In this 2017 I‑COM Data Startup Challenge presentation, Adam Spector, co‑founder of LiftIgniter, argues that personalization is essential for survival in today’s digital ecosystem. He explains that every user interaction—menus, images, titles, ads—is effectively a recommendation that must be tailored in real time using machine‑learning at scale. Spector critiques traditional analytics, cohorts, and A/B testing as inadequate for the volume and velocity of modern data, and outlines LiftIgniter’s two‑step integration process (send non‑PII data, receive personalized recommendations) that promises a guaranteed minimum 20% uplift and typical 80%+ improvements across key metrics. He backs the claim with case‑study results: an e‑commerce client saw 240% click increase, 150% bid increase, 102% conversion lift; a media client saw 228% click increase and 320% engagement lift, all measured via continuous A/B testing. The talk then shifts to future product direction—expanding personalization to every digital property and device—and a brief Q&A where Spector discusses roadmap priorities. Finally, a Barclays representative, Yasemin, describes banking’s need for predictive, intelligent personalization, especially for Millennials and Gen Z, and how combining proprietary banking data with LiftIgniter’s platform could enable social‑scoring‑driven risk models. Spector confirms that proprietary data can be securely fed via API to enhance personalization for financial services.
Questions this video answers
Adam Spector says the focus is on expanding the platform so that every digital property will have built‑in personalization within five years, adding more refined products and covering all verticals, devices, languages, and data sources.
Chapters
Hi there, I'm Adam, the co-founder of LiftIgniter. We're a machine learning personalization company. Our core goal is pretty simple. We want to personalize everything, every action, every user interaction, everything that someone sees on that site at that moment in time. The reason you need to do it is because personalization equals survival in today's digital world. Literally, your users have come to expect a personalized experience on Facebook, on YouTube, on Amazon. Wherever they interact with users, they expect to get what they want right then and there. If you don't give it to them, another competitive site is frankly a click away. And here's also the big thing that you have to keep in mind. Every single thing is a recommendation. What you see on the site, the menu items, the page, the pictures, the title, you name it, what to show, when to show it, and how you show it to the users is a recommendation. The question becomes, of course, how you personalize it for that individual user, and frankly, the only way to do that is scale is via machine learning. Here's how we work at a high level. It's a little bit of a confusing picture, but the core idea is this: your sites or apps or digital properties are filled with tons of users who are frankly interacting with a site in a variety of ways. You have lots of different types of content, be it video, articles, images, items to buy, ads, you name it, and they're always in flux. Essentially, your site and your users are evolving together. We look at all that information, put it together, and then learn from those users in real time as they flow through the site and take action. Here's the thing, a lot of the people here, especially at ICOM, are spending a lot of money on a variety of digital spending. The problem is a lot of that money is wasted because it's frankly not optimized. You get someone to your site or your digital property and suddenly you don't really know what to do once they arrive. You forget about the whole optimisation chain once they get there, which to me is sort of silly. It's a big waste and inefficient use of resources. The other— the reason though that a lot of you guys do this is because frankly it's really tough to personalise. Frankly, your teams are overwhelmed with a massive amount of data. There's too many items on the site. You have billions of impressions across every single touchpoint every single device. That's just an overwhelming amount of information. On top of that, we all know we're all diverse here at this conference, obviously, but frankly, we're all diverse from ourselves at every single point in time. In fact, if I do my job right, every single one of you will be different at the end of this presentation than you were before it even began. You're changing because of what I say, and when you ask me questions, I'll change as well. The final piece is frankly current tools that are out there simply don't do the job right. Analytics, cohorts, segments, A/B testing, all of those in a machine learning personalization world are essentially dead. They're unable to deal with the fact that user segments are far too wide. The idea of targeting a 25-year-old male is silly when I could be someone different at every single moment. The fact that I would look at social data and to tell me what I want to do right now also doesn't make a ton of sense. So my team, my co-founder is a PhD in theoretical machine learning from Princeton, built the world's first machine learning