems that one always comes with the other (a bit like security and freedom …). ?
Indeed, we cannot know the proportion of prospective vs remarketing neither the performance of either, nor how to optimize one or the other.
Specific case of Smart Display campaigns with product flow
You can go into dynamic prospecting and remarketing as soon as you add a product feed to your Smart Display campaigns during configuration.
To do this, obviously, you must have a Merchant Center account associated with your Google Ads account. And a valid product feed in Merchant Center.
So, through analysis of user behavior, flow performance history and machine learning, Google will distribute the most relevant products to potential new buyers.
It’s ultimately the Google equivalent of Facebook Dynamic Product Ads campaigns.
Smart Display campaigns automatically remarket.
With a feed, they will do dynamic remarketing automatically.
As a reminder, dynamic remarketing allows retargeting past visitors to your site showing them the exact products they saw or added to the cart without buying.
Please note, for this to work, you must have a Google Ads remarketing tag on your site which has custom parameters.
I explain how to do it on Prestashop and how to do it on WordPress in my Google Tag Manager guides.
For the general procedure, it can be found here.
I’ve used this type of campaign many times and I must admit that the performances are there.
These are usually always profitable campaigns and often above the goal we give them.
Example with an SDC + feed. ROAS 150% objective (we are twice as real):
So this is a type of campaign to test first. It can quickly become a Hulk. ?
Performance of different auction modes
Target CPA (pay per click)
The most classic method.
She gives the best results most of the time.
Performance is generally less volatile than the ROAS strategy.
Be careful however to give a realistic CPA target, otherwise the campaign will stop on its own relatively quickly.
I have examples like this in spades.
Sometimes it’s worth putting a higher CPA than desired at the start, so that she has enough time to learn.
Then, we gradually decrease the CPA (in steps of 10 to 15%) to reach the objective.
I’ve already had some rare successes like this. An example :
For the record, the campaign does not perform for a month.
Before cutting it (because not on target), I try to increase the CPA for 2 days.
And there it is a box and the campaign becomes a bestseller. ?
Target CPA (pay per conversion)
This method is not may not be available in your account because it is very framed:
- You must have generated more than 100 conversions in the last 30 days in your account;
- 90% of your conversions must be saved within seven days of clicking on your ad. (If your customers often wait more than a week to convert after clicking on your ad, you cannot use this payment option);
- This payment method does not work with shared budgets;
- It is highly recommended to have a target CPA below 200USD (or the equivalent in your currency). Note however that this is not a crippling criterion.
Google says check and adjust eligibility daily.
I had the opportunity to test this strategy twice and I was pleasantly surprised at the longevity with which Google still gives clicks even though they give little or no conversions.
I have used it very little so far for 2 reasons:
- Often, this billing method was not available to start (but it seems to be the case now):
- In general, we can do better by paying for clicks if the campaign works well.
Therefore, I used it as a last resort, when the campaign didn’t really work.
But as the results are quite good and now this payment method is widely available (including to start with), I think I try it much more often, right from the start.
ROAS Target (pay per click)
Obviously, one of the best choices when selling online.
But not only.
You can also give values to your conversions (however not monetary) in order touse the ROAS strategy even in lead generation (as I explain here).
The results I have seen in ROAS Smart Display are mixed. ?
- A third of the time it works super well;
- A third of the time it works but we do not reach the ROAS objective (and the campaign does not stop there);
- A third of the time it doesn’t work, basta.
Compared to classic Display campaigns where for me ROAS works almost every time, I was a little disappointed.
Target CPA with inclusion of post-view conversions (beta) (payment for visible impressions)
Currently in beta, you must request whitelistage to be able to take advantage of it.
Concretely, this option means that all conversions after display will be included in the conversions column.
By default, they are not.
Now, as said before, post-display conversions are sometimes the majority.
Some examples :
As you can see, depending on the campaign, you could miss 80% of conversions. ?
Now, in a machine learning world, the more signals you have, the faster the algos learn, and the better they will perform.
In my opinion, this option is therefore essential because it significantly increases the likelihood that the Smart Display campaign will work.
I have succeeded on several occasions in running campaigns thanks to this.
Without post-display conversions, the CPA goal wasn’t high enough and the campaign would stop.
But by including post-display conversions, you get there and the campaign becomes a hit. ?
Analysis and optimization
As said before, the possible optimizations are limited.
I still give you a few tips.
Optimize your creative assets
First, at the ad level, fill in everything that is possible :
- 5 titles;
- 1 long title;
- 5 descriptions;
- 15 images;
- 2 logos.
If you have done your job well, you should see this:
Then let your campaigns run a bit to accumulate data.
Then after a few days, you will be able to access the report on the elements:
You will be able to see the elements that perform and those that do not:
Replace items that perform poorly with new challengers.
Consider conversions after display AND cross-device
As said before, if you can, activate the beta to include post-display conversions in the conversion column.
If it is not possible, always consider these 2 metrics in addition to “normal” conversions to decide to adjust your CPA.
