Hello everyone. My name is Yana Lyashenko and I am a Google logistics specialist. And I deliver it to the target audience of the business with the necessary parameters. What does it mean? This means that I can find your target audience on the Internet. Using Google advertising tools and producing your site. Accordingly, you will already receive sales.
General dynamics since the beginning of cooperation
And in today’s video I would like to show one of the cases that may be interesting and indicative for someone from the point of view of what I talk about Performance Max either in individual consultations or on the YouTube channel. So you can understand whether it works or doesn’t work. Plus, this is really, well, not just an indicative, yes, case. When it is truly possible, what Google creates with its own codes can play both a good positive and negative role.
So where do we start? Here I show the general dynamics of the advertising account. The client has been with us since 1919. Somewhere in this period. Now we will show exactly what dates were. Here you can immediately see the analytics mountains in terms of conversions that exist here. You can immediately see how this period, which we placed an order here, is much higher here, and the cost per conversion, here, accordingly, you still need to navigate and look.
We have a completely different strategy here. Well, in any case, now for October 23 we see 5755 conversions and a total average cost of 48.39. There is a lot being tested there. Now we will all talk about this.
The request the client came with
The client came with a request for what it was not possible to stabilize conversions after switching to the fourth analytics. That is, they were spasmodic. They were very expensive and, in principle, even an ordinary product did not show itself to be stable enough. Here, if you look at the types of conversions that the client tracks, there is a lot here in a row. This is due to his efforts to somehow stabilize and systematize everything that was in the advertising account. Yes, we’ll see now.
Operation results since July 2023
But this is the period from well, you can see it here from July. These are the works of our Performances. Whatever has not been tested here in general, here you can see something more or less stable. Here, the period of failure will now explain what it is associated with. And here we finally managed to break through a certain ceiling. Basically, the client came with a specific request to stabilize advertising campaigns and reach a hundred sales per day. That is, his main task was ours – to reach a hundred sales.
We started with little. There are 60 conversions visible here. I can even show you the conversion type. If anyone suddenly doubts that these are purchases or not duplicate conversions.
This is all the purchase tag is done. This is a conversion tag, which is analogous to purchase, but Google Ads. Here we see 87. Here we have finally reached this period. The client didn’t really like it. He was not ready for these volumes. We’ve shrunk a little. And again, well, 85 is also visible, the volumes have increased. And our task now is to very significantly break through this ceiling of 200 conversions. Here is precisely the period of breakdown of the global tag, from which a certain instability began, unstable operation, in principle, of any automatic strategies.
Well, in principle, there was a lot here, tests were carried out to ensure that different product categories were approximately the same, plus or minus, in terms of volume and conversions. And here, in principle, for now, the ceiling has been broken for now, and we will move on. Well, here the conversion price is reduced. For example, Wednesday November 8th there are 227 conversions. Their price is 43 UAH, respectively.
Main accents during the work
So what did you want to emphasize here? What would you like to focus on? Firstly, we started with Performances, despite the fact that the client came with certain problems in regular product advertising. He came with a regular product where the ROI strategy was used. It was not possible to scale this result. It was either completely leveled out due to an increase in the cost per conversion, or this amount was significantly small.
So when I say Performance is one of the tools with which you can scale. Here, for example, based on the example of this client, I show that this is indeed the case.
The need for narrower segmentation
The second aspect is that one of the typical mistakes of projects that large, large ones want to reach a fairly large turnover is that they do not go into some kind of narrower segmentation. That is, they usually have one, maximum two, two Performance Max campaigns of the same. And that’s it, they are no longer moving anywhere in this direction. This is also a signal for you that I now have 19 active campaigns here. There are many more of them here. A lot of tests were carried out here.
Now I’ll tell you how many of them there are in total, excluding the remote 45. And all this was carried out during this period. Certain tests in order to understand what works best, what type of grouping, what type of bid setting strategy, what profitability, structures, etc. work best for this client.
