Hello, everyone. My name is Yana and I’m a Google logistician, I deliver targeted ads to businesses with the necessary parameters.
In today’s video, I’d like to talk about AI Max once again, which is our search engine with AI functionality. I posted a video on my YouTube channel announcing what it includes, why AI Max is present, and what campaign it is derived from. Today I will show the first results in AI traffic. How to find it, where to see that it is traffic that came from AI. Let’s look at several variants of advertising campaigns to see what? First, the quality of traffic generated by this AI Max. Second, whether we have increased conversions from what we launched. Of course, we will see this only on one or two campaigns, because they were the first launches. The third one, maybe, we will look at the general concept that we have.
How many calls and sales will I get by ordering contextual advertising from you?
I need to calculate the conversion of my website Describe
the task
in the application
Calculate potential ad revenue Google
contextual advertising calculator
What indicators will we analyze?
Remember that AI Max is still a search with artificial intelligence functionality. As we know, artificial intelligence is very smart and cool, but it still needs to be trained to generate the right result. That’s why we’re going to go through a little bit now, i.e. these are the first launches. We’re not going to be super picky or critical about it, but we’ll start.
I think we’ll start with a variant of an advertising campaign, perhaps this one. It was the first to be launched, and the AI functionality was enabled on June 4. We will compare it by several parameters. The first is performance in conversions. Right?
There are five and two conversions. You can see that the success is unequivocal. There are a few fewer impressions, but this is most likely due to budget constraints, because the bidding strategy was changed.
Our conversion rate has increased, which is quite good. Our cost per click shows that it has increased, but we will compare the before and after cost per click in terms of what we generated there.
Keep in mind that the bidding strategy has also been changed. What do we do?
Example 1: Campaign with an irrigator
We have a product here, it’s called an irrigator. Let’s use it as an example to see what we have. We have a type of match. We have this AI Max. Its average cost per click is 19.29 UAH. It generated two conversions in total. Not very many.
I spent about UAH 1000 on all this, and generated about 52 clicks. But we realize that this is for the queries that I can see. We also have a block of other queries that we cannot see. AI is likely to be present there as well.
What can you notice? You can see that it doesn’t match some keywords. That is, it will be incomprehensible to us. Most likely, as I said, it will be an analogue of the DPO. The type of semantics that the system picks up is as wide as possible so far. For irrigators, I see that it also shows up for some grammatical errors, like “aregator”, this is “origator”, something like that.
I mean, it’s okay, let it be, there are “top budget irrigators” and so on. Something could have been substituted, right? But before that, classic search didn’t do this, these conversions.
Now let’s compare other types of relevance besides AI before and after. We have almost no change in conversion rate between before and after.
The volume of traffic has increased, but this may be primarily due to the fact that we changed our bidding strategy. The cost per click has been growing everywhere, so AI is not necessarily more expensive, it has more to do with the betting strategy.
But if you look at it, in my personal opinion, less of this total traffic was generated by keywords. There are more phrases with “buy” or some kind of brand component and so on. So far, it’s not clear from this ad campaign whether it’s good or not.
Example 2: Campaign with trimmers
How many calls and sales will I get by ordering contextual advertising from you?
I need to calculate the conversion of my website Describe
the task
in the application
Calculate potential ad revenue Google
contextual advertising calculator
As an example, I show a campaign in which the AI machine was launched on June 12. Today I’m shooting on the 13th, so it’s hard to say anything. I am interested in the quality of the traffic. Here we have female trimmers, which is a very key factor. I just want to look at the quality of traffic.
“Beauty device” – I have no complaints. “For cleaning heels” – I don’t understand why a heel trimmer is for cleaning heels, okay. “Heel trimmer” – okay. “Razors, epilators” – okay. “Depilatory” and “foot files” are out. “Heel files,” which we don’t have. “Trimmers” – I don’t like these keywords for one simple reason, because we don’t know what kind of trimmer it is. Because if you’ve ever sold a power tool, you know that the phrase “trimmer” can show a lot of things, including “mowers” and “grass trimmers.” And here, among all these brand-model queries, there is indeed something similar to “for grass”.
There is a transition between “for scrotum shaving”, “intimate areas”, and “electric scythe” – this is an example of how well or poorly the system can understand how good or bad the structure of your site is. It chooses the right or wrong words. I can’t make any complaints here, because there was only one show, one impression of an electric scythe, but, for a moment, where is the electric scythe and where is the women’s trimmer? There are also “facial massagers,” which means that the intersection is very different. “Rechargeable scythe” also slipped through the cracks here, with three shows.
The total number of impressions was 156 yesterday. We had only three clicks. We can try to see what the clicks were for. These are “hair clipper” and “heel clipper”.
Here you go. These are the ones we have here. “Female trimmer”, “female”. Now pay attention to the type of keywords we have selected. “Women’s”, “women’s”, “women’s trimmer”, “women’s”, “women’s”, “women’s”.
Clearly, the electric scythe couldn’t keep up with these queries. I won’t show you the site.
Example 3: Other campaigns
We still have campaigns that have been receiving traffic since June 5. Unfortunately, I don’t see any conversions yet. There were 55 clicks and almost 1000 impressions. We’ll have to look at CPCs a little later.
We can see how the average CPC, or cost per click, has changed. But this is not yet an indicator, because there may have been some strategy changes.
Is it inadequate? No, I wouldn’t say that it’s super. It’s probably high somewhere, low somewhere, but within certain limits.
“Massagers”, yes, we are looking at it. Well, in general, if you go through this whole story, a very broad type of traffic is being picked up. I’m even wondering why this type of traffic is so broad, maybe it’s just the semantics that are associated with it all.
“Massagers”, “massagers”, “massagers” seems to be good, but when you select keywords more purposefully, you will most likely have better semantics than what AI Max generates. That’s why I don’t see any conversions so far.
If we take, for example, “Broad”, “Exact”, “Phrase” match, then, of course, there will be much more conversions in this whole story. So for now, the classic keyword selection wins. So, for now, these are the first results for two or three product groups. You can give me a rough estimate.
First impressions and conclusions
Of course, this is all AI. To say that AI should immediately give you a very good result is to say that it should be trained to work. Perhaps the change of some strategy variants overlapped with each other. Okay. But the question also remains open that the “electric scythe” and a little bit of “trimmer”, the selection of semantics is a little bit unsuccessful, and the AI does not understand what this product is about. Most likely, you need to analyze additional semantics somewhere, keywords that are set in the title, meta headlines, meta descriptions of your site, because, after all, it will primarily take semantics from there. If we take the AI that was advertised to us and don’t make a separate feed of the pages we need to target.