Updated Demand Gen Strategy in Google Ads - Life Hacks in 2025

Test campaigns for testing creatives

Test campaigns. They are also called passive, right? Or scaling. Although here, we are probably talking about the initial test.

Yes, exactly the initial one. They have one goal: to be as fast and, importantly, as cheap as possible. To test as many creatives as possible.

Creatives, not audiences. Exactly. The focus is on creatives.

That's why the environment for them is so specific. How so? If we are testing creatives, the audience must be ideal, a benchmark. Absolutely right.

We use the most accurate and, most importantly, proven audience segment. The one we are 100% sure will work for our product. For example?

Audiences based on competitors' keywords or brand queries. That is, people who are already looking for something like that. Or data from Search Observation Audiences.

This is when we look at how existing segments, based on interests or intentions, behave in search. The idea is simple. Stabilize the audience variable to assess the pure impact of the creative.

It's logical. We isolate the variables. And how does that work in terms of structure? Do you need a lot of these campaigns? Usually, these are small, very focused campaigns.

This is often how it's done. One campaign, one ad group. And there are, well, about five similar creatives.

Let's say five video options or five banners. I see. And so many small campaigns are launched.

A lot. A lot. The numbers are shocking, to be honest.

How many? 10-20 such test campaigns. Per day. How many? 10-20 per day? That's a huge amount.

Is this really feasible for, say, a medium-sized business? Yes. The volume is truly enormous.

It's more for large players with significant resources for both creativity and management. But the principle itself is important for smaller scales as well. What principle? Get ready for a constant stream of tests.

Maybe not in such large quantities, but the point is to quickly try out different options. You need to try a lot of things. I see.

What about the budgets for these tests? Are they also huge? No, not at all. The budgets are actually quite small. And they are strictly tied to the target CPA — the cost per acquisition.

Right. The recommendation from the materials is a daily budget for one such test campaign of approximately 3-4 of your target CPAs. If your CPA is low, say a couple of dollars, then you can set 7-10 CPAs per day to collect data faster.

How long does such a test take? Not long. Usually about a week. It's like a sprint.

And what should we see in a week? What is the goal? To find the winners. That is, creatives that brought, say, 10 conversions during this week — this is a benchmark and at an acceptable price. And if not? And if not, then we ruthlessly delete them.

This is where ruthlessness is important. If the creative doesn't work, produces expensive conversions, or simply doesn't deliver on the budget, it's out. No need to wait for a miracle.

Quick elimination. I see. What bidding strategy is best for such quick selection? Often, they start with maximum conversions.

This allows you to get impressions and initial data faster. But. If conversions are too expensive right away, then you need to immediately switch to a target price per conversion, to a target CPA, in order to control costs.

The main thing is speed and clear screening criteria. Okay. We seem to have figured out the test pipeline.

We found a few creative stars. What do we do with them next? Let's move on to the second type — scaling campaigns. This is where, you said, the most interesting part begins.

Scaling with selected creatives rather than targeting.

This is where it gets really interesting. The goal of these campaigns is to take these winners and give them maximum exposure, to really promote them.

And this is where the main break from the norm occurs. Especially in terms of the audience. Right, right, right.

What kind of audience is needed for scale if the tests were narrow and proven? Here, it's as broad as possible. This is the key point that many find counterintuitive. As broad as possible.

That is... Minimal targeting. Just the bare minimum. We only leave geo.

Well, maybe the most basic demographic restrictions. Gender, age, and that's only if it's critical for the product. That's it.

Wait. No targeting at all? No interests, no looks, no likes? Nothing? Just the whole country or region? But that would be a waste of budget. On an irrelevant audience.

We were taught something completely different. Exactly.

This is one of the most surprising conclusions from our experience that we are sharing. The whole bet is on creativity itself. On creativity.

Yes. The idea is that strong, tested creativity is capable of finding its audience in this broad mass. Thanks to Google's algorithms, of course.

