Case: promoting a sporting goods store through Performance Max

Кейс: продвигаем магазин спорттоваров через Performance Max Cases

Hello everyone. My name is Yana Lyashenko, a Google logistician. I am engaged in the delivery of the target business audience with the necessary parameters. Audience delivery for those who don’t know what that means? That I know how to find and bring it to your site with certain characteristics in the shortest way. Already, accordingly, on the site, a person makes a decision whether to buy or not. My task is to find more people who can buy directly from you.

Brief announcement of the case

So what do we have today? Case of the store, which has been with us for one month so far. I hope we will continue our cooperation. But here in December, the end of December, everything fades a little in terms of productivity and effectiveness. This is a classic story for everyone. But I want to share. I will show the dynamics, what the client came with, what difficulties there were.

What do we have here? An online store, we will conventionally call it sports equipment, because there is a lot of everything here, a lot of everything, and we have to somehow cope with it.

Indicators with which the client came

Difficulties. First. First of all, we were surprised that the client came with a rather excellent result. That is, the base here was pretty cool before that. By the way, I want to emphasize to followers who do not believe in the work of several performances, for example, I show you almost a month of work of our team here, 2 thousand, 1910 rounded amount of Purchases/Sales achievement, this is a purchase. The average ROAS is 882.70%. There is somewhere lower, somewhere better. Cost per conversion here, on the green graph, you can see that it was jumping, jumping, but it came down.

The client came with a very good base. It is impossible to produce such a result, even the good base that was before this client at one performance. Therefore, these adepts who believe in two performances, especially with copies on different feeds, guys, I want to disappoint you, there are a lot of campaigns here. There are only about 20 or 21 test campaigns here, not taking into account the basic part of advertising campaigns, which have simply not yet been significantly moved. I will tell it now. Maybe it will be quite useful for someone.

What is the complexity of supporting existing advertising campaigns

It is problematic to manage a large number of campaigns

The nuance is that it is very difficult, first of all, to manage such a large number of campaigns. There is quite a decent amount of them here, it is impossible to take them out in two or three performances. Google’s ROAS would drop to two or three by such an amount. If you wanted to bring it out in three performances. There are some who perform alone and believe that this is not a way out of the situation.

Increase indicators without disturbing the existing one

The second difficulty. That is, it is necessary to break through this ceiling and make it stable. The second difficulty for us personally is that it is important not to break what was in the account. That is, I will tell you as a specialist, it is very easy to reach the target KPIs for your own campaigns simply by dragging traffic and conversions from the base with which the client came. To show that, well, they’re broken, they’re old, but our campaigns are growing there.

We in the accounts do not welcome this method. Our task is to support the general dynamics and, in principle, how to say this correctly, better not to offend, there is such a principle. Our task is general growth. All the campaigns were basically managed by us, but we did not redo the structure by completely demolishing everything, as some specialists like to do for some reason. Do not tear it down, support it and smoothly transfer it to the structure that would suit us, for example. And we would like a certain categorization or breakdown by some other criteria so that it would be convenient to scale it further. That is, the life cycle of this performance was much longer. That is, so that it does not stop, but gradually increases, increases or maintains some stable plateau. At the same time, the entire product range worked. I will say that it is very difficult to do just with some primitive breakdown. This is very difficult to do simply by dividing the goods into groups. This is very difficult to do just by doing a price breakdown. There is a very complex set of big factors. The task is not to break, but also to increase. Perhaps now I will show where even the correlation of growth was in the general dynamics, so that you understand what was happening.

Conversions were generated only in certain categories

The next difficulty was that due to the fact that a very large number of conversions were generated only in certain product categories, other categories were simply not very active, the first – they went to shows, the second – they directly received relevant sales. The reason is that your performance always takes account-wide user behavior analytics. This is a fact.

Well, whether you want it or not, even if you don’t have dynamic remarketing there, you just have a goal connected there, a purchase, it will also take the mechanics of user behavior from it. Therefore, if you have a very large store with a large assortment of products, it is important to remember that some categories must be manually forced to be displayed. The second is to bring them to profitability, because mere performance alone cannot do it. If it was a new account, most likely these categories would have gone in faster because there was no analytics. On the other hand, it does not make sense to do this on a new account, because it is better on the existing analytics. That is, your task is not only to support the result, but to increase it. I will show that our budget has increased a little, and in conversions there are almost 70% of these conversions of the total account. Well, just for a minute, there is only 14% in the budget, and in conversions there are 62 with something of a percentage, from almost 70. Well, if it was not done simply due to the increase of the budget. I can even show it in the history of changes, if anyone is interested. So there is a set of difficulties, there is a base that it is important not to break and it is important to increase it.

