Кейс google shopping товары для детей Italy: сайт на шопифай
Hello everyone. My name is Yana Lyashenko, Google logistics specialist. I am engaged in delivering business to the target audience with the necessary parameters. I help businesses scale their online sales using Google advertising tools. In today’s video I will show the team’s case.
Brief description of the case
Market – Italy. Currency. Don’t be alarmed if anyone doesn’t understand – the Czech crown. What are we selling? Conventionally, handmade goods for children. These are not toys. Something closer to furniture and more.
Difficulties in the project
What do I want to demonstrate in this case? Italy. The first difficulty of the project is that the Shopify platform. The Shopify platform has issues with conversions being reported correctly. The number of conversions that we see, in reality, is much higher. Why is this happening? What we haven’t tried – various plugins, conversion tracking schemes, then we tried adding different tracking formats to UTM tags. Still, the Shopify platform from certain payment systems, primarily Shopify payments, incorrectly reports the number of conversions. When more than one person wrote a request to Shopify support to check the versions, which may differ, in each case Shopify wrote that the problem has been around for a long time and they are trying to fix it. Perhaps somewhere I’ll put here a screenshot of one of the correspondence with Shopify support. Conventionally, they answer the same thing.
This is the number of conversions that exist… There is no choice of what to see, there is only one conversion – a purchase. It is actually significantly more conversions, and, accordingly, the cost per conversion is also much lower than what we see here.
Work to maximize analytics collection
I will show the numbers that exist. The Italian market was very slow in terms of conversion. The different structures that they tried were a bit tight. We managed to put together analytics, with the help of which we were able to scale further against the backdrop of interruptions in conversions – this is a significant disadvantage. Why is this a significant disadvantage for Performance Max? They do not receive enough analytics to guide them.
We stuffed the account with additional transmission of dinremarketing data in additional conversions, which are in the additional tracking mode, not as the main optimization goals, so that at least he could see the indirect results of the work.
Of course, this is not an ideal situation. Ideal when initially he sees all the analytics from beginning to end, what happens to the product after switching to it and, accordingly, finishing the purchase. Well, it’s possible, it’s possible, conditionally. This is the main difficulty.
Categorization of products based on collected analytics
Having collected analytics, the advantage of a niche is that well-known “cash cows” appear, products that can generate conversions, you can increase traffic on them. They don’t grow as actively. There are “star” products. There are “dead dogs”, which can be translated in this project. Not to turn it off and forget, but to transform it into conditional “stars”, teaching performances to generate a certain type of traffic. Plus, there are “zero” goods that act as a resource for increase. I’ll decipher “zero” products, this means products that, relative to the rest or array of the data set, receive fewer impressions or do not receive them.
There are “months” – not zero months, wandering back and forth. Because there is no situation where in a new performance, where zeros have already been unscrewed, new zeros appear, the situation where all goods were evenly shown is not eliminated. Otherwise, this will increase the cost per conversion.
Shots as a result of the collected analytics
The situation is something like this. Accordingly, analytics were collected. Next we try on cross-sections. I call them lumbago, this is when we specifically try to increase the amount of budgeting for a set of campaigns, for certain advertising campaigns, for campaigns en masse. Here the structure is voluminous enough to understand whether cash cows will be able to proportionally increase the amount of conversion in pieces. Secondly, will the stars be able to, will the performances be able to generate more products of the stars. Star products are products that receive single, irregular conversions; anything less than 15 sales per month is considered irregular conversions.
Can we get conversions? So that they fit into the critical KPIs that the client set for us directly. Right here you can see the shot that took place. Then it was liquidated, it was reduced. It can be seen that plus-minus conversions remained fairly stable. Somewhere the cost per conversion was missing, but it’s logical. The performance does not fully receive the entire volume of sales conversions, but it is relatively satisfied with the performance of the period. In the situation with the image, when there were few sales and quite expensive cost per conversion, there were conversions in the Shopify admin panel. If we relate it to expenses, in principle, on the verge, conditionally, of payback. Not to zero, not a big plus what the client wanted. You can see a huge scary figure of expenses and conversion alone, while there are quite a lot of clicks. Now I’ll show you the picture in a slightly different format. I like to look at any performances in dynamics.
Analysis of performances in dynamics using average values
I will remove the clicks for clarity. There is a period of time. Now I’ll take some pink one. We measure by conversions. The average conversion value was approximately at this level. Now the average value for conversions is approximately at a different level. The increase delta is significant. Plus, if you look at the average value of the second period, these upward breakout deltas significantly exceed the share of downward breakouts and downturns in conversion pieces.
There is a similar situation here. You can raise this delta a little. Let’s try to lift it somewhere here. It should be a little lower.