You just add conversions + conversions after display + cross-device conversions and divide this sum by the expense to get the “real” CPA.
Make a rate of change between this “real” CPA and that of “normal” conversions (conversions after clicks).
You can increase your target CPA by that percentage.
This will give you a much better chance that your campaign will perform.
She will try harder and longer to find the ideal people.
One last thought that I would like to share with you.
I could see that the Smart Display campaigns generated more post-view and multi-device conversions than traditional Display campaigns.
Hence the importance (again) of taking them into account in your analyzes.
Learn from my tests / errors
Since my job is to test and learn, I have tried many, many things.
I’d like to share with you the tests that didn’t work.
The best of the masters is the young Skywalker chess.
Do not start in ROAS
This has been a good lesson. ?
I created several campaigns and wanted to start them directly in ROAS since others that already existed in the account were performing.
No clicks, no impressions during the learning phase.
“Normal” tells me someone from Google’s support (however Premier Partner).
In the end, once the learning period ended (7 days?) And still having no impression, I switched to CPA.
Startup was instant…
Don’t split mobile / computers
A good lesson there too.
Everything starts from an observation.
There is not much you can optimize on these Smart Display campaigns.
But it looks like we can split them by device type.
After segmented performance analysis, I realize that my SDC campaigns with ROAS objective do better on computer and less well on mobile, but that the average of the two gives the objective.
As a reminder, we cannot make bid adjustments on these campaigns.
So I decide to split.
The logic is relentless: by bidding better for each device, we will boost the results. For sure !!
Yes, except that no.
I have never managed to get so much performance that the unsplit campaign.
And I tested on about 10 campaigns. Same result every time.
I had to back off.
Remove micro conversions as soon as you have enough macro conversions
I have been advocating for years using micro-conversions (ex: seen at least 5 pages, seen the entire page, add to basket) to be able to enjoy smart bidding.
This allows you to quickly give artificial intelligence leads, to learn up to 5 times faster:
Yes but here it is, true and effective until it is no longer.
In other words, you should remove micro conversions as soon as you have enough macro conversions (prospects, orders). Otherwise, the micro-conversions will start to scramble the best direction (that of macro-conversions) because they are much more numerous.
Suppressing micro-conversions generally results in a noticeable increase in campaign performance.
In summary : useful and essential at launch but to be deleted as soon as possible.
See in the example below how the transaction rate increases when I remove micro conversions:
Strategic reflections on Smart Display campaigns
Smart Display campaigns keep their promises.
They are simple to set up and the time saving is relatively strong compared to the previous generation.
In terms of performance, as said above, when they work, they are war machines.
But they don’t work every time.
And the configuration has the right to life or death. Do not miss it. ?
But opacity may be hiding the truth from us
However, it should not be forgotten thatthey are forward looking AND remarketing.
However, we know that remarketing campaigns work very well.
And since we have no way of knowing the proportion of prospective vs remarketing, it is also possible that the campaigns that perform best are also the ones who do the most remarketing in proportion.
We don’t know (and that’s the whole problem).
Time saving vs performance gain
And in most cases, I succeed where the Smart Display campaign failed.
And when I compete, I do better most of the time.
Yes but here it is.
Does the best performance I get cover the time I spend there (and bill my customers, of course).
It’s not always the case.
Optimizing classic Display campaigns, it takes time.
It’s a lot of research, refining, splitting, etc.
So to be completely honest, this is a calculation that is not easy to do but must be done.
It should not be forgotten that the real ROAS of a campaign must also include the cost of the provider.
But there is one last point.
A point that leans in our favor for the moment.
Smart Display campaigns take away our ability to learn about customers
You see, I was giving a Google Ads workshop in an agency, just last week.
And the message I wanted to get across is:
If you’re just optimizing campaigns like monkeys, you will no longer have a job in 2 years, when artificial intelligence will do it for you.
: I like monkeys? I use the formula simply to make an impression.
And even today, if you are just that, you are replaceable.
Google Ads agencies / freelancers are everywhere.
You must be obsessed to learn about customers.
You must have a double objective:
1. Satisfy your customers’ need for short-term results;
2. Learn as much as possible about their customers in order tohelping your customers from a strategic point of view :
- How to make an offer that sticks even better to customer expectations? (Impossible without knowing them);
- How to improve / reformulate the messages on the website, and more broadly on all communication media so that they reason even more with customers? (Impossible without knowing them);
- What questions do potential buyers ask themselves throughout their purchasing process? (Impossible without knowing them) => the answer gives the what of the necessary content strategy.
You see, when for each customer, you do that, you are no longer the replaceable Google Ads provider.
You are an indispensable partner, that we are happy to pay dearly.
That makes all the difference. ?
What I like about classic Display campaigns is that you learn the best audiences, the best locations, and more.
We have a lot more ways of learning what works and what doesn’t.
These Smart Display campaigns leave us with very few lessons in the end, I think.
Now your turn
- Do you see the same things I do in your campaigns?
- Do you have any insights to add?
- On this last part of strategic thinking, do you agree? Disagree ? Why ?