That is, the task is to form a set of advertising campaigns at the beginning, which can act as a traction force, a stable generator of conversions to increase and improve these volumes. Thanks to it, you will, relatively speaking, accumulate enough fat with which you can use other campaigns. Quite different strategies were used here.
Starting from the launch on the category, by the way, it didn’t work out for him right away, although in an analog project, in another niche, this is in another project, another site, it worked out quite well. By category. Then the price. Then brands. Strategies, so you understand, even several strategies worked for the same product range. Restarts, etc.
A properly optimized feed is required
The bottom line is that a properly optimized feed does a lot of the work for you in terms of optimization efforts. Well, not so much for you. This result, in principle, could not have been achieved without work. But a properly optimized PID, properly tailored, especially for product types, allows you to not play with signals at all. That is, in this project there are very few similar products. In almost every product category there may be one or two product positions. Everything else is in different categories.
That is, the product range is very diverse. Therefore, in principle, to create performances for all categories there will not be enough for us, in principle, the limit on creating advertising campaigns. Secondly, this makes no sense. And thirdly, it is also impossible to create a universal signal for a very diverse product range. Plus Performance itself, it takes more targeting from your feed. And here, the right high-quality feed and correctly grouped advertising campaigns allow you, in principle, to grow further without playing with any signals.
Here, in principle, there may be some category or remarketing alone with a signal, but here in the tops, for example, of these campaigns, I now don’t see a remarketing campaign at all. That is, these campaigns are completely signalless. If you go into a group of objects now, you may not even see the signal. There are no signals here at all. This is one of the nuances that almost all campaigns here have, they are all without signals.
Disadvantages of advertising campaigns
Yes, the only downside for us began to occur when we had a period of failure of the global tags, which we personally found to be the same in different accounts. And our task was to break through the ceiling of 200 sales. Here you can see 140, 126 conversions at 72. This conversion price was even more expensive. We managed to do this only here to stabilize these performances.
This was precisely due to the fact that in key moments like these, the very robotic ability of Google itself plays a very important role. From Google’s side. You can see it here.
A summary that can be useful to take away from working on this project
What else can be useful, if we summarize, what useful things can be taken out of this? The first thing is to never think that Performance alone can achieve the results you want. Yes, he can. But all the same, sooner or later the question will become either to keep it profitable or to increase the number of sales. And to keep these two blocks the same is both profitable and sales.
Therefore, the client’s main focus was precisely on the cost per conversion, which he wanted to receive on a budget, respectively. How much is not profitable, how much is the price per sale. So you need to go deeper. That is, dividing your product into different advertising campaigns. At the same time, you can have the same product range in different types of bidding strategies and this will also be OK. And they will generate completely different results in different ways in different periods, time periods, etc. And this will also be quite good, relatively speaking.
You need to test different structures for yourself
The next thing is that you will need to test different structures for yourself. That is, you will not have a perfect structure. A structure was launched here that worked very well on another project. And then it didn’t work and had to be remade into another one. So be patient and don’t think that Performance can generate, I don’t know, millions of sales for you in one second. As my client said: “Make me rich,” in one second, and it will be very soon. You need to be patient.
The more analytics you get, the more presentable the next result you get and you don’t think that everything will be so easy. Even in this case, simply by increasing the budget it was not possible to break through the ceiling of 200 consistently. During the period when the global tag broke, this also came out during a certain period or tests of some tiny or split structures and others.
It is necessary to control products that are not displayed
And one more important nuance, always control which product range is not shown at all. That is, there is a large part of the products in your Performances that are either not shown at all or receive a very small number of impressions. Don’t expect that if they sit in the same Performance, they will, at some wonderful moment, start working. No, this is a completely broken story.
You will still have to make them work. And they earned it – this is certainly not the case with all existing advertising campaigns. Well, that is, for you to understand anyway, it comes down to the fact that you will need to duplicate the campaign.
If you have a niche in which the store has everything in a row and there is no representative selection in one category, there are thousands of products in one category, then this is also not a problem. It is not necessary to group them by product type. You can group them all together. And this will work for you too. That is, you can also focus on this.