personalization system at Google. They launched this system on YouTube. I'm sure you all have seen videos on the right-hand side rail of YouTube. Those are personalized in real time for you. When they launched this system, it had double-digit percentage improvements in engagement time sitewide. and added billions of dollars in incremental new revenue. We're now bringing similar technology to the rest of the world. We average over 80% improvement for our customers in any key metric they want— click-through rate, engagement, conversion— for 2 hours of time to integrate. We're on over 5 billion— actually, it's more than when I produced this— over 5 billion page views a month. We actually guarantee a minimum of a 20% improvement on your digital property. Our view of the world is simple: one platform for all personalization. Every vertical, all types of content, all devices, all languages, and all data sources. Everything that touches the user should be personalized in real time. It's pretty simple to integrate. It's 2 steps. You send us data, we— you then receive the results. It's all non-PII, and the data is fully privacy protected. We have large customers in Europe and globally, so we know how to deal with those situations. How to personalize is pretty simple. You choose what you want to do on site, app, search, and email, type in your URL, choose what you want to personalize, choose a goal KPI and different rules that you might have set, and then good, you get started. And that's essentially how our system works. It's that easy to get 80+ percent improvements. Here's an example from one of our e-commerce customers. 240% improvement in clicks, 150% improvement in bids, and 102% improvement in conversion. Here's one of our media customers, 228% improvement in clicks and a 320% improvement in engagement. These are all A/B tested in real time, 24/7 for our customers. So we're LiftIgniter. The simple fact once again is you need to personalise, otherwise you'll be crushed by the big guys who are doing this today. Time is a zero-sum game, and if you don't get your users to win on your site, they will bounce off and find somewhere else that is giving them what they want. Thank you so much.
All right, now what we're going to do is we're going to have a question from the juries. If there's a question from our jurors. Any takers? Yes.
Hi there, I'm really interested by what you see as being your next step in terms of either feature or what's coming next.
Yeah, absolutely. So, you know, from a feature perspective, a lot of our focus is really just on expanding this. We truly believe that every single property, digital property in the next 5 years will be required to have personalisation built in. Now we have over 5 billion page views a month and some of the largest customers in the world. But that's just the tip of the iceberg, right? Every single way you touch a user needs to be personalized, and that's just a matter of expanding our base of products. So it's doing that and then making the product even more refined.
Very cool. All right, and now we're gonna hear from Barclays, from Yasemin, who's gonna share her perspective.
All right, thank you so much. Very exciting. So in banking today, in financial services, we actually— personalization is a big topic for us.
Yep.
Millennials are coming in, it's the next generation, they all have ADD, and it's multi-device, omnichannel all the time. So I see a huge, huge sort of application use case for us of how we can accumulate all the various channel data, but then enrich it with social data, lifestyle, any type of other, you know, behaviors that we can observe in digital channels to go after them really in a more intelligent way. It's all about predictive intelligent banking, and if we can actually combine that type of, you know, platform service with our own pricing models, propensity models, to go and provide the right product with the right pricing at the right time, then that would be the ultimate, I think, future for us for banking.
Yeah, I know that. I totally agree. I mean, and one of the things you talked about, Yasmin, I mean, Combining the proprietary data that you guys have is something we actually didn't talk about in the presentation, but something we actually do with some of our customers where they have their own specialized datasets that can be really valuable in doing the personalization really well. And if they have that, they just pass it to us via secure API, and then we're able to turn that on to add more personalization and once they take that action for your users to give them that fully personalized experience that they expect from a high-quality bank.
And this type of service for us, especially as we are going to the next generation of banking services, so our entire lending products, are finding new ways of how risk scoring is going to be done. But again, millennials, Gen Z coming in, we're looking at social data, social scoring rather than our traditional risk methodologies that we use. So hence, this would be a start and then how we enrich it further with other data points and go after every click counts.
Every click counts. Exactly right. You got it.
Very good. All right. Thank you, Yasmin. Thank you.
