Med Ad News spoke with Nick Bartolomeo and Bryan O’Malley about their new roles at Fingerpaint and the creation of the SHIFT Performance™ Center of Excellence, the capabilities of SHIFT’s AI, the launch of Actionable Intelligence Engine™, and how they think SHIFT will be able to help clients in a proven, science-driven way.
Med Ad News: In light of your promotions, how would you describe in an elevator pitch what you do for a living?
Nick Bartolomeo: That’s a great question, it usually depends on who is asking me, because most people have no idea what we do for a living. We’ve been at this for quite some time, working to help Fingerpaint provide an offering that we feel is very differentiating in the market. And this SHIFT Performance has become a platform that we have recognized has so much value for our clients and the way that we are able to reach and engage customers and build an experience that really is meaningful, because at the end of the day what we’re trying to do and the reason why we have “Performance” in the title is deliver beyond on what your typical approach might be of impressions and clicks, and really try to get on to return on investment, and what impact this is really going to have for brands.
So in this pharmaceutical industry, as to what we’re delivering and what I feel like my role is, is looking at the landscape of how we’re able to connect and engage with target physicians or patients, to really become more educated about the product or the brand, so ultimately we can help shift their mindset and behavior to the direction that we want them to go, because if performance is a critical piece of what we’re doing, then we know how important that experience is, and we’re leveraging multiple technologies to allow us to do this. And data is probably one of the most important facets of how all this works. It helps fuel our ability to deliver relevant, timely content.
Bryan O’Malley: It’s something we’ve been working on for a while, we’ve got a lot of people brought up to speed on how this system and technology platform works. Ultimately what we’re trying to do is helpour clients shine and achieve their goals. And I’ve used this analogy, it’s kind of a mix of people and process and data and technology. And out of that magical formula you get measurable performance. And that’s what we’re really trying to get to, we’re beyond clicks and impressions and in a place where we can actually prove that the work we do for our clients contributes to their bottom line. We can demonstrate how this mixture of great technology run by industry experts through this process we’ve honed over years and years, with things like data and artificial intelligence, allow us to really dig deeply into what’s working, optimize it, and make sure that we’re helping each one of our customers’ ultimate targets, their doctors and their patients, are getting the information that they need to make informed decisions about their treatment options. And we can demonstrate that it leads to increased revenue for our clients. So that’s what we do all day, we tweak the numbers, we look at the data, we look for ways to keep adjusting and optimizing to be able to squeeze the best performance. And that’s why we called it SHIFT Performance, it’s to pull the best performance out of the investment that our clients are making in their promotion.
Nick Bartolomeo: But if my dad asks, I usually just tell him that I work in advertising!
Med Ad News: Please explain the function and capabilities of the Actionable Intelligence Engine.
Bryan O’Malley: Actionable Intelligence is the thing we’re most excited about right now. It’s the newest component to this whole SHIFT Performance platform. It is a machine learning tool, and the beautiful thing about machine learning is that it’s a great way to augment what we’re already doing. There’s a phrase that I love, “Human expertise augmented by machine intelligence.” We’ve got over 60 people in this center of excellence who are solely focused on making this platform and this experience the best it can possibly be. It expands the gamut from data scientists to digital strategists to software developers and media experts and so on, and that’s a lot of brain power we’re able to put at a problem.
As smart as people are, they look at the world in a certain way. One of the ways that machines do a better job than we do is looking at giant data sets and picking up some of those trends that are really hard for us to observe because there are so many variables and so many data points that even the smartest of us would struggle to look at millions and millions and millions of records of data and be able to identify patterns. And that’s where machine learning really excells. It can look at that data and churn through it more quickly than we can and it can identify the things that we struggle with. It’s helping us identify things like, “Hey, given this huge data set, tell me which doctors are most likely to start trialing a drug?” Or, “Which ones are mostly likely to look beyond trialing and start adopting?” Or, “Which ones are mostly likely to start lapsing?” Or, “Tell me what content they’re most likely to engage with and when?” And those things now get us out of the area of being backward-looking with analytics and saying, “Tell me what did happen” and allow us to be forward-looking and saying, “What is about to happen?” And that lets us react in a different way and be more proactive, and get our salesforce involved at the right times, get our messaging in front of the audiences at the right times, adjust what that messaging says. It allows us to be much more forward-thinking, much more proactive, and much more effective in the way that we communicate.
This system is really remarkable, it’s on the cutting edge of technology now. We’ve really been able to do a great job of demonstrating its capability, we’ve got some data models showing more than 90 percent accuracy in predicting some of these events, and the sky’s the limit. It’s really early days in machine learning and we’re taking advantage of some of the best and brightest in the industry and the technologies available to us. We’re really excited about where this will be able to take us.