Under the same conditions for launching campaigns, Hoshopop wins compared to Prom
And one more important nuance here is this. With analogue launches of two projects on Prom, Hoshohop, Hoshohop wins here. I attribute this to the fact that Prom is an excellent, wonderful platform. I actually like it because it’s a great starter. But due to the fact that they do not transfer normally adequate data for the same dynamic remarketing, they themselves clip the wings of their own users of their platform.
I don’t know whether they will correct it or not, but under analogue conditions, conditionally, in stores, in niches – Good now, as you can see, is winning a little. Yes, it also suffered due to the breakdown of the global tag site here. But it can still be stabilized.
The need to control performances in movable goods
And one more important nuance. This is that if you have the same flexible product range as in this project… Mobile – this means something runs out of stock very quickly, the status is “Out of stock”. Something else appears, a new assortment is constantly appearing. Something is added, then you must constantly monitor the effectiveness of your Performances, because often your Performances have the same error.
The same positions are converted. And as soon as they are no longer present, the entire Performance Max breaks down. This is because it is not trained to correctly distribute impressions between all products and is very dependent on only a limited number of these products. As soon as they simply drop out of advertising, it completely breaks down. Because his entire learning process is based solely on this.
How long does it take to learn performance?
And this is the answer, by the way, to the question of how much Performance Max studies. He studies constantly. You have to remember this, this type of advertising campaign is so cool because it can simultaneously use Google AI. At the same time, he uses machine learning and builds models with which he finds new audiences. But he is constantly learning.
That is, the first five days when he draws you the learning status – this is the basic status, which is initial, but it continues constantly. And if in this case you have some product assortment dropped out of the auction, and a new one has arrived and it breaks. This suggests that no matter how long it has been running, it is unstable. It depends on a specific product range.
Therefore, if you try to break it artificially at least once a month, all the more so you will take over the control mechanics of even these manual performances, regardless of what is going on there from the point of view of the artificial intelligence itself. So that he is constantly in good shape, and does not become dependent on a certain product range. Basically, this is a small case.
There is one campaign right at the top. But it also worked the longest, which is why it generates the most conversions. And there are several more advertising campaigns that are at the top. And our main task, in principle, is to ensure that the last categories of products are close in terms of the number of conversions to the point that they are all almost identical in terms of the number of conversions while maintaining an agreed upon target price per conversion in certain product groups.
That is, in each category we have an agreed price for the conversion that exists here. Somewhere it stands out very much. Somehow here, for example. Who stands out less and these campaigns in general have now reached almost 50. There are currently 19 active ones left. Some of them have just begun their own career path.
Brief conclusions about this case
Basically, this is all that can now be revealed about this case for you to tell. If we cut again, first of all, let’s not focus on just one Performance. Not just Performance, but deepening your segmentation.
Next. Segmenting your products according to different criteria. This should be done in test mode. Tests only. Because we never know whether the same strategy that worked on one project will work or not work for you.
Next, always monitor how adequately Performance distributes impressions between products. If he receives impressions and conversions only from a small sample, then it goes out of stock, everything collapses for you, this is unstable Performance. It cannot even be continued at all. Most likely, you need to create a new one.
The next thing is that, if possible, try to switch to all conversion tags related to Google Ads, Google Ads. That is, these are purchase tags, these are conversions imported not from the fourth analytics, but directly from the Google Ads tag. They do not guarantee one hundred percent stability, as practice has shown. But still, unlike the fourth analytics, they still work better and generate conversions better, which are more stable overall.
And the last thing is that you always control what result Performance gives you. Remember that he is constantly learning. This is constant work on its structure, on it. If you do not control, do not monitor the result, do not assume that after a while it will pay off. He either gives you a result that you like or doesn’t give you. With smartshopping, there was still a time when you could advertise at a disadvantage, collect statistics and dance further. Now this kind of story doesn’t work. Either everything is right at once, or everything needs to be radically reworked.
