Exploring the AI Revolution: How Artificial Intelligence is Reshaping Amazon Selling

Transcript

Exploring the AI Revolution: How Artificial Intelligence is Reshaping Amazon Selling

Note: This live presentation was designed to be watched, not read. We strongly recommend that you watch the on-demand video, which includes charts and visuals, not in this transcript.

Victor 00:03
I’m going to talk today about AI, so I’m not going to talk explicitly about Feedvisor. I probably will give you a couple of use cases, what we do, and how it relates to AI. But this talk would be a little bit scientific. It will give insights into what’s going on behind the scenes, how the AI operates, and what it really means.

Victor 00:27
When you think about AI today, after the revolution of chat GPT and everything else, there’s an underlying AI that basically lives within the models and that powers the chat GPT engine and generative AI, but AI is much broader than that.

Victor 00:51
We’re going to touch this during the presentation to really see and understand. But then there is a layer of tools that basically lives on top of the chat GPT. The chat GPT basically powers the whole AI thing by modeling, by being able to generate these artifacts, be it text messages, emails, content, or any other sort of automation that generates.

Victor 01:22
These tools basically just connect the API. So, there are a variety of tools out there that leverage generative AI or other types of AI, but at the very end, all these tools just connect the API.

Victor 01:39
So, today, I’m going to go into the basics of what AI is today and what the AI landscape is. I’ll try to give you a simple way to look into the software and say if it’s true AI or more like a quasi-AI, which pretends to be AI. There is nothing bad about it, but that is where the line is. Then, I’m going to dig into a few use cases, and we’ll see if it’s true AI and why and how you really judge it. We’ll finish with a Q&A and then recap.

Victor 02:24
So, let’s start at the very basics. When I started Feedvisor 10 years ago, my vision, my idea, was: How can the advancement in AI be brought into Amazon selling? Because, in the end, selling is a concrete, logical process, right?

Victor 02:48
As you get the good, you put it on Amazon; someone buys it, so where does AI come into it, right? It’s not so trivial. AI usually comes in our ability to use computer systems capable of performing complex tasks that historically only humans could do.

Victor 03:10
These tasks usually involve reasoning, making decisions, and solving problems. So, when you think about this for a second, AI doesn’t mean that everything, that calculator, is not an AI. But when you, as a person, think about a problem, when you think you need to make a decision, this decision-making is what differentiates between a human on one side and a computer on the other side.

Victor 03:40
If you create a prescriptive rule, “so if event A happens, do something.” If I get a sale, lower the price. If I get the sale, I increase my advertising. That’s not an AI because you are the one who made the decision.

Victor 03:55
You created the logic. On the other hand, if you give the computer your objective and say, okay, I want to increase my sales, and I want to watch my margin, and I want this to happen given the reality of the marketplace and the computer makes certain actions, that’s where the AI comes into play.

Victor 04:18
Again, the definition is AI replacing human activity in this case. So, let’s go one step deeper and see what it really means. As you all know, the whole thing started with movement from what we know as an analogical world, where everything was written on paper.

Victor 04:44
Everything was verbal, not necessarily recorded. This data couldn’t be processed or analyzed, so the first step of innovation came with the fact that data became digitalized.

Victor 05:01
This data started to go through a computer, and as it started to come in, it started to get recorded. This is where the concept of “big data” came in, so there’s a lot of data.

Victor 05:10
It’s called big data for a reason. But then—and this happened around the 1990s or 2000s, more or less—as the data began to be stored, and everything was great, what do we do with the data?

Victor 05:25
I have tons and tons of data. What should I do with it? Then, machine learning advancements started to come in. Machine learning basically takes a lot of data, scans it through a machine, and then tries to figure out how to make sense of it.

Victor 05:43
If we see, for example, that event A usually coincides with event B, and we see a clear causality, then we have that event A always precedes event B. As a human, we always understand that that means “If I need to make event A happen, then event B will fall”, and then there can be more than one event — can be event C, event D. Once that starts to happen, then we understand that there is a certain flow of causality between them, and that’s where we come to the reasoning.