So the algorithm will figure it out itself. Yes. It learns from conversions and optimizes impressions itself.

It shows them to those who are more likely to respond. And narrow targeting at this stage, as practice shows, can only get in the way. It can prevent the algorithm from finding unexpected but effective segments.

We trust the creative and the algorithm. Sounds bold. But how do we do that? We simply copy the winning creative into a new campaign with broad targeting.

Yes, that's right. A new campaign is created. One or several winning creatives are added to it.

But it's better not to mix too many different ones in one group. And this campaign is given a budget. A significantly larger one.

How much larger? Significantly. And here's another point that goes against what was previously advised. What is it? Probably about the speed of budget increase.

I remember they used to say no more than 15-20 percent every few days, or even every week. Exactly. The old school.

Increase slowly, don't scare the algorithm away. But here the approach is much more aggressive. You can increase the budget by 20-30 percent.

Wow. Every two days. If the campaign shows good CPA results, of course.

Every two days? Yes. They even give an example of manual management, when the budget for a successful campaign was raised from $3-4 thousand per day to $10 thousand by the weekend. Well, when they expected peak demand.

That's very, very dynamic management. Wow. 20-30 percent every two days.

That's really aggressive. And what is the bidding strategy on such a scale? Also maximum conversion? Often yes. They use maximum conversion.

Especially at the start of such a scaling campaign. To give it a boost. But how do you control CPA? CPA is controlled manually.

Through the budget. If they see that the CPA has crept up, they cut the budget. If the CPA is stable or falling, they aggressively increase it.

I see. But then, when the campaign has gathered enough data and is running smoothly, switching to target CPA is also a viable option. For greater predictability.

Let's dig a little deeper into tactics. What else is important to consider when setting up these two types of campaigns? Especially when it comes to creative formats. Yes.

There are several critically important points. Really critical. First.

And again, this is a counterintuitive insight that often contradicts what Google itself has previously recommended. Intriguing. What is it? It's the separation of formats.

Video and images? But Google advised mixing them in one group for better performance of responsive formats. Isn't that right? That's exactly what they advised. But practical experience shows that

the results are much better and clearer if you strictly separate them. That is, separate campaigns for videos and separate ones for images. Or, at a minimum, separate ad groups.

A separate video, a separate image. When they are mixed, it is very difficult to understand what exactly worked. All the statistics become blurred.

It is extremely difficult to optimize this. Separation gives a clear picture for each format. This is a really valuable observation from practice.

What about placements? In DemandGen, you can now choose where to show InStream, Shorts, and InFit. Yes, you can.

And here, too, there is an observation from practice. For video advertising, the InFit placement, well, it's in YouTube, Discover, and Gmail feeds, often shows the worst performance. Worse than InStream and Shorts.

Yes, compared to InStream and Shorts. The recommendation is this. Start testing with all placements enabled, but keep a close eye on the stats.

And in subsequent iterations, disable the ineffective ones. For example, uncheck InFit if it consistently fails to convert or is too expensive. Returning to the volume of tests, you mentioned up to a hundred creatives per day for large budgets.

This emphasizes how important the volume of testing is. Absolutely. To find those diamonds in the rough, you need to test a lot.

On average, only about 5% of tested creatives become real winners. Only 5%? Yes. The ones that can be successfully scaled.

This means that 95% of ideas and variations are discarded. And you need to clearly understand this when planning your resources. Both for creative production and campaign management.

It's a game of big numbers. Are there any specific tips on image formats in these materials? Yes. They recommend testing different aspect ratios.

Horizontal, landscape, square, one-to-one, and vertical, portrait, 4 to 5. And always add a logo. Which works better, adaptive or drawn? Here's an interesting trend. Creatives that a designer has drawn specifically for each size often work better than when the system itself crops one image for different formats.

So it's better to go to the trouble? It seems so. It's also important to test different concepts or ideas in separate ad groups. And present each idea in all formats.

