Marketers Get Superpowers, Creative Gets Smarter, Campaigns Get Faster
Why you should listen
Performance marketing has reached the boss level. Time to grab your AI supersuit.
Wangxin Li, VP of Decision Science and Global Ops at Aarki, is the powerhouse behind that suit. In this episode, she breaks down how Encore was built to help marketers take back control in a world of signal loss, SKAN delays, and fragmented funnels.
From real-time bid decisions to creative that adapts fast and hits harder, AI doesn’t just assist. It orchestrates every stage of the funnel: user acquisition, retargeting, and everything in between. Supervised AI (which is AI + human oversight) leads to smarter campaigns, less guesswork, and way fewer wasted budgets.
If you’re doing user acquisition and retargeting, but are tired of wrestling with SKAN constraints or flying blind post-install, this conversation is your cheat code.
Top 3 takeaways
- AI ≠ Autopilot. It’s an Iron Man suit.
Encore fuses deep neural networks with human oversight to orchestrate every touchpoint—creative, bid, placement, context, frequency. Marketers stay in control, but with superpowers. - Creative isn’t an asset. It’s a strategy.
Performance data feeds directly into both human teams and a custom AI model, slashing creative testing time from weeks to days. The result is faster iteration, smarter storytelling, and less budget waste. - Retargeting + UA = Full-Funnel Flywheel.
Encore’s roots in retargeting gave it an early edge in behavioral insights. Now, that data shortens training time for UA models and lifts campaign efficiency end-to-end. It’s not just about acquiring users, it’s about keeping them, profitably.
The transcript
Peggy: It’s pretty much a given that AI is out. The core of competitive advantage, but is all AI created equal? And what’s the difference? Is it the technology? Is it how marketers work with it? And what is the future of AI? What’s coming next? These are the kinds of questions that I personally get very excited about.
I imagine you do as well, and that is why I’m also excited to be talking to someone who understands how we’re gonna be collaborating with AI, what it’s going to be like, and ultimately, how powerful is AI really going to be behind the scenes? So I have Wangxin Li and you are, love this title, VP of Decision Science and Global Ops at Aarki.
So, it’s great to have you, and it’s your first podcast. So, in a way, I’m also honored to have you, as well.
Wangxin: Thank you. Peggy is, I’m glad to be here. Thank you for having me. And, as I said, I am, the VP of Decision Science and GL Global Operation at Aarki. I work closely with our product and machine learning team to bring our product to life.
My goal at Aarki is to make sure that we are delivering against client KPI goals and, make sure that we are bringing the best product to market
Peggy: Your expertise. You’ve been doing this for a decade. And that in this industry means that you’re one of the first, because we’re just getting our heads around AI. So there you are. You’re building it and you’re bringing it together because you’re blending quantitative analytics, market intelligence, bringing that all together in AI. And that’s also coming all together at Encore. So what is it exactly? What does it allow marketers to do differently? Because we are talking about AI for a while. This is a new development at Aarki.
Wangxin: Before we talk about Encore, let’s make one thing very clear. At Aarki we live and die by ROAS, and everything we do is to make sure that we are bringing more laws for our client. And help them to spend their budget most effectively. And to achieve that, Encore does a couple of things.
One is that we use very advanced live DNN model, which is the deep neural network. Which constantly learn and adapt with the live attributions from campaigns. And Encore is not just limited to predicting user behaviors with gear model. It also orchestrates all different other aspects which are crucial to campaign success.
Wangxin: For example, it determines what is the right creative format to use. Is it video or is it banner or is it interstitial? It also determines how many times we need to serve an ad and in what context we need to serve an ad. And lastly it also need to figure out how much do we want to pay for an ad?
What is the real price for it? And these are all discrete systems. Encore just orchestrates all of these system, make them work together. And also Encore is also supported by our own data center which processes 5 million KPS. We listen to the word inventory. That is also what makes Encore super, super unique.