Victor 06:22
That’s what the machine learning does. Machine learning helps us to understand how the whole thing works from point A to point C. If I’m changing, for example, my price, how is it going to impact my revenue, given the competition, given the market, given the responses of either, given the advertising, the traffic, there’s so many things that come into how, at the end, I can achieve the goal.

Victor 06:48
As humans, we usually look at data, Excels, and try to figure it out. Machine learning, combined with big data, tries to achieve the same thing.

Victor 07:04
So this is where the birth of AI came in. How the machine can basically replicate or replace part of what we do. That’s number one. The second huge innovation that actually happened relatively in the last decade, I would say, is the concept of reinforcement learning.

Victor 07:24
This concept was introduced about 10 years ago in academia and started to be implemented. Today, it lies pretty much behind a lot of what chat GPT does, but not only.

Victor 07:40
So reinforcement learning means that when something happens, you encounter a result. And when you encounter the result, how will you respond to it? That makes machines much closer to humans.

Victor 07:56
So, we humans think, okay, this is the action we take. This is what happens as a response. As a result of that response, I need to change my course of action, evaluate the result, and then continue this process until I reach the outcome.

Victor 08:15
That’s what’s called reinforced learning. The system can take action, learn, take action again, learn, and take action again. This leads to the concept of autonomous systems.

Victor 08:29
An autonomous system is something that can act on its own. It’s not like a crazy robot running in the movies, but an autonomous system means it can act on its own. It doesn’t require you to give clear instructions.

Victor 08:44
It doesn’t require you to prescribe the behavior; the system can itself, through reinforcement learning, understand the impact and make the decisions. Finally, the last thing that really happened in the last years…

Victor 09:01
The explainability of how to interpret the model. Everything is a black box. It’s something that runs on the data, makes these Inferences, makes the decisions, and eventually takes research action. So a lot of people will say, okay, so how did you reach that conclusion, not this conclusion?

Victor 09:20
That’s where the model of interpretability comes into play, and there was also a lot of technological advancement in that field. So you combine all of that, and now you have the AI revolution.

Victor 09:33
It’s something that can take the data, understand the data, and act. It can reinforce its action through learning, and it’s also explainable. It can also show you why it achieved its actions in one way or another. So, when you think about all these concepts, how does it really transform into the Amazon selling world?

Victor 09:59
What we’re dealing with day-to-day transforms a number of fields. First of all, it transforms customer service, which is basically using chat GPT. It actually existed a few years before, but chat GPT really improved. It is an ability to understand natural languages and respond. So when you today again, it’s not only an Amazon—almost everywhere. We know today that eighty percent of customers interact with a customer representative.

Victor 10:29
It will not be human in eighty percent; again, it’s using the same concept it knows to reinforce to learn and reinforce itself. It knows to respond, it knows to process the information in real time. So all of this is incorporated, and this is where the AI comes in. Inventory management or even more inventory forecasting is another example. So you’re running a certain set of inventory, you forecast how much inventory is left, your current inventory velocity.

Victor 11:04
But then, you take certain actions: you advertise, you change prices, seasonality—there are a lot of factors that come in, and the system learns. Again, it’s capable of forecasting more accurately given the change in the world and continuously, and you proceed with this continuously.

Victor 11:22
This enables, again, revolutionizing inventory forecasting and inventory management with the help of AI. Listing optimization is probably the most trivial application of all. It basically runs the chat GPT and helps take certain information and create your listing.

Victor 11:42
This is the most known and trivial application of AI. Dynamic pricing is what we pioneered. It’s how to price. I’ll talk about this a little bit more in the use cases — how to price correctly, given the constantly changing world be it you do or don’t face competition. The other area that we, feedvisor, operate in is advertising optimization. When you want to optimize advertising optimization campaign management: where does the AI come in? It finds the right keywords, and we’ll talk about a little bit more, it comes in being able to set the right beat, given the competitive environment, and obviously in managing budget allocations. There are obviously more tools, but here are the certain aspects we’re going to touch today.