Site visitors do not buy products from the landing pages

And one more important nuance. In most niches such as this one, there are many others, for example, the same auto parts, there is a very large nuance of user behavior. It consists in the fact that people can go to your site for certain products. I call them entry positions, introductions. You can call it what you want, but they can sell completely different goods. How to check it yourself? You just go in, take some great period, at least 30 days, or it’s better to take more. What do you do next? Be sure to go to the “Product groups” tab, look at the products there, segment by those that had a conversion.

That is, if the number of conversions is at the level of the performance itself, provided that you have a performance without objects. Because there is also a need to make adjustments to this error. I have them without an object. I am a believer in smartshopping, and not in smart types of all campaigns. Although it can be tested. This should be tested. If you see that the number of conversions you really have less on products than on the campaign level, congratulations, most likely somewhere you have entry positions through which people very actively enter, but walking around your site, they buy completely different products and also see this performance.

Why duplicating feeds is a very bad idea

That’s why I emphasize a lot. I don’t like when you duplicate feeds. When you’re doing a feed index. Especially when you are a very large store, 70-80% of your product range is simply not advertised, nothing is shown and, accordingly, you do not get a result. It is different in different niches. For example, take a large power tool store there. It was enough to just get you into impressions, because if you don’t have enough analytics for conversions – just make maximum impressions for all products – it will be simply expensive. And if there is analytics – it will be much easier.

So, here we concentrated our attention as much as possible on bringing, at least during this period, the advertising account to a more or less stable picture in terms of less competition between products. Remember that each of your product IDs is a separate product item, it doesn’t matter if it is loaded in Russian or Ukrainian in three, ten languages – it is one product item. If you make a duplicate of it, it is a completely different product. You make the work of the performance itself so difficult. I’m sorry, the performance changes the algorithms of its work. On duplicate feeds, it can’t scale in the long term like without duplicate feeds.

The client had an overlap of the same product in the campaigns

The second is that in most of the basic advertising campaigns that the client came with, there was a lot of overlap of the same product. And here it was important not to break the general dynamics by assigning priority bids, dragging, making, scaling certain product positions in impressions, respectively, in clicks and conversions. Well, because one campaign at a time will be a complete failure. On the other hand, some significant growth. And you are essentially killing this campaign.

Therefore, it was necessary to maintain this balance and gradually bring advertising campaigns to the point where there are no identical IDs in different campaigns. Or if it is only done with a specific purpose there, testing some types of conversions or helping some specific product categories in conversions, etc. That is, it was also a certain difficulty.

The task is that the entire product range competes as little as possible for the same, the same product at the shows. Because some will get conversions, some won’t. Somewhere you will squeeze the betting strategies, somewhere you won’t and, conditionally, the result will be anything.

Therefore, there is a set of certain difficulties. This is always on large accounts, where there are many or many conversions per month. Or on accounts where, for example, a very large product range always collides with this. The more conversions you collect, the more nuances of instability you have.

How the account grew by category as a result of work

Yes, what can be emphasized here. What to pay attention to? I would like to show you here, a demonstration of how the entire account has grown by category. Categories will be overlooked here. You won’t see which categories, well, to preserve customer privacy. But look at the period of equalization of conversion, our purchases there, there are periods of duplication of purchase, but there is a very small percentage, maybe up to 5-7 percent of this amount at most. But what do we see here by categories? Well, according to categories, the main categories here are divided into Russian and Ukrainian. Here they blur a little.

But let’s take the first category. This is Russian. And this is the Ukrainian version of the feed. For a minute, just look at how much they have grown. Just how much they have grown. Plus 50%. This category, I will tell you, converted very poorly in general. Here, please, the increase in conversions. And next, if you look at the main mass, here we have a certain category, which there remained at the same level, but the main mass, you see, it has grown. Well, it’s just not the season for this category. I will not reveal what it is. This one indeed, but there were few conversions here and before that. It is very difficult to scale.

What you would see is just the general dynamics. The same general dynamics can be seen on this chart if I just take, for example, a slightly longer period. And here you can see a jump in conversions in the period when we start working. Well, I think, here you can see with the naked eye, you can see how much the price for the conversion has decreased a little. Here you can see, yes, how stable it has become. It can be seen, yes, there are jumps, breakthroughs, conditionally. But how stable has it become?

Well, it has increased here, because we had both a Global Site Tag breakdown and problems in Kyivstar. All this also affects all this work, but it is clearly visible that the slide has gone. Mountains of analytics, conditionally, have gone. Therefore, it is approximately visible that the total price for…

Comparison of the period November 17 – December 16

If I take, for example, the seventeenth. Yes, I even compare 17, 16 with the previous period. It will even be seen here that the price per conversion has fallen. ROAS we raised. The number of conversions has increased, and the number of purchases has increased.

