14 Jan 2024
Mastering Amazon Ads: Integrating Profit Analytics, SEO, Inventory, and Market Insights with Tarik Berrada Hmima from m19
Join Tarik Berrada Hmima, CEO and CO-founder of m19, as he delves into the intricacies of Amazon PPC Automation. ⚒ 🛫
Explore the blend of strategy, data, and the future of e-commerce advertising. Gain invaluable insights and tips for optimizing your Amazon campaigns. 📈
👉 Also, don’t miss a chance to explore more about m19!
Today’s Navigators: Tarik Berrada Hmima & Oleg Zaidiner 🧭
(The soundtrack belongs to BoDleasons / Pixabay)
Team members:
Episode Transcript
Oleg: Hi everyone, welcome to the Navigator podcast. Today we have something special for you. We have an amazing guest, Tarik, and we have a little bit different opinions on PPC Automation, and I believe it can be a very interesting discussion. So welcome, Tarik, and can you please give us some background on how you ended up in the Amazon space and your journey?
Tarik: Okay, so thanks, Al, for having me on your podcast. I am Tarik, one of the co-founders of M19 and CEO of M19. Just a quick note, M19 is an automation platform mainly for Amazon advertising and is providing other tools next to it in order to make broad and the smartest Amazon advertising using all Amazon data like profit analytics, inventory, SEO, etc. My story with M19 started about four years ago. I used to work in a French adtech company, and I was in R&D, research and development, in the machine learning department, working on optimization algorithms for mainly Google ads and retargeting. About four or five years ago, the idea started. I was discussing with another friend engineer from the same department, machine learning, and we said, "Okay, let's do something. We know how to do optimization on Google. Why don't we do that?" We started researching. It was the beginning of Amazon advertising APIs, which is the ability to connect directly with Amazon in an automated way. That's how it started. There was an opportunity, Amazon advertising was fresh new with their APIs and enabling other software to automate. So, we started like four engineers, friends, said, "Let's build the technology."
Oleg: Nice, and how big is your team now?
Tarik: So the team is 15. We grew pretty quickly. We were pretty lucky. In the beginning, our plan was just to build the strongest technology possible, just the backend, and sell it directly to agencies or integrate it with big vendors. We saw quickly that it was a hard sale, and the market was not mature to integrate directly with the technology. So after a couple of months trying to sell it, plus engineers trying to sell to do enterprise sale, I can let you imagine that it's not a good match. So we said, "Okay, there's a big sellers market." We got some beta testers, they saw great results, they were willing to pay from day one high tickets. So we said, "Let's build the interface, start to do the SaaS model, and build the platform and give them all the information they needed." This is how we grew, mainly at the beginning with sellers, and after we had big traction with agencies that found it pretty useful to gain time in a super automated and controlled way. They were happy to streamline their operation, to hire quickly, and onboard people without being too technical about operations. They wanted to have people only concentrate on the strategy to put in place, not to focus on bid by bid or keyword by keyword. This is how we got the traction, and now our main focus for the last year and a half is to serve agencies because this is where we are bringing the biggest value, like agencies and big brands.
Oleg: Yeah, this is definitely where the market goes, right? The most sellers, they cannot manage it themselves. It's getting more and more complicated, more data points.
Tarik: Yeah, so it's a market getting to the level. Nice. So, and by the way, just a small interesting fact for you, we are getting more and more big sellers reaching to us saying, "Hey guys, we know that you don't do managed service. What agency would you advise?" Because we want someone to advise us on the strategy from the beginning. We are saying, "Hey, we are a tool. We are an Automation and optimization tool, but the business strategy, it's either you as a brand owner or you have an agency that is helping you do that." An automation software, I know that some people don't agree with me, is not there to replace a human to come up with the business strategy, the ads strategy, but it's there to streamline the operations in an efficient way, not in a non-efficient way where they see something they're not happy with, they jump in the advertising console, change it, blah blah blah. So, the position of our software is pretty tricky.
Oleg: Yeah, we definitely get exactly the same position. So, you need a brand, you need a strategy, and you need the execution. In our case, with tried many different AI solutions, and Tarik said he tried to convince me that it's doable. So, I want to start with some things that we tried before, and it didn't work. We even had, like, last month, we had an experiment with one another company in the PPC automation space, and so what we did, we took our campaigns that we manage manually because it's a very important client, it's like huge volume and very competitive. So, we spend a lot of money, and we want to be organically and on top. So, we split about 50 campaigns, like 25-25 more or less in number, and half we gave to the algorithm, and half we keep doing ourselves with a clear goal. We want to keep the ACoS, get better, okay? Right, we want to increase ACoS from 2.3 to 2.5. So, the algorithm, after one month, got it from 2.3 to 2.2, and we manually got it to 2.5, keeping the same volume. Right, so for me, it doesn't work, right? And again, we get this like, "Yeah, we need more time to learn," and I have knowledge in machine learning, right? So, what...
Tarik: So, okay, so first, first thing, I don't like, even if you might see it on our website, but I'm not doing all the marketing stuff. I want to, and we can discuss this at the end, let's not use AI because it's not real AI. We are doing some very basic machine learning. Anyways, with the data provided by Amazon, you cannot do crazy advanced machine learning. If someone is claiming that they are doing, I don't know, deep learning or generative models or whatever, to be honest, I'm super appreciative of your humbleness and honesty because, you know, I might be the only one that really says it. Many on say, "Yeah, we use super smart AI model," and I'm very skeptical. They don't show, no, this is...
Oleg: No, no, no, this is... However, the tech team and the R&D team spent a lot of time in R&D working on machine learning topics related to advertising. And to be completely transparent with you, like all the advanced recent research and algorithms were, in most cases, having worse results than super basic stuff like logistic regression that you might know, Al, and some basic statistical prediction models. They, in most cases, they are the same, and sometimes even better than most advanced algorithms. Why? Because the sophisticated algorithms, they work in other areas like speech recognition, image recognition, content generation. This, yes, you cannot do basic machine learning, which is not the case in Amazon. I'm talking about PPC, which means finding the right keyword and finding the right price to pay for the keyword in order to optimize either RoAS, margin, market share, or whatever. So, for this specific use case, it's simple stats that is used.
Second, and you know it, as with a computer science background, having a proper framework to test two systems is very tricky. Because to properly test two systems, you need to have completely isolated systems with no influences between the two systems. And these two frameworks, or contexts you are comparing, they need to be the same, like the same overall, the same. You need to have lab context.
Tarik: Yeah, so it's like clear.
Want to know how this episode concluded, dive in into our podcast channel on Spotify!
our activity
News and podcasts
LATEST UPDATES