Peggy: I like the idea of orchestrating because, you know, I’ve been speaking with a lot of marketers and what they’re saying is, you know, it’s not a question of what is in control? Is it AI or is it the marketer? It’s really how can AI enable marketers to do their job better? So that is about making right decisions, but not always having to make them by themselves.
It’s great to understand that to orchestrate all those capabilities, but they also have to, and I’ll stay with the pun, yeah, strike a chord with the user.
Wangxin: So let’s talk about creative. One thing this orchestration layer does is it feeds the real-time performance data to our creative team, and we are able to quickly look at the features of what creative is performing the best. And in the meantime, our creative optimization algorithm will continually serving more and more impressions to the best performing creative and then this information is also fed to our creative team who has access to our custom-built GPT model, which not only help us to generate new versions of creatives, but also it creates brand new, creative concepts. And what it does is it can help us to reduce this creative testing time from a couple of weeks to a few days.
Peggy: The real magic is the efficiencies. Right. So maybe you can give me an example of how this plays out in the real world.
Wangxin: The beauty of Encore is it’s a full-funnel platform and it not only acquires new users but it also retains existing users. Through our retargeting product. Yes. Because they’re layered together. Yes. And, Aarki initially started as a remarketing platform, so most our clients naturally started working with us for retargeting campaigns. And [with] our initial focus on retargeting, [it] provided our model deep insights into user behaviors early on. And this actually help us to reduce the model training time, model calibration time, and then it can help us to reduce the waste of spend during the early stage of user acquisition campaigns.
Peggy: At the core of this is always something cool, and here you have it, an algorithm XYAB. I mean, it just sounds like something that you even have to have clearance to talk about. It just sounds too cool. It’s like, it’s like Black Mirror meets marketing or something. But, I won’t ask you necessarily where that comes from, but what I am interested in is understanding exactly how does that work behind the scenes? I mean, not going very deep, but let’s face it, you are an architect of this, so you know how it’s built. Lift a little, little.
Wangxin: Yeah, that’s a great question. Well, SKAN levels the playground for our platforms. We came up with something unique with our XYAB algorithm. we enhanced the post backs we are receiving from Apple with multiple advanced data resources in a privacy friendly manner, which include the a call data and transient signals, which can help add more context to the post backs that we are receiving from Apple. And our algorithm also adjusts for the delayed conversion postbacks that we are receiving from Apple that is generally set up by Apple and advertisers to protect user privacy. And with all of these. additional signals and whatever Apple sent us, we are able to deliver exceptional results for our client.
Peggy: So that’s really important because that’s an enhancement that market has been talking about. Another thing they’ve been talking about is that creative is the heart of the campaign and AI is the motor a growth motor there? Yes. It is something difficult to orchestrate. How do you approach that?
Wangxin: So I think creative is often one of the most overlooked part of, marketing strategy. And I think that is a huge mistake on the marketer’s part. Creative should be front and center. And to achieve that, we have made significant investment into not only creative production, but also into our creative optimization algorithms. We also have a team of industry experts who not only produce creatives for our client, but they also create brand new creative concepts for our client, which is industry-leading.
And, we married this deep human expertise with this advanced AI model, which is constantly tagging, marrying, and refining the creatives. This powerful combination can make sure we can consistently deliver the most relevant and most engaging creative to our campaigns, which can help us drive the overall campaign performance.
Peggy: because that tagging is very, very important, because it’s one thing to win. It’s another thing to understand how did I win that? How did that work? Because that’s the only way you can repeat that success. So tagging is a big part of that. And the human, yeah, the human inspiration, the human role in the storyboarding, the AI tagging, bringing that together, it’s really important also to have those capabilities in-house, which you do as well.
So I wanna look forward into the future. We talked about the role of human and AI, how important that is. Again, the harmony, the orchestration going around in full circle here, zero touch campaign. What does that, what does that look like? What’s it going to be? What can be automated? What is it that human and AI can be doing? Yeah, and also scaling together.
