Ecommerce Advertising Made Easy With AI | #237 Mati Ram

Please support the show if you liked today's episode:

• If you love the podcast, please get someone else to listen, too!
• If you enjoyed this episode of the Ecommerce Coffee Break podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.
• Share the podcast with your family, friends, and co-workers.
• Tag the podcast on Instagram @clauslauter and let me know what you like about it.
• If you like the content and would like to support the podcast, you can buy me a coffee here.
• Become a guest on the show or sponsor an episode.

Claus Lauter: Hello, and welcome to another episode of the Ecommerce Coffee Break podcast. Ecommerce advertising is a difficult topic for a lot of merchants. And when AI came out or basically launched big last year, I was thinking about, okay, that's finally helping us with advertising and making things easier. And we want to dive into this topic a little bit more today with me on the show to discuss this.

I have Mati Ram. He is the CEO of AdScale, a leading ad tech company e commerce advertising with AA driven solutions. A tech entrepreneur with over 20 years of experience. He's also passionate about bringing innovative solutions to market to solve real world challenges. Before AdScale, Mattie was the founder and CEO of Dynasec Limited and continued to run the division after it was acquired.

Why the checkpoint software technologies. So I'm a really expert when it comes to advertising in AI, and I would like to welcome him to the show. Hi, how are you today?

Mati Ram: Hi, Klaus. Nice to meet you again. Pleasure to be here.

Claus Lauter: Great to have you on the show. Let's dive right into it. E commerce, advertising, all these different platforms. We see that third party cookies are dying. There's a big shift in there. Things are not getting easier. Tell me a little bit on what's happening in the market right now.

Mati Ram: I think it's all started in March 2021. That was the exact month the advertising ad tech world, if you want to call it, changed forever. what happened on that month is that 14 was launched. Apple kicked out Facebook from getting data from the apps, unless people give consent.

And I think that it was the first time where we've witnessed the death of the third party cookie. maybe we need to spend some time to explain about the third party cookie. So, basically, ever since the internet was invented, the third party cookie served... As the main vehicle to track user behavior and from the early days of Internet marketing, Google and later on also Facebook heavily used the third party cookie to offer personalized ads.

, they could have known anything about any click, any page we visited, any product we've added to the cart, everything we've purchased like. Everything was very clear and very open. That made advertising very personal, more accurate. And if you look, holistically, I think that's one of the main drivers why budgets shifted from TV and print and radio to the internet.

So Google and Facebook became what they are today, and it's all around the ability to track user behavior. Now, following the privacy regulations like the G D P R and the C C P A and some others, The usage of the third party pixel started be limited. So it started with Safari and who banned the third party pixel usage.

The issue with Apple and Facebook had to narrow the attribution window to seven days. Apple declared that it's gonna ban the use of third party pixel in 2022. Now it moved to 2023, probably will move to 2024, but eventually it'll happen. And they also declare two years ago that there are gonna ban the data also from the Google Play, which leaves us in a totally different world.

It means that Google and Facebook are collecting less data, means that they are less accurate. And that is the main reason why everybody sees drop in ad performance ever since March 2021. And it's going to be even worse once Google will ban it from Google Chrome, which is 70 percent of the browsers.

So I think that's a huge issue in the industry that everybody heard about it. People know about it, but... I'm not sure that everybody grasps how dramatic it is when it comes to ads performance. Now, if you look at that clause, you can see that Google and Facebook are becoming more and more black boxes.

Google came with the PMAX campaigns. Which is basically, give us your budgets and your assets and we will advertise for you. Facebook came with the Advantage Plus, which is basically the same ID. And the idea is, use our own Google and Facebook first party data. We will get you the audience that you want and the people that you want.

And we will get you the results. And I fully understand that because have less data. That's a long answer for a short question, so cut me from too long, but, what we need to discuss is first party data and how first party data can help advertisers to be more efficient with advertisers.

So what is first party data? First party data is the data that we collect from our customers orders. It includes the order information, the customer information, and the product information. And the nice thing about first party data is that, first of all, it is our data. Nobody else has it. It's not something that either Google or Facebook have because they are not connected to the store.

So they don't know what is the average order value and what is the customer lifetime value and what is the repurchase frequency and what is the repeat customer rate. Which products are frequently bought together and all the huge information that you can extract from a store is not something that Google and Facebook have because they are simply not connected to the store.

And more than that, the pixel gives you broken data and that's why everybody is so busy with trying to implement analytics and getting the data and there is... Duplication of conversion and trying to solve the data. That's an issue, but the first party data is a very accurate data. It is my data. I know exactly who are the people.