Horizontal, square, vertical. What about video formats? Here, the main focus is on vertical 9-16 for shorts and mobile feeds and horizontal 16-9 for instream. Square videos currently generate significantly less traffic, so you can spend less effort on them.

There are nuances. There is one critically important nuance for horizontal videos — custom covers. Thumbnails — previews.

So, don't rely on a random frame that Google will choose itself? Absolutely not. You need to create and upload your own. Attractive ones.

After all, users see the cover before deciding whether to watch the video or not. Especially in the feed. It makes sense.

This has a huge impact on CTR and further interaction. Testing different covers for the same video is also part of the job. What is the best way to group videos in tests, with five different ones in one group? There is no single recipe here.

You have to try different things. We offer different options. For example, a group with five different vertical videos.

Or a group with five different horizontal videos. Or a group with one vertical and one horizontal video, but both about the same idea. And what about this Google metric, AdStrength? Is it important? That's an interesting point.

It's worth noting that it's becoming less important than actual CPA and conversion metrics, especially if you separate formats, as we discussed. So it's probably not worth chasing Excellent AdStrength at the expense of results. And one more clarification on targeting.

Returning to test campaigns. Although the audience should be precise, it should not be restricted by a multitude of conditions. What do you mean? It is recommended to use a maximum of two or three targeting signals simultaneously.

For example, demographics plus interests plus location. That's it. There's no need to add language, parental status, device type, and so on.

This can narrow the reach too much, even in a test, and prevent you from quickly collecting data. What audiences are considered a good starting point for tests? There are usually three main types.

The first is affinity audiences. These are people with stable interests, such as cooking enthusiasts. The second is in-market buyers.

These are people who are actively searching for or comparing products and services. Buying a car, for example. Their effectiveness, by the way, can be estimated in advance through search observation.

I see. And the third? The third is special audiences based on intent, custom intent. They are created based on keywords that people search for on Google or the URLs of websites they visit.

This allows you to target people with specific recent intentions very accurately. What about lookalike audiences? They are also mentioned, but they are given lower priority for the very first tests. These three — affinity, in-market, custom intent — are considered more reliable for the start.

Okay. The tactics seem clear. But any creative, even the coolest one, burns out over time.

Let's talk about that. How long do they actually last in Dimension? Burnout is absolutely normal. It's to be expected.

Here are some rough estimates. Video creatives last about a month of active use. Static images last a little longer — 2-3 months.

But these are, of course, very average figures. You have to look at each specific case. And how do you know when the moment X has come? That the creative is burning out? The most important thing is the CPA dynamics.

If you see that the price per conversion is rising steadily, say, for 3 days in a row or longer. 3 days. Yes.

Three plus. It's important not to panic over 1-2 day spikes. These may just be market or auction fluctuations.

What you need is a steady CPI growth trend. And the second signal? The second signal is when the costs or impressions for a particular creative start to fall steadily while the campaign budget remains the same. This means that the algorithm has stopped actively displaying it; it is losing effectiveness in its eyes.

What should you do if you see these signs? Just pause it? Here's some important advice from experience. Don't pause a burnt-out creative with the idea of “let it rest, then turn it back on.” It won't work.

Most likely, no. It will not return to its previous effectiveness. The system has, so to speak, remembered it as less effective.

In other words, resuscitation is useless. What if you refresh it a little, change the music, background, text, just a little? Yes. Refreshing is a viable option.

Take a successful idea and make a variation on it. A different background, different voiceover, length of the video, call to action on the banner, but... Key point. What is it? These updated versions must be run through test campaigns again.

Tests again? Absolutely. You can't just take the old creative and replace it with the refreshed one in a scaling campaign. You first need to test it to make sure it really works.

And if it wins the test again... Then what? Then you need to create a new scaling campaign for it. Don't try to squeeze it into the old one. Got it.