And, as you can see, the amount of traffic has decreased here. I can even show you the bones, that is, we have 53.3. Yes, a dice increase of 53. I’ll show you now at the campaign level. Here, all names will be obscured, of course, with the aim of preserving the client’s incognito. Here it will be. Let’s do it. Here it is. It will be seen here that the expenditure part for all basic/non-basic performances increased by 18%. Conversions, for a minute, by 75%.

She said that there are some conversions that are duplicated, so I will demonstrate it so that everything is objective, and not as if I just showed conversions. I’m just a non-believer. Yes, it will all be obscured. You won’t see it all.

We take Total Performance Max Campaigns. Let’s see Conversions will be observed. I’ll tell you where she’s shopping in a minute. The buying target here is 68.82%. Although I said 62. I must have watched previous episodes.

Look, the cost part for all performances plus 20%, let’s round up the larger side – 20%. The cost per conversion has dropped. Profitability increased a little and conversions increased. Attribution model respectively. Well, income will also be visible here. It is impossible to do this with just two or three, maximum five performances. If you see here together with all, even deleted 81. I will not show how many of these performances there were. But so that you understand, approximately the overall scale of the picture. Well, there are some periods where conversions were duplicated, but you can see the result. That is, by how much… That is, we are not only slightly under budget… Yes, we slightly exceeded this budget within the limits of what the client asked for, but how drastically we increased in conversions.

We have the next accounting period for which we will launch, our task is to further reduce the price per conversion. Because it’s all right here, a data-driven attribution model. It does not completely satisfy me in terms of the adequacy of data display. I still plan to reduce it in the next accounting period. When will it be? Will it be after the new year or now? Let’s see. Maybe we will have some period of support, but conditionally.

What measures increased conversions

That is, here is an example when a person comes already with a base. He wants her to be managed, not broken. They increased and even brought the conversion to the top. Here is an example of how it can be done due to proper segmentation, selection of certain categories. Starting with small budgets is what I always say. It’s always a question of profitability, guys. This is not a maximum conversions or target cost per conversion strategy. This is direct work with Target ROAS on campaigns without objects, without pictures of all of them. This is a launch on advertising campaigns, which in most cases are without signals, if you have a good feed.

In general, I think that the signal is a bit rudimentary already in some cases. It is possible on new advertising accounts. Maybe until you install the optimal version of the feed. Here we had a pretty good feed with a base. We received an account with a fairly good conversion base, but you know, everything can be hacked very easily, conditionally, in literally a few minutes.

Why product IDs must match

And here, by the way, we had the same IDs. This is a very important nuance, because we use commodity optimization, commodity analysis, commodity optimization in our work in performances. Therefore, it is very important that you do not duplicate, do not do any nonsense in your feeds. Because there won’t be a fairy tale in your feeds, when you make one feed for standard shopping and one for a performance, or one feed for one performance and another feed for the second performance, and depict it in some ID. And the system does not see these IDs on the site, because you take one product item, bring it to the site, and then everything is dark. A person receives analytics by another ID. Plus, take into account that the competition is based on the search queries of the audience that enters them.

In the previous case for men’s accessories, I showed how sharply the conversions on the performance, which was with native IDs, increased. For those who don’t remember, I’ll leave a hint somewhere here. Just look at this case. It is long, I know how to listen for a long time. But this is a problem when you use non-native IDs. Because it’s a pain. Not only that, it turned out to be a pain for me when I found out that some clients exchange them artificially on the site itself. That is, in the feed there are some ads, and on the website they are completely different. It is also pain. Because the functionality of commodity work is disappearing from you. Because the product is an advertisement. And your task is to teach the performance to operate. It is very important exactly the commodity analysis. Because the more your goods, the more it is, conditionally… How to say, how to name this product, the more it is not that it is flexible, but changeable. The Ukrainian word is this. “Changeling”.

Yes, when something that was in stock ran out, was “out of stock”, was thrown out, it is necessary that the performance does not fail completely due to these conversions. And he adequately switched places, that is, he pulled up another product position and changed it in a different way. Sometimes this cannot be done even within one category of goods, because the performance simply dies due to the fact that one or five product positions are out of stock. The more such changing goods, the masks, the more difficult it is to teach performance. Therefore, this vibrant analysis is needed.

? If you do not have the opportunity as a prom, we will do within what we have. If these are other platforms, such as good. Do you have the right Aidyshnika somewhere – it will be a little faster, a little lighter than our team cases show.