Victor 12:39
As I promised, there is a very easy way to tell the difference between true AI and quasi-AI. Now, when we say quasi-AI or rule based, there’s nothing bad with software that is not AI. Nothing bad, but it doesn’t mean that if you have an Excel — that it has AI.

Victor 13:00
Saying something is not AI doesn’t mean it’s a death sentence. It’s not to say “okay, I should stop using that software forever because it’s not AI.” But still, it’s important to understand what AI is and what it is not, so how can you figure it out?

Victor 13:20
There are only four questions that you can ask in order to understand if the software is a true AI. The first one: Does it or does it not rely on data and algorithms — and does it make autonomous decisions?

Victor 13:39
If I need to prescribe or define rules, or if I need to set certain conditions or I need to set certain limitations, it’s not really AI. It’s not really autonomous. It’s not really making decisions.

Victor 13:55
It follows the guidance that I defined. So I’m the human, I’m driving the logic, and the software automates. There’s nothing bad about this, but it is not AI. AI is where the system acts autonomously.

Victor 14:11
It can process the information, it can learn, and it can act. That’s number one. Number two is: Do you see clearly there isn’t reinforced learning? Or, in a more simple way, does it learn from its own mistakes?

Victor 14:28
We, as humans, learn from our mistakes. That’s the blessing we got. But one way or another, the software doesn’t always do that. If you prescribe the sequence, I’ll give an example from advertising; if I see that my advertising sales increase and my accounts increase, I want to drive the bids down.

Victor 14:53
I put this logic into the tool. There is nothing here. AI prescribes the logic; I’m the one driving it. There’s nothing bad about this. That’s a rule. The rule has its own presence and own place in life.

Victor 15:08
But it’s not AI. That’s what I mean by reinforced learning. If I say, “If I raise the price and someone acts, I need to program the system, in that case, lower the price, increase the price, keep it the same”, that’s not AI.

Victor 15:31
It’s logic or a rule. That’s number two: the ability to learn by itself. The third one, and probably the trickiest of all, and this is where it really helps you understand that this AI really stands out versus others that claim to be AI, is that it can adapt behavior to the various objectives.

Victor 15:56
This is much harder to do than all sorts of rule-based software. Think about this: if I give you an example of an objective, let’s say, increase sales and keep the margin.

Victor 16:14
Another objective can be increase the margin and keep the same level of sales. There are two different objectives the AI needs to adapt through learning. But if you’re using a system that you basically coded, that you decided there are certain rules and requirements, or there is a certain one-way black box that created rules but is being sold as an AI, then this system cannot adapt to these various objectives.

Victor 16:40
Adapting to objectives requires reliance on reinforced learning, and because my objective is different, I need to learn in a different way.

Victor 16:54
It also requires attention or autonomous actions that, again, rely on learning and take objectives into account. In the end, it needs to replace humans. I spoke with somebody today. We spoke about if AI is good or bad for society.

Victor 17:14
AI is not bad for society, and the fact that it replaces jobs does not mean that someone needs to learn new things so we can move forward and be more productive. That’s the whole meaning of AI, but AI must replace some of the work you’re doing manually, because if you don’t replace it, you’re wasting your time.

Victor 17:36
If you introduced AI, and you have exactly the same headcount with exactly the same people doing exactly the same tasks, you just wasted money, plainly. There’s absolutely no impact of the AI on your business.

Victor 17:50
AI must replace some of your activity. You may say you have five employees. Now you introduce the AI and you keep those five people, but they’re doing different things and your sales skyrocket. That’s great. But if you have the same five people doing exactly the same thing — you simply wasted money on the tool. With that said, let’s look into the use cases.

Victor 18:14
When I selected these use cases, I selected it from my own experience. Every use case is what we try to solve and the way we have solved it. Then, I will try to come from your perspective. I’ll put myself in your seat and say: is it true AI and why is it this way?