You can try it here. All the same, the downward deltas are much smaller than the upward deltas. A fairly positive signal, in principle, of the work of the performance. This means that the traffic generated in the general array is of sufficient quality. This means that globally the situation with regard to teaching mechanics and selecting A/B performance tests is also going quite well.
Why is it important to pay attention to attribution
The only nuances are when they shoot through the cost for conversions. They are tied to the attribution model. Plus on this mechanics of customer behavior. I’ll show it now. Let’s go, on the way. Here “purchase” is selected. Here is the “purchase”. This is one single conversion.
What do we see? We see the most critical thing – what affects the cost of the conversion price. Average number of interactions before conversion. It’s 4.1. Just imagine, this Italian clicked an average of four times before making a purchase. Of course there are smaller numbers. Eat more. This amount was earned, relatively speaking, by people who took 12 interactions. Our advertisement was clicked 12 times.
Keep in mind that these analytics are only for completed conversions. We cannot target nose to nose, those who will buy immediately. We target those who want to buy a product, but they are asking the price, comparing, surfing websites, terms of purchase, delivery, colors, etc. This does not take into account the return rate, which is usually even higher in Italy. On average they clicked 4 times.
Here we add our metric as cost per click, where it is avg. CPC. Look, it’s 9, almost 10 Czech crowns. Just a minute, at least four times. They spent at least an average of 40 crowns just to convert it. This is average. There is more or less. The performance adaptively tries to adapt to the guys who are buying.
This four-interaction situation is made worse by the average number of days to conversion. The average number of days is 2.5. I had to increase the quantity for Italy. When the average was 1.2 days, there were fewer conversions than after. What does the number of days until conversion mean? This is the number of days when a person made a purchase decision. We haven’t looked at purchse in analytics yet. Purchse in analytics is significantly less than this conversion. There is an interruption, data transmission failure. There the figure can reach 8 days. There they collect data a little differently and there are significantly fewer conversions than these.
There are 104 conversions that were completed in the first 24 hours. Nine conversions were completed in more than 12 days. On average I take 30 days. Analysis always. Ideally take 7 days. 7 days usually does not show relevant analytics. You can see the real number of conversions for yesterday only within 72 hours. Google will change, recalculate, etc.
It’s the Italians, guys, how they buy things for children that are advertised now. Quite a big problem. Not a problem – a difficult task for most projects. Initially, it’s easy to spend a lot of money and have some conversions. Bring the performance to the form where you simply increase the number of clicks and you get a relatively good picture of an increase in conversions. Already a task with an asterisk. It is almost impossible to exclude it. We must remember an important big nuance.
Some features of setting up Google advertising
If everything comes together, all the cards are aligned, there is excellent demand, hot for this product, there are not many competitors in the niche, the client is hot and ready to buy now, then the most clumsily configured advertising will simply start generating sales. You will be completely satisfied until you reach a certain budget ceiling. Then profitability will fall. If there are difficulties or nuances in the purchasing process, this purchase is quite long. You just increase your budget. Profitability simply drops to zero. There is no result, conditionally.
How to see? Do you see the first shot, what was done in traffic that led to a significant increase in conversion? Well, supposedly not twice. If you observe, the volume of clicks here is about the same, relatively. They increased it – immediately the cost part simply increased, profitability fell, and the cost per conversion went through the roof. You don’t know what to do next. It is important to control the analytics that you feed your advertising account with. Artificial intelligence is built into the performance, which kindly learns from the data you give.
Features of the results of artificial intelligence in performances
It’s easier for him to handle any critical situations on his own, with a minimum of 50 conversions per week. Minimum 50 sales. There is no minimum of 50 sales per week. Perhaps we will reach fifty sales a day. He needs such analytics. If the market were in Europe, and this is Italy, where there is a Consent Mode for collecting cookies/without cookies. We need even more statistics. There, the system asks for at least 7 days in advance so that there are a thousand users who have denied access to cookies and a thousand that have given to enable machine learning. I got out of critical situations of subsidence and increased conversions. In short, to be more uniform, beautiful and stable.
Results
Naturally, you can’t just input 1000, 2000 events, which has no value; Google will always make money on you. Pay for every click. Whether you will make money is a big question. There are projects and niches where you need to be patient in the process of optimizing performance. Immediately, after launching a bunch of tests, they can blur the real avatar of the client, which the performance is about to touch. You can’t make sharply large shots frequently.
As you can see, it has been reduced to zero. They brought us to a picture that was more or less understandable, and did it again. You just can’t do that either. This is a huge block of simply intellectual work. Not just Ctrl+Z – Ctrl+V changed the campaign or signal just like that. This is a fairly large set of works, and they are more intellectual. This is thanks to the constant analysis of the advertising campaign.
I would like to demonstrate a complex interesting case, which I did. See you in new cases or issues.