What do they buy? When do they buy? Which products? Everything. So it's a very accurate data. It's much more extensive than the pixel data. Think about it. Pixel data is what, 30 days long?

The first party data forever. We can use the entire history. So it's much more extensive. Having it much more extensive means that statistically, it's much more significant.

There is more data with more significance. So if we can take that data and we can make Google and Facebook more accurate, we might be able to get more results in advertising. And in order to turn the data into that, I think we need the AI part.

Claus Lauter: I like that. You basically summarized what every merchant went through in the last two years and everyone's can see declining results, increasing ad costs, having the problem that the beloved audiences are not working anymore or lookalike audiences are not working anymore. So I think everyone had a very good feeling on.

What was good in a good golden times when everything of that was working and marketing was easy and now things are becoming more difficult. Now, obviously collecting your own data, with third party data is a great thing. Tell me a little bit about the implementation, so how to do that. And then the step going from your own data back into the advertising platforms, because obviously then you need to connect that somewhere to make it work.

How does that work? 

Mati Ram: So it actually works in several steps of the process. I think that what connects the BI, which is, look at the first party data as two layers by itself. There is the raw data, which we collect from the store, and then there is what I call the BI data, which is the, insights that we learn from the data.

Because there is a lot of data, and there is a lot of noise, we need to summarize the noise into actionable things that we can use. I think that's the first layer. Now, what connects the BI data into the AI advertising are mainly the segments. There are customer segments and product segments. Now, what are customer segments?

Customer segments, basically, everybody knows that, are a group of people with similar buying behavior, buying patterns. Which we group together in order to offer them the relevant offer in the relevant time on the relevant channel. If we can inject this customers, if the AI can help us find the correlation and say, okay, this is a segment of big ticket spenders, let's offer them big ticket products.

This is a segment of, let me give you some very interesting segment that I see quite often. Segment where the AI finds correlation for people that are more likely to purchase soon. Based on their repurchase behavior, based on the products that they are buying, based on other people that are buying similar products.

So the AI takes that into consideration, all these features. And comes back and say, well, here's a list of, I don't know, 3, 000 people. They are more likely to create a purchase soon. once you have that data, you have a competitive advantage. You can target these people, and people like them, we will discuss it in a second, before your competitors will do.

Which means that... You have more chances to get them as customers than others. So the segments are connecting the BI data, BI first party data, to the, what I call AI advertising. And maybe later on we can discuss what is AI advertising because It's a puzzle that I throw right now, but I didn't explain what does it mean.

 So segments are definitely the answer, and maybe to elaborate a bit more about that, it's important that the segments will be, dynamic segments. What do I mean by dynamic segments? We want to create segments that people get in and out of the segment based on their behavior, and based on the entire people's behavior.

So, a segment of lost customers would be people that... The last purchase was more than two standard deviations than the average, for instance, okay? So that's something which is dynamic. It goes with the store. It lives with the store and it means that everyday people are getting in and out of the segments automatically.

But you offer each and every segment what's good for him in the right time on the right channel and that makes the advertising much more accurate. All with data that Google and Facebook don't have.

Claus Lauter: think a very important point that you mentioned before is that Google and also Facebook, introduced dynamic ads there where you throw your money into, and you mentioned that very nicely into a black box. So it's like, okay, you do it and you don't have any control. I think at scale is a little bit different because you have the control, what's coming and Coming to the AI, which I'm very interested in and how you do it, there's so many levels where I can think AI can help you with doing your running your ads, give me a bit of an overview of how it works, what kind of features you have, and what's the best scenario to for immersion to get started.

Mati Ram: So the AI advertising is actually divided into three stages in the process. Stage number one is the data training. We somewhat discussed it before where we connect, , AtScale to the store. We learn all the first party data and we inject the insights. The insights are the segments, but also some more, , advertising related insights that can be very helpful for us.

For instance, Interest, gender, and age as targeting doesn't work anymore, right? You don't get the good results. But if you can take the segment and you can look at like a segment and then narrow it with interest, age, gender, you might get better results because you make Google and Facebook more accurate.

So you need to prepare that data. That's in the first stage. And you also need to look at business KPIs like AOV because sometimes, there are businesses that they will never get to the relevant to the expected return on ad spend only because the AOV is too low. So you need to find ways to increase the AOV, otherwise the correlation between cost of click and conversion rate and the cost of the product will never work.

So we need to see how you increase the AOV. Sometimes you need to shorten the repurchase frequency. So you also get some KPIs that are business KPIs that you say, Okay, in order to improve that business, we need to do 1, 2, 3. And then how do we achieve that with advertising? So that's like the first stage analyzing the business and understanding which things, the advertising can help with in order to achieve the goals and what will move the needle.