Victor 18:36
Maybe it’s not. So let’s look into a few examples. The first example is this: I’m a seller. I have various products. I manage a portfolio of products. They have different demand patterns.

Victor 18:57
What that basically means is, in some products, people are sensitive to price. With other products, people are not sensitive to price. Sometimes, there is a competition in place, so they’re sensitive to price up to a certain point. There are millions of millions of different scenarios that can happen.

Victor 19:19
But then I, as a seller, have one simple question: what’s my optimal price? Let’s see how we solve it and let’s ask Feedvisor if it’s true AI or not. First of all , we can solve this through dynamic pricing. We use Ai to drive these dynamic pricing decisions. Let’s look into this.

Victor 19:44
First question: is it autonomous? Yes, you don’t need to go and say “you need to change the price by this, see the result, okay I need to do this, then I need to do that. ”

Victor 19:56
It’s autonomous. It’s acts on its own. Excellent. Is it learning from number one? The question you always ask is if it is learning. If it doesn’t learn, it doesn’t work.

Victor 20:13
We build the demand models. What demand models mean is what kind of demand you have given to the price point. So, if my price is ten dollars, there is, let’s say, one hundred orders. If my price is $20, I Get 20 orders.

Victor 20:30
Clearly, there is a decline. It can be another story. It can be $20, and guess what? You get exactly the same amount of orders — people just don’t care. That’s what we call the demand model. We built this demand model for every given product. Now, if you have one price point — if you don’t change the price — you cannot build a demand model.

Victor 20:53
It will be guessing at best. You need to move the price. How do you move the price? That’s where the AI comes into place. It moves the price. It sees the impact. It moves the price down. This is the impact and this is where it learns.

Victor 21:08
It reinforces itself. It says “okay I lower the price. I saw that much of an impact. I raised the price and I saw that much of the impact. Let’s run it over a period of time and see what the actual output is.”

Victor 21:20
We need to take into account the environment, which is very much not static. Everything is changing, right? There are competitors who may change their price. Maybe I’m running advertising.

Victor 21:34
Maybe competitors run advertising. Maybe there’s a change in seasonality. Maybe my discoverability has changed. There are millions of things that can change the influence of the model because, obviously, if I’m running the advertising, the sales skyrocket. If not, it’s because I changed the price or because I put a lot of advertising into it.

Victor 21:56
You need to take this into account. Then, you build this dynamic model. The model learns and knows to respond to these various conditions. It knows how to respond because machine learning basically runs and understands the sensitivity, understands if my price stays static, but my orders jumped — that’s the reason it correlated to. All this comes under reinforced learning. Being able to adapt to objective is the most important thing. We adapt to three objectives: Increase margin, maintain sales, increase sales, maintain margin or maximize sales regardless of margin. Same data, but the same algorithm will act differently depending on the objective you have so we can help you achieve what you want to achieve. That’s the last thing. We eliminate the need for a person to manage pricing.

Victor 22:59
That person can do something else. But you as the owner or as a VP of a group, you can just set the requirements, what you want to achieve, you can set your objectives, you can set your ranges, and then the model executes.

Victor 23:13
Let’s visualize this. Here are the three examples I talked about. The first example is when you think of demand, think about the quantity of unit sold. So you see that clearly as the price increases, unit sold go down — the most common scenario I would say.

Victor 23:34
If your objective here is to increase sales on the expense of margin, then move the price to the left, to the lower, and then if you want to maximize margin, move the price to the right. Sounds simple, right?

Victor 23:50
It’s not so simple as you move to a different case. In a different case, and this I would say is the most common bias to almost every seller on Amazon, is what happens if your demand is not very sensitive to the price?

Victor 24:07
This can happen if there is a lot of demand for a product or there is not substantial competition out there or there is certain seasonality in place or Ssmply you price yourself too low in your mind, so you think it’s the right price, but you price yourself too low. In this case, what will happen is you increase the price and you’re getting exactly the same amount of sales again. The AI here knows to experiment and raise the price and see “Oh there was no change. Let’s raise a little bit more. There’s no change. Let’s raise again — oh now the demand drops. Let’s come back and let’s find that point where your margin is maximized.”

