AI-Driven Paid Media Strategies for Healthcare Marketing Success

Announcer: Welcome to the Ignite Podcast, the only healthcare marketing podcast that digs into the digital strategies and tactics that help you accelerate growth. Each week, Cardinal’s experts explore innovative ways to build your digital presence and attract more patients. Buckle up for another episode of Ignite.

Lauren Leone: Hey, everybody. Welcome back to Ignite Healthcare Marketing Podcast. I’m Lauren Leone, Chief Growth Officer here at Cardinal. I have with us again today, Alex Kemp, our Director of Applied Analytics. If you listen to our podcast regularly, you saw that we just dropped our Performance Max episode with Alex, so take a listen. Today we’re going to be talking about AI. This is the hot topic. We’ve seen it at every conference we’ve been to this year. We just posted or held our live stream event on AI with our chief strategy officer, Rich. We want to talk about how AI can be used in content production, predictive analysis, and user research. Let’s open up with just a general understanding of AI capabilities. Give me the top ways that firms like Cardinal or our clients are using AI right now.

Alex Kemp: The top ways are going to be creating content or copy at scale. That’s going to be one of the biggest levers as copywriting has always been the hardest part of marketing, I think, at least in a paid search realm. Being able to give a lot of context to a model and allow it to develop that creative for you rather than you have to start from scratch. I think the other big one is forecasting, predicting outcomes to better align and adjust your strategy to those outcomes. Looking at historical data, using machine learning to predict outcomes and look at different scenarios.

Lauren: You said the perfect thing when you opened that statement, which is, to give it context. The biggest misconception of AI is that it can or will replace marketers, can replace the functions in what we do. Talk about how important that is, and what doing it wrong can yield.

Alex: I always say AI is not magic. None of these tools are to the point or maybe ever going to be to the point where you can just plug in a very short sentence and it gives you exactly what you want. Context is key, that garbage in, garbage out principle of if you don’t give it the right inputs, you’re only going to get as good of the outputs from those inputs. I always err on the side of too much information when I’m trying to give those models context because, again, the more context and details you can give them, the more useful the output will be really for any application, not just copywriting or predicting a forecast or anything.

Lauren: Let’s talk about a few of those examples and what good going in looks like versus bad going in. Let’s start with content. A couple of the ways you might be using some of the long-form content tools to write blog posts or on-page content. In your case, Alex, you’re using it thinking about ad copy at scale, for example, where you’ve got 100 locations, and you’re trying to build out a new account and you’ve got to write copy. We want copy to be very specific to the keyword, the user, and the intent. You’re talking about writing copy across different ad groups and campaigns. What does it look like giving an AI tool good inputs to produce something that you can actually use in ad campaigns?

Alex: There’s going to be some staples to that. I think things like the brand voice, how you want your brand to be actually portrayed, the voice it’s supposed to be using, what are your key value props. What makes you different from the competition? If you just give them a website and say, “Write me some headlines for this website.”, they’re just going to determine that you’re a dental website and we’re going to give you a bunch of very generic, bland-sounding headlines and descriptions.

Whereas if you give them all the context about your business, what your goals are, who you are as a practice, and how you’re different from other practices, it just gives you so much better of an output because, again, it can’t read your mind. You have to just upload a bunch of information to it and then refine from there. It’s much easier going that way versus starting off very broad and then having to refine from there.

Lauren: You said the other critical piece, which is refined from there. You can do all the great work to give it the context going in. You’ve done the good in, the out still is going to require refinement and human oversight. What kind of lift are you seeing if you’re giving really great inputs for ad copy and then you’re still reviewing it? How much work is that?

Alex: There’s the study that was done where basically, by using AI, campaigns have improved by about 20%, but then there was about another 15% gain on top of that when there was a human insight intervene. Obviously, that’s a study that you can’t apply to every single situation, but the idea is just that AIs plus humans are always going to be better than just AIs on their own.

Lauren: Just humans, maybe.

Alex: Yes, exactly. That too. Yes, having the human intervention piece and refining it is what gets it to that 100% where you want, I think AI can get us to 70%, 80%. We’re still not to the point yet where humans are not needed in the equation.

Lauren: You may be getting things as simple as catching phrases or statements that we don’t want to say accepts all insurance if maybe you don’t. There’s the tactical things like that. Then you’re also reading that copy and saying, “Is this really what we think our patient is trying to get out of this experience with us or not?” Taking out the things, even sometimes that feel like that’s not really how a human would say that statement.

There’s a little bit of all of that. Copy is something that people understand AI can be used for and probably are most commonly using it for now. I think the really interesting AI that we have come across are in the user research and in the predictive analysis. Those are fascinating ways to use AI and data. Tell us a little bit about how you’re using it in those two realms.