Nick Bartolomeo: We’re calling it “Actionable Intelligence” for a reason, it’s our take on AI. We know how important the human aspect is to all of this. The machines can only tell you the point of where you’re leaving and when you’ve gotten to your destination, but they can’t really tell you how to get there. That’s where the human part comes in, for us to be able to help chart that path. And the aspect of what we’re able to do is leveraging all of these years of people’s experiences across media, data, digital, creative, and creating true, practical use cases that are going to be able to make us be that much more connected to the audience and providing an experience that’s going to be able to help them get to that next step. For us it’s really about being able to take the best and brightest of our human intellect, and now adding on top of it and augmenting it with machine learning has proven to us that we’re just on the cusp of some amazing work, being able to have these bright people work with now a technology that really points them and predicts some of the future.
Med Ad News: How long did it take to train the Actionable Intelligence Engine before you launched it?
Bryan O’Malley: It’s an interesting question because it dates back prior to technology, really. The tool is all well and good, but knowing what to use the tool for is part of the secret sauce that goes into it. Because you just can’t throw data at a machine learning algorithm and say, “Tell me something.” You have to know what it is that you need it to tell you and how to train it to be able to do that. It’s probably been six months to a year that we’ve been working on this to get the models working, to talk about the kinds of questions that we think machine learning algorithms can answer for us, to look at which models we think would be the right ones to apply to it, what data sets we need to gather up in order to make it work, how you isolate out the data that it’s going to learn from, from the data it’s going to test against, from the data that we actually want it to apply to. So it’s been a bit of a journey and a bit of a learning exercise, and it’s been an area where with every step we see another possibility. We’re at a point now where we feel really good about what we’ve done and we’re really excited about the capabilities it provides us. We already had so many conversations about what it could be and we have a guy internally who is constantly monitoring the latest news about artificial intelligence and sends out emails almost every day about what is going on at Amazon or Microsoft or Apple. It’s one of those parts of the industry that is moving so quickly that we’re really excited about where this is going to take us.
Nick Bartolomeo: When you think about the data that’s required, and so many of our clients are underutilizing those data – they might have built a data warehouse to have a repository of every interaction with a physician or a patient. Those actions are being analyzed to understand that if you put sales and marketing together what kind of lift and impact would you get. But what they’re missing are the opportunities, how you can leverage those data to be that much more specific and deliver communications on a one-to-one level. Now, take machine learning and put that one top of it, if you don’t have an infrastructure that allows you to have a lot of these data aggregated, machine learning is only going to be so smart. That’s where we feel that we’re ahead of the game because we started this five years ago, had a vision about building out and centralizing data in a way that we could predict from our own standpoint and prescribe what we think the next best action would be, to now having the ability to run this through the system because we do collect just about every piece of information that is available to us, to help this model be this much smarter.
Med Ad News: What kinds of data do your machine learning algorithms look at?
Bryan O’Malley: We gather a lot of different information. We gather visits and interactions, we seek out scrip data. We gather information about competitors, how our drugs’ prescription rate matches up against our competitors. We get web data, email interaction data, media data, action to click, those kinds of things. There is a wide swathe of information that we gather into our data warehouse and analyze. We feed that into the Actionable Intelligence Engine and one of the things that is great about machine learning is that in addition to trying to figure out what the right models are and the right data sets, once you isolate those and start running them through the engine it will also start to tell you which data points are the best indicators. Let’s say we were going to ask the question, “Tell me which doctors in this data set are most likely to start trialing,” Not only will it give you a ranking for each doctor, it will also tell you that these six points in this giant data set are the ones that are the best indicators. It might be past prescription behavior, with our drug or with a competitor’s drug. And as you apply the different models you start to see it move around the different data points that it thinks are the most relevant and what variability each one of those contributes to the overall calculation. It’s nice that it not only gives you the one number and end result, it gives you the indicators of what data might be most valuable in generating that result. And that helps us figure out where to get more data or what data is most critical when start working with a new client, in order to be sure of what kinds of decisions we need to make. And of course, with every new client that comes in, a new situation comes in, of what data they have available and what data we can access.
It would be nice to say that’s this black box, we’ve got it all set up so we just flip it on and it spits us out the answer we need, but there’s a lot of art that goes along with the science of making sure that we’re able to track down the data that we need – the right data in the right volume. Obviously if you have five data points it’s not going to be as accurate as having 5 million data points.
We could have come out a year ago and said, “We’re doing AI!” Well, OK, we were, but it wasn’t at a place where I would have trusted to make a business decision off of it. Now we’re at a place where we’ve done enough work and demonstrated enough capability that this is a real thing. It’s math, it’s science, it’s looking through large data sets and being able to pull out numbers and do that kind of computation that human brains are not great at. Yes, if you take a small data set and a poorly trained model you’re going to get some craziness…but we don’t want to do that when it comes to the dollars and cents, the things that our clients are spending their money on. We want to make sure that we have really put the rigor behind this, we really tested this thing thoroughly, and we have really been able to demonstrate repeatedly that if you put in this set of data, you get out these set of results. We’ve put controls in place to measure what those results mean, and that means measurable differences in the decisions we make on how to address your audience and hopefully guide them in their journey to become an adopter of your product.