 That's the main point, where to shoot, what will move the needle. Stage number two is the campaign creation and within the campaign creation. You have also two types of AI. You have the mathematical AI, which is getting everything that we've gathered in stage number one and inject it automatically into the campaign creation stage.

So not just give the data and say, okay, success. But when you create campaigns, say to the people, well, these are the products you want to target on these channels. Here are the audiences that you want to target. You need to work on new customers, like acquisition, remarketing, and retention. Here are the segments for that.

Here are the segments for that. Here are the segments for that. here are the, geographical places, and here are the keywords, and let's exclude these things. So all the things that you usually do manually are now being done with AI that analyzes the data and helps you to improve the conversion rate.

So that's. One part of the AI in the campaign creation stage, which is a mathematical AI. But then, and that's one of the nice things in the past, systems like AtScale had suffered from the fact that you also need some soft skills. Copywriting, creative. Now we have generative AI, so it makes things much more easy.

So for instance with Edscale, we have added ad copy with ChatGPT. Not only ChatGPT, ChatGPT plus a version that makes sure that what you get Is relevant for the different, lengths that you're allowed to use and meets Google and Facebook criteria, so you'll not be blocked, which is a big problem by itself.

So, ad copy can be done with the system. , ad creative, video ads, image ads, it's all something that the generative AI can do. Part number two is generative AI or AI campaign creation that includes generative AI plus mathematical, insights that are injected into the campaign creation stage. And then comes the most important one, which is the optimization stage.

 In the beginning of all days of Google and Facebook optimization was like a big thing, right? We all optimized the budgets and the beats and play with that. And then it was too big and too complicated.

And Google and Facebook came with their own, optimization. Most of the people use that. I agree with that to some extent, but. The idea is not only optimize Google and Facebook, is to optimize the entire portfolio. So how do you optimize your advertising?

There are three ways to optimize. It's budget optimization, bid optimization, and campaign structure. Download this stop targeting this interest, add new ad, all kind of these things. A. I. Can substantially help. Let me give you an example of how, for instance, we do budget optimization because I think that's very interesting.

Let's just understand usually is to maximize the revenue in the given budget, which means that the advertiser gives at scale the budget, call it X and says, give me the maximum revenue that you can. Now, we are in a world that, there's a wall garden between Google and Facebook.

Google sees only Google. Facebook sees only Google who sees the client. We sit on top of them we use ai and I'll explain what is the AI about it, because AI could be a lot of things today, but, we use AI in order to optimize the campaign. So in order to optimize the campaign, what do we do? of all, The idea is to generate a distribution for every campaign in Google, because that's the budget holder and every ad set in Facebook, because that's the budget holder over there. And the distribution generates the regression line. In the distribution, We get a lot of data, we draw the regression line is the line that represents the minimum variance between all the optional lines, call it the average, probably. Once we have the regression line, we can now predict, because we know in a decent level of certainty, what would be the outcome of revenue for every given budget we give, based on the historical data. So if we can predict that, and now you, Klaus, as a customer, come to me and say, Mati, here I have 10, 000. pen it to me in the best way. I just throw the number into the input. You give me 10, 000 and you say I need a return on ad spend of 500%. Okay, that's the input for the model. Now I use your input and the model looks at all the possibilities.

It has all the, curves. So it knows what probably we'll get. from each and every campaign. Now there is a mathematical model known as the knapsack problem , that tries to find what is the optimum of all the optimums. So each and every campaign has its own optimum, but we want to optimize everything. And that's how we divide the budget. And then, the AI listens to the Google and Facebook auctions 24 hours a day, and sees if the prediction is right. If the prediction is right, all is good. But if the prediction is not right, and there is an anomaly, or a trend, then it automatically reacts. Google and Facebook are, 24 7 live auctions.

So imagine we discussed about Facebook. Let's discuss a bit about Google. Imagine that at eight o'clock in the evening, everybody's home watching football. The agencies are home, the customers are home, everybody's home. But the AI that is there 24 7 identifies that the correlation between the CPC, the cost per click, and the position in Google.

is much better now than it is usually in the same hour in previous days. It means that our competitors are out of budget. For today, which means that if we increase the budget, let's take some budget for Facebook and drop the bids, we can collect all the impressions for almost no money.

Now that's like an inter date opportunity that you can exploit with AI. So that was a very big and amazing mathematical explanation to how the AI can help you get better results by actually sitting on the auction and managing your campaigns 24 7.