Victor 24:41
That’s an example of of the second situation. The third situation is a mix of both.

Victor 24:56
I would say this is exceptionally common. You see very inelastic demand. Yu can change the price, you can raise the price, and demand stays the same. At a certain point, it drops, but drops very quickly. You need to find that optimal point where your margin is maximized, but demand didn’t drop. That’s an example of using AI for pricing where you have your own private label or branded product. You don’t face direct competition on it but you want to leverage the AI to learn to understand the market and make the right decisions and you don’t need to get humans involved in it.

Victor 25:39
Let’s take another case not in private label or branded product. It’s actually a competitive product. There are a lot of sellers selling this competitive product and I need to be able to win the buy box, but I also don’t want to give up all the margin, I don’t want to get to the price wars. That’s an example of what feedvisor offers: the the buy box repricing. The buy box repricing, let’s ask the same question: Is it AI? Number one: it does real -time price changes completely autonomously. You don’t tell how exactly to change the price if the competitor is higher or the competitors lower, I can rely on feedvisor and feedvisor will make the right choice. That’s that’s the first attribute of the AI. The second one is that there reinforced learning. The system is learning or basically we prescribe, so we don’t prescribe, the system learns. What it learns is how competitors react to the price changes, which is very important. For example, you raise the price, how will competitors act? If you lower the price, how will competitors act?

Victor 26:52
The next thing, never less important, is the speed. So if I am the algorithm, I see that everyone responds slower. I can raise the price. I can win a lot of buy box in between while they are still responding.

Victor 27:15
By the time they responded again, I can respond very fast, lower the price, regain the buy box, and then raise it again. Being able to learn the speed of response, or the pattern of response, enables the algorithm to basically trick the competition.

Victor 27:29
We saw cases where people were using sort of a rule -based system, or they were using some sort of primitive algorithms, or even repriced manually, and they just couldn’t believe the results because this is where the autonomous pricing comes in.

Victor 27:47
By understanding speed, by understanding patterns, it can just outsmart the competition. So that’s where the learning comes in. Again, it follows the very same business objectives.

Victor 27:59
You can increase margin, maintain sales. You can increase sales, maintain margin. Or you can maximize your sales regardless of margin. You follow all of these, and the algorithm will behave entirely differently in each of these scenarios.

Victor 28:16
Same algorithm, same market, just different behavior. Finally, it eliminates the need for a person to manually reprice or to manage multiple pricing rules because the system does it for them.

Victor 28:30
Now I’m going to show something we have never showed in any conference. This is the visualization. We talked about the interpretability of the AI.

Victor 28:51
This is the visualization of the actual competitive environment for a buy box product for an average buy box product taking in the market. What you see here is a number of competitors. You see the the timeline, you see the price moves. Some of you will be surprised.

Victor 29:04
There are so many price moves because when you look at this, you don’t necessarily see that we receive price moves through notifications. Amazon sends notifications. We capture each notification. We put it in and then you see the competitors.

Victor 29:19
The red line is Feedvisor customers. The other lines are other customers. As you can see, we initially set a certain price.

Victor 29:36
It was $33, we won the buy box, everything is great. We’re trying to see what happens if we raise the price further. This is the actual action that happened during these days. So, we raised the price, we lost the buy box.

Victor 29:52
Why do we raise the price? We raise the price because we’re trying to invite others to reciprocate. Maybe we can share the buy box at much higher price and then you will get more margin. So they don’t.

Victor 30:06
They didn’t reciprocate, in many cases they do, but in this case they don’t. So you see the green line, the competitor below keeps price the same, the others keeps price the same, we lost the buy box, we learned, we’re coming back, we test again.

Victor 30:19
We say “maybe these guys are just slow, maybe they just don’t know to respond that fast. Let’s raise again, wait a bit, so now we’ve been waiting a substantial amount of time above, they don’t respond. We have to win the buy box, we’re going all the way down, we’re going to $30.”