Alex: For user research, one of the analyses we will do is voice of the customer analysis. It’s actually when you start stacking AI on top of other tools. Like we have a tool that can scrape reviews from clients or your competitors in your area. Then basically use AI to disseminate that ocean of information to determine what do people like about your competitors, what do they not like about your competitors, what do they like about you, what do they not like about you, so you have a better idea of how to position yourself in the market and to better speak to that user.

I think, again, using tools to gather lots and lots of data and then using AI to understand that data at scale, is a really streamlined way for us to understand our users better and to adjust our strategies. That’s one thing for user research.

Lauren: Something that we probably don’t even think of as AI now, is we talk a lot about call tracking and our partners in call and lead scoring. Even like line or patient prism, when we say that we’re automatically tagging calls based on what the outcome was, whether they were new or returning, and then analyzing the reasons not booked and using that then to inform our paid strategies, that’s AI in our daily lives.

Just another example. Probably the AI use case that I know you’ve spent the most time with recently is predictive analysis. Talk about how you’re using AI to both inform investment strategies and what’s going into campaigns, but also in maybe reporting out to clients on the value that their investment is driving.

Alex: Having forecasts are really important because it allows you to understand what is good and what is not good, rather than just using historical data or month over month or year over year. What we’re doing right now is essentially using one of Google’s newest tools called Vertex AI, which uses Gemini, which is their AI model. Essentially what it allows you to do is use machine learning models to train a model and then predict outcomes off of that model. What we can do is basically take clients’ data, their ad spend, whatever outcome they’re looking for, and then layer in all these other variables like seasonality, other things in the market that will affect the outcomes that you’re wanting.

Then what you can do is the model can train on that data. Then we can basically say, “Okay, if we were to spend 300,000, what is our expected output here for the next month?” It’s much more nuanced than just, “Well, our cost per lead last month was $100, so let’s think it’s going to be $100 again this month.” It’s more like, “No, there’s actually going to be, due to seasonality or due to recent changes in campaigns, the forecast will actually look a lot different.” It’s a much more nuanced way of looking at forecasting. That’s where we’re leveraging AI and machine learning to take it to the next level.

Lauren: What about compliance in all of this? In those instances you just mentioned with predictive analysis, are you taking into account in that model, any data that would need to be protected? Is that something we need to be aware of or thinking about?

Alex: Yes. Really the area you want to be careful with in terms of that is probably around user research. If you’re uploading patient data to any AI tool, that’s probably something you want to avoid just for any vulnerability in the future. I think that really comes into play when you’re doing user research and maybe you’re batch uploading your patient list to see what is your typical patient demographic look like. That’s where you want to be careful about having de-identified information, anonymized information so that you’re not exposing yourself potentially to any HIPAA violation.

Lauren: Or considering solutions that are localized, maybe on your device instead of cloud-based that are shared so that you can use the model and it only exists for you. It doesn’t learn off of other information, but it also doesn’t take your information and allow anybody else to learn off of it.

Alex: Exactly.

Lauren: Just some things to think about as y’all are assessing AI solutions. I know you mentioned Vertex and Gemini. Any other favorite tools you want to give a shout-out to or suggest that anybody listening can take a look at?

Alex: Anyword is a copywriting tool that will, again, going back to what we were talking about, it allows you to input information. It uses ChatGPT in the backend, but it’s a really nice tool to, again, streamline your ad copy development. It also does blog posts and other content types like that as well. It’s good for that. I think Anyword and, obviously, ChatGPT. I’m a daily user of that.

Lauren: One other cool one I would mention to the group, Synthetic Users. It’s a concept talking about user research where you can essentially create AI users. If you want to poll, let’s say, 500 parents with children ages 2 to 18 to get their feelings on maybe some messaging options that you have for your pediatric client, you could pull that then audience. You can give demographic, geographic, psychographic information to the model to create those users and then collect basically survey responses without it having to be real human beings.

Not something, again, I would rely entirely on for all of your user research, but a really cool way to maybe validate some concepts that are based on it to do it quickly and on the cheap too. That’s another one that we’ve been using lately. Alex, thank you so much for joining talking about AI. These are just some really high-level concepts here. Definitely take a look at our livestream on our website that went a little bit deeper into the topic and let us know if you have any questions. Thanks.

Announcer: Thanks for listening to this episode of Ignite. Interested in keeping up with the latest trends in healthcare marketing? Subscribe to our podcast and leave a rating and review. For more healthcare marketing tips, visit our blog at cardinaldigitalmarketing.com.