It all comes back to the idea of continuous testing. How else can you optimize campaigns? Here's the “Budget Limited” status. What should you do with it? It's a direct call to action.

If your campaign is budget-limited and you're completely satisfied with the CPA, then you're losing conversions. Simply because you don't have enough money. The solution? Increase the budget.

Remember those aggressive steps — 20-30% every two days. And if the campaign is limited by budget, but the CPA is too high, then either your target CPA is too high or your budget expectations are too high. In that case, either lower the target CPA or, well, accept the current conversion volume.

It is also important to check whether the “Optimized Targeting” option is enabled. What is this option? Can you briefly remind us? It allows Google to show ads to people outside your explicitly selected segments of interests, intentions, and so on. That is, if the system believes that these people are highly likely to convert.

Ah, so it's about trusting the algorithm? Exactly. This fits perfectly with the philosophy of broad settings. However, if you enable this option on an already running campaign, there may be a slight dip at the beginning while the system adapts.

But in the long run, it usually pays off. Let's summarize the bidding strategies. What are the main ones for such dynamic demand generation? There are two main workhorses.

These are maximum conversions and target CPA. Maximum conversions for what? It's great for quick tests. And for an aggressive start to scaling campaigns.

When we control CPA manually through the budget. And target CPA? It's good for more stable scaling. When you already have data, but the benchmark is around 50 conversions per week.

And when you need greater predictability in terms of cost. By the way, it is recommended to take the starting target CPA as the average actual CPA for the last 30 days of work at maximum conversions. What about value-based strategies? ROAS.

However, they are considered less suitable for this approach in demand gen. With its frequent changes, focus on conversion volume, and not always on their direct value. ROAS strategies are better for more stable environments.

Well, search with clearly trackable e-commerce. It's good to try to tie it all together. Separate campaigns.

Broad audiences. Volume of tests. Aggressive budgets.

Focus on creativity. What is the overall philosophy here? What has fundamentally changed? The overall philosophy is a transition to a model of, let's say, continuous high-speed testing and learning. Test and learn, as they say.

Constant search, rapid iteration. The emphasis is shifting from carefully selecting an audience to creating many creative hypotheses. Hypotheses in the form of creatives.

Yes. And quickly test them, and then aggressively scale the successful creatives to the widest possible audience. And we kind of let the creative and the algorithm find the response.

It sounds like a “volume game.” Exactly. It is a volume game that requires serious resources.

Time, expertise in creating different creatives, analytics for quick evaluation, and, of course, budget. But there is a plus. What is it? The system learns not only within a single campaign.

It learns at the level of the entire DemandGen account. That is, the more you test, the smarter the algorithm becomes within your account. And over time, the search for effective combinations can accelerate.

So, to summarize for our listeners who work with DemandGen. The main paradigm shift we discussed today. Forget about tests within working campaigns.

Create separate ones for testing and for scaling. For scaling, use extremely broad audiences. Rely on the power of creativity.

Creativity itself is king. You need a lot of variety. The volume and speed of testing are decisive.

And, importantly for the purity of the experiment, we separate videos and images. Exactly. You have to admit, this is quite different from many familiar practices.

Yes. It really requires a rethinking of many well-established approaches. And that brings us to our final thought.

Key point and question

If success in DemandGen now depends so heavily on a large volume of creative content and the speed at which it is tested, how should internal processes adapt? In marketing and design teams? How can this DemandGen machine be effectively fed with new ideas, variations, and formats at such speed? What becomes the main bottleneck? Idea generation? The speed of video and banner production? Or is it analytics and quick decision-making based on the results of hundreds of tests? So, the question is no longer so much about Google Ads settings? Yes. It's now a question of how work is organized within the campaign. How ready is it for this pace? An excellent question to ponder after our conversation.

That concludes our dive into the nuances of DemandGen for today. We hope these practical insights were useful, gave you food for thought, and inspired you to try new experiments! If you need help, please don't hesitate to contact us.