Victor 30:36
We win the buy box, we start to raise, see how they respond in intervals, until we get again to a certain point, we lose the buy box, but we lost the buy box because a new competitor came in.

Victor 30:50
It’s hard to see, but a new competitor came in and that competitor took the buy box at a very low price. There’s no choice, we’re going all the way down. We fight this competitor, but guess what, this competitor sold out in an exceptionally short period of time, maybe like an hour.

Victor 31:04
We’re pushing price all the way up. There are sales happening with higher margin in the middle, and then we get another competitor, we compete with them in the middle — this is around July 20th. Guess what, it comes down, they don’t care, this is their price. We have to go all the way down, but then he sold out.

Victor 31:31
We start to raise up again and win the buy box. When you look at the average, they average like $33. If we’ll keep the $33, the results will be substantially lower. But, if we get the average, when you look at the performance, you get the average amount of sales for the $33 as if you kept it the same.

Victor 31:51
But you cannot keep it the same. If you keep it the same, you’re going to get 50% of the buy box of the best, assuming everyone is at the same price, and then you will lose it to all the actors below. If we don’t raise the price, it will deteriorate the margin.

Victor 32:07
That’s a very complex game in a very complex arena. This is how the true AI behaves. You see, nothing here is prescribed, not in increases, not in decreases, not to do what not to do when it’s learning. Nothing like a person, it can make mistakes. But in the end, it outperforms the person, and it outperforms rules obviously as well. So let’s move to the next case, and this is the case of advertising. Now, this case is hard to visualize.

Victor 32:42
I’m just going to talk it through, but at the end, in advertising, you want to achieve your best performance. So, what autonomous actions do you take? Your autonomous actions are keyword harvesting: being able to find the right keywords. Your autonomous action is dynamic bidding: to understand the bidding environment and bid properly and budget balancing and budget allocation. All three are then again captured by the platform. What do we learn? We use generative AI to understand which keywords are relevant and we give recommendations so we can supply keywords that that are relevant. This is where the learning happens and we understand which keywords are right, as well as the search volume. We’re trying to find the most appropriate keywords. We also look at the daily, weekly, seasonality and we’re trying to understand what kind of bid is needed and how various bid prices influence the performance. The last thing we look at is the budget utilization cross day and over time to set the proper way of budget allocation. We have two different objectives. You can focus on efficiency and say I want to achieve certain rows or echoes or we want to focus on what we call the usable visibility to achieve the highst brand awareness to help you with the new product launches. Again, the same algorithm behaves differently given the objective. As a result, we replace humans. We don’t eliminate the person needed for the advertising, but that person doesn’t need to take these mundane tasks that they will do if they have to do it manually. The last use case is probably the most interesting. You have pricing optimization. We spoke about this optimization for buy box.

Victor 34:53
You have a pricing optimization for private label branded product when the competition is not on the product itself, and then you advertise at the same time. How are you making sure it works right together?

Victor 35:06
No one, except Feedvisor, does this. We’re the only ones who are capable of combining the sections because we understand our AI — it talks with each other. This is why this is possible.

Victor 35:25
Let’s look into this. The first one we do is dynamic price changes. We know to understand the price elasticities, we know to understand how pricing influences demand, but we also know to take into account the advertising spend.

Victor 35:40
On the other hand, we know to take the advertising optimization actions, we know to set the bids, we know how to set the right targets. When these two happen hand -in -hand, how do we really coordinate between them?

Victor 35:56
First of all, it happens through treinforced learning. What it’s learning here is the elasticity of demand that we spoke on the pricing side. That elasticity of demand helps us to align better with the advertising spend.

Victor 36:11
It can choose different objectives. It adapts to the various business objectives. You can say, “okay, I want to do my sales objective, I want to increase sales, or I want to go more for the efficiency, and I want higher conversion rates, I want better echoes.”