Claus Lauter: I think it gives a good overview of how strong AI actually is because it helps you in so many different levels, the campaign set up, the artwork, the copy and so on. And then when you mentioned that it's very important to point that out for our listeners is the bit management because who has ever worked with Facebook knows they go in, you waste a lot of time in there because you're trying to manually find out what works best.

Yeah. And as you said, a very proficient ads manager goes home at some point and then whatever, and we all had it, Google or Facebook, the algorithm goes crazy and your ad spend goes through the roof with no results. Now you have the AI tracking 24, seven, what's happening and adjusting accordingly to what's happening on the ad platform.

So I like that a lot. And obviously that will give you much, much better results. Now, what is the kind of, , Homework that emergent needs to do before they can get started and what kind of timeline, how long does it take before the first data really kicks in and the, AI kicks in to give optimal results.

Mati Ram: The good news is that once the AI gets in for instance, we connect our system, we have all the history open for us. So if you have data, we can use all the historical data both in the store and in the ad platforms. If you have zero data, then the main advantage is the beginning will be the ease of use and the automated automatic collection of data.

But the learning curve will be longer because we need to learn the data. When you have a new store. If you already have a store and you're running it for a few months and you have sales and, things start evolving, or even if you're big and you have a lot of data, basically the more data you have, the shorter time you'll see results coming in.

Now, it's very hard to give a concrete answer only because... The variance is very high. But I think that using AI in advertising general, if I have to give a ballpark figure. You can see about 30 percent increase in revenue or decrease in cost per acquisition.

Depends what you want to within one year. Now along the year, it could be either more quickly or less quickly. It really depends on the store. It's also dependent in a time of year. Like Q4 is a very strong Q quarter for e commerce. So usually results are better. Q1 and Q3 are lower, so it depends where you start, when you start as well.

But I think that as a ballpark figure, 30 percent in one year is doable.

Claus Lauter: Okay, that's huge result there. How does the onboarding process look like for a shop of immersion?

Mati Ram: So the idea is to, first of all, connecting AdSkill with Shopify is just to download from the App Store. We are in the App Store, so it's one or two mouse clicks. And then there is the integration with Google and Facebook. So the system guides you through the process. You must be an admin in your assets in Google and Facebook so you can connect up.

So the system just guides you through the process. Click here, give us an access. It's about five or six clicks process. And then we are connected , both to the shop and then to Google and Facebook. And sometimes people are not admins, so they need to get the password, but if you have the rights, this process should take not more than two, three minutes.

If you don't have the rights, get the rights and then do it, but usually it's an easy process.

Claus Lauter: Okay. If you don't have the rights, find someone who's better than you.

Mati Ram: Exactly. or call the people who have the rights. never believe how many stories we hear of people with enormous amount of assets in their Facebook page and they don't have the admin rights and nobody knows where it is, so it happens.

Claus Lauter: What's the pricing structure? How do you charge for the service?

Mati Ram: Our positioning is to help people to, increase their revenue and be much cheaper than any other option in the past. You have two options, either to go for an agency, which is, has pluses and minuses, or to do it yourself, where you need a lot of time and knowledge. And we're coming back with a third option and we want to be very competitive.

So we are basing our pricing on the ad spend simply. It's a, so the minimum is 129 per month and it goes with the ad spend, but you pay us. a few percentage of your ad spend and In any case it will be much cheaper than in any other solution that you have In order to manage your ad so pricing is very competitive

Claus Lauter: Yeah, that speaks for itself. You're getting all the know how, you get the AI and you save a ton of time. So I don't think , It's really money well invested there before we come to the end of the coffee break today. Is there one final thought that you want to leave our listeners with?

Mati Ram: Yes, think that the most important thing is to understand that we are living in a different world and Whatever you decide to do with your advertising First party data comes first The only way today to increase the causes with just to increase the results is with first party data. There was a very interesting , study of Boston Consulting Group.

I think it was even in 2020, they've analyzed organizations that are using first party data for marketing and advertising compare them with others that didn't. And I think that the average difference was that the people that used first party data was. 2. 9 X better results than the ones that didn't.

And that was before the third party cookie, was starting to die. So my number one advice to everyone, first party data comes first and AI really helps.

Claus Lauter: Yeah, makes total sense. And I think there's not really an option to ignore first party data because it's the one way to go. Where can people find out more about you guys?

Mati Ram: On our website, www. adscale. com, or if you're a Shopify merchant, we are just, a quite popular app , in the app store.

Claus Lauter: Okay. I will put the links in the show notes. Then you're just one click away. Marty. Thanks so much to give us basically a masterclass on how AI helps with advertising. And I think that's the way going forward. I don't think there's any other way around and I hope a lot of people will check your system out.

Thanks so much for your time today.

Mati Ram: Thank you so much, gloves.