Wangxin: If you think about ad tech and ad operation in general, this is a highly manual world with many hands stretch work actually with many hands on keyboard. Yes, this is an inefficient ecosystem. Our goal at Aarki is to make Encore to orchestrate all the different aspects of Aarki, all the way from media planning to creative generation and personalization. And to day-to-day continuous campaign optimization. We want Encore to be an AI agent, and we want to offer Encore to our clients as a marketing assistant in a fully integrated fashion so they can use it directly as one of their team members. And, that is a lofty goal, but we want Encore to be an Iron Man suit for retargeting and user acquisition marketers.
Peggy: That is such a cool metaphor, if you can even see it. But yeah, it, it’s about augmenting the marketer, you know, enabling, empowering. It’s not about the question of what is AI going to do? What is the human going to do? It’s more like, what are we gonna do together? How are we gonna collaborate? Together. What is your vision of how it’s going to look, you know, collaboration with AI? It’s one of the things that you’re designing in the work you’re doing with Encore and the way you see the workflow of the marketer. What can that look like? Just help us imagine that.
Wangxin: Considering what we are doing right now, I think AI can be a very powerful tool, and what I imagine is we will still need humans involved, that we will have to feed AI all of the information and the data.
We should come up with all of the problems that, what are the problems we are trying to solve. But AI should be able to help us gather all the information and try to come up with potential solutions, but again, humans have to evaluate what is the right way to solve the problem and do different types of tests.
There will always be human touchpoint in this AI collaboration. But AI can be super powerful to come up with some solution and answers, some potential answers to, the problem that you’re trying to solve.
Peggy: And the way you’re doing this, you’re looking at. The complete process, you know, bids, creative, everything. Is there something in that process, isn’t there something in that workflow that you are particularly excited about? Because we haven’t looked at automating it before? We haven’t looked at bringing the power of AI to that.
Wangxin: I think most of our main algorithm or system. They all have at least some of the AI involved, like the pricing, the user behavior and creative. It’s just some of them have more involved, some of them have less involved.
Peggy: And what’s exciting here is bringing it all together because if we’re talking about full funnel marketing, we’re talking about retargeting and UA. Then it’s not the individual capabilities, it’s again, coming back to it, that orchestration of all of them to come to the right decisions.
I wanna wrap it up with a little bit of a rapid fire. We’re looking at what’s happening under the hood because we have you here, Wangxin. A little bit geeky, but I think some people in the audience do appreciate that.
Query per second. What’s the peak query per second capacity here?
Wangxin: We process 5 million QPAs.
Peggy: Okay. That sounds impressive. I don’t know what the benchmark is for the industry right now, but that is certainly significant.
And how often do the DNN models, how often do they update?
Wangxin: It is real-time. It will keep updating.
Peggy: So the system is learning constantly and the speed is off the charts with the QPS we talked about.
So you said before, it’s like an Iron Man suit for the marketers. Let’s imagine we meet a year from now. What do you wanna be telling me about what you have helped or enabled marketers to do? That is mind-blowing and, puts that suit, puts those superpowers to better use.
Wangxin: My vision is we hope that they can use Encore as a fully-automated AI tool that helps them orchestrate all different aspects and they don’t have to make lots of human decisions on what creative to use, what format to choose, and where to serve the ads. And Encore can help them make the decisions and they just need to collaborate with Encore as one of their team members.
Peggy: And you personally, are you into superheroes? I’m just curious because you bring up the Ironman.
Wangxin: It’s so, no, no. I don’t really.
Peggy: And Okay. You are all about empowering the marketer and making something possible for them to do things differently or better with the help of AI. And so with that metaphor of the Iron Man suit, yes, it’s exciting and it’s exciting to speak with you because you’re building it. So thanks again for all of your insights and time.
Wangxin: Thank you, Peggy.
EN 
KO 
ZH