Victor 36:34
The AI helps you save, again, the human’s time of coordinating the two. An example would be if you’re running a campaign. You pour money into a campaign.

Victor 36:54
You put money into the campaign, and now what do you do with the product price? You raise the price, you lower the price, any ideas? It depends on the context. If you’re running a campaign, and you detect the price elasticity, that is the price. It is pretty much elastic. It would make sense for you to actually lower the price, achieve much higher volume, improve the conversion rates, and achieve much better echoes.

Victor 37:27
Because if you have more sales, your echoes substantially improves. Because the very same spend now is against many more sales. That’s the first scenario, the elasticity.

Victor 37:42
The second scenario is that there is very little elasticity. Lowering the price is not going to help, but you can raise the price. If you raise the price, you actually can go and offset some of your advertising expense because you spent on the advertising of that product. If you can sell the very same amount, the conversion rate is not going to suffer. If you’re going to sell at a slightly higher price, what’s going to happen?

Victor 38:15
You’re going to offset some of the cost that you put toward the advertising through your increased margin. That definitely helps to offset costs and and achieve better results. If, at a certain point, that leads to elasticity actually increasing and it leads to the fact that your price will now drop the demand.

Victor 38:40
You should stop and get back to the point where demand is optimal because otherwise, your echoes will be through the roof. Because you increased your price, your orders and sales went down, and now you have fewer sales against advertising.

Victor 38:57
That’s an example of where you really need to watch the two together, and this is where they’re combining everything. So Feedvisor offers a product called Feedvisor 360, where we do 360 optimization. That’s exactly where we take all of this into account, and we help you visualize this through analytical means.

Victor 39:15
You can see exactly what I’ve done in advertising for each product, what I’ve done in the pricing for each product, how it influenced the performance, what actions did I take, and how I’m going back and adjusting my objectives accordingly.

Victor 39:29
To sum it up, we spoke about true AI. So the true AI, there’s four questions. Nothing bad with not true AI, nothing bad with the quasi AI.

Victor 39:44
You can still use it. It’s not an AI, but it’s okay. The four questions: is it autonomous? Does it do self -learning? Does it close the loop and do reinforced learning? Is it adapting to objectives?

Victor 40:00
It’s very important, and this is replacing the human. How is my organization going to change as a result of the AI? If it doesn’t, that’s a problem. We do it to improve the performance.

Victor 40:14
If you do AI and your performance stays the same, something is bad here. The first thing you should see is a boost in revenue.

Victor 40:36
The second thing you should see is that if you’re using advertising optimization together with pricing, it strengthens your brand awareness and helps you achieve better visibility. It’s combined with good performance and also delivers very strong RoAS and TACoS.

Victor 40:56
This is what feedvisor is all about and how AI is can be used. It’s not the only way to be used. There’s multiple ways of using AI, but there’s one way of using AI to really achieve outstanding performance. With that said, before the Q &, one of our clients said he would stand up and say a few words about Feedvisor.

Lucky21 41:47
We’ve been using Feedvisor in my company, Lucky21, since 2017, initially for the repricing.

Lucky21 42:01
And later, as they developed their advertising, I could not give a higher endorsement for the software.

Lucky21 42:17
Everything that Victor spoke about here is 100% backed up by his software. It’s miles ahead of any other tool we used. I know we’re focused here on Amazon, but my company also operates on Walmart. When we initially started on Walmart, they had locked us into two of their API partners only.

Lucky21 42:40
It was very, very rough going. I won’t mention who they were. But as soon as they opened it up to other API partners and Feedvisor was on it, they did what they do best. They ran with it. I think that was about early January.

Lucky21 42:59
And here we are now in the end of July, and we’re fully functional with all our campaigns on Walmart. The proof is in the numbers when it comes to the Amazon campaigns. Many times we’ve tried outsourcing to agencies for particular tests.

Lucky21 43:23
Everybody seems to have a methodology that makes sense on paper, but nobody could beat the campaigns that Feedvisor is optimizing. That’s my little pitch for him.