Why should there be many performances

In principle, I posted another case at the request of the bosses, conditional. To see what I say about, for example, whether in telegrams or chats, it is not one performance if it is scaled. If there is a certain loan in sales that you cannot break through or which you punch with the decline of profitability, then do not think that simply relocation of the advertising campaign or reduction of the budget, or simply increase the budget somehow edit this whole situation.

No, you don’t have to think that you will take out several performances. No. The larger the store or more range – immediately get ready that it should be many performances. Not for new accounts, not for accounts, which are less than 100 sales per month. Most likely, there is no point in playing it only to some limits. Do not marry immediately behind the pile, the number. The structure here is not broken in this account. A lot of nuances. She personalizes under each of the projects to understand.

What are the parameters to analyze the product range

There are tops, there are not tops, there are weaknesses, there are doy cows, cows, died dogs. This is the whole structure, which, in principle, is used to analyze its product range there. Here is the same story. Your task is to analyze the product range, at least three parameters. Effective cows, your task not to kill them, to feed them, to bypass, so that they give more giving, happy flowers, cinema. Dogly dogs, they need to be overdo it. Because the dogs are dogs, they can take on costs. Potential asterisks that can go into active cows. It is important that the potential asterisks are very quickly passing into actual cows. This, in principle, structure is done. Not just by type or price, but on the calculation of these all stories.


nuances when it doesn’t work out. This is when you have very little analysts in your account. Or it seems, but in the categories that you launch it is not bother. This means that the audience that converted to that it had other indicators as a result. Other indicators in behavior on the site. The second factor. Maybe you do not have the Idshchi. The third factor maybe people come to you through one product, buy completely different. Your task is to test these other products, in principle, whether they can generate traffic on their own, whether they buy them because they look more presentable in the background of the previous ones. And without equation with the previous ones they simply will not convert. Even this happens. If, for example, a very single -shaped product range, such as paints. You have more than 8,000 headings. It seems to you that advertising is very expensive. But look just or the same electric tools. New accounts where analysts have had very little or very unprofitable conversions. You are married by traffic volumes. It is also wrong. Because not all of them in proportion to conversion correlate.

Case when a conversion price can be expensive

It is important that you need to understand that sometimes the price for conversion is a road not because performance is inadequate, but because you have 8,000 goods there. Each will spend, for example, 5 UAH. Give 1 hryvnia. Each heading will receive one click of 1 hryvnia in one day. 8000 hryvnias What is the probability of buying from one click? Well 50×50. Conditionally, this can be 20% of this traffic. And may be 5% from this product only. The likelihood that they will pay off both yes and no, conditional. Looking at what price there, which will buy from you. For you to think large -scale, they were approached by the Analysis. I do not demand from you that you go to each heading, analyze the readings whether it has or does not have. Performance will understand this if he has the correct analytics, the correct correct fid.

Why don’t you need to do duplicates Fid

Do not marry duplicates. The tap should only be done in some cases where you do not have a way out or you have a exact task to promote a certain small product range. Then there is a sense. But don’t expect it to be once once, doubl there, I don’t know how many times.

Correctly break the goods by categories

the second of the recommendations. This is if you have the same product, such as paints, no matter who. These can be spare parts. It can be, for example, power tools or stores related, garden there. Don’t think, go to your performance in depth. If you have these expenses, shows between all goods, then it is expensive not because it just does not understand what you want from it, but just a very large product range in one campaign. Each at least two, three players spend. It seems very much. Very expensive because you are given an average. You will try to cut on some small spare pieces. Not everyone has enough patience for it.

But this is the nuances of some categories. Although he uses artificial intelligence, he is not perfect. It is not suitable for everyone. Yes, he has an algorithm. This is, by the way, why you still get sick of Fidi. You have an algorithm that is sewn into this artificial intelligence. You just need to give the right database to give this artificial intelligence, it will give you a cool result.


Do not engage. For example, you have a Tesla conditional that has a certain basic functionality. Here she was released from the plant, she has changed there. Google is doing something good, something is good for him. Your task is to drive the machine. Not so that you put another rudder manually (duplication of Fid). Turns one in one side, the other to the other. Wheels, if conditionally, they diverge. Or you are attached there for some reason, I do not know, some horns on this carpenter or there, I do not know, the parachute is done behind. This is what you are artificial there, you see something about performing performance.

It is very difficult to operate these performances sometimes. But basic, if you sell the stretch, do not marry the volumes, budgets in the first weeks. Because it can stretch for a certain period. But it will be worth it. Because you just slowly collect analytics. It will be the easiest method of precision of your performance to generate the same base.

this in principle everything I wanted to say in this case.

Яна Ляшенко
Rate author
Add a comment

Sergey Shevchenko Google Logist GoogleLogist
Google Ads 90 Days Package, will help make your advertising campaign not only profitable, but also increase sales from it