How Can Healthcare Marketers Navigate Reporting in a Post-OCR Guidance World that Aims to Protect Patient Privacy?
Last year, a healthcare client told me that they would need to stop tracking their paid advertisements due to new rules from the U.S. Office of Civil Rights. Many other healthcare organizations made the same decision.
The OCR guidance aims to protect patient privacy by strictly enforcing the Health Insurance Portability and Accountability Act — widely known as HIPAA — when it pertains to tracking user activity. This means that healthcare organizations cannot use the usual tracking technologies, like cookies and pixels, that could potentially reveal personal information about patients. As a result, legal departments in healthcare organizations are acting to eliminate digital media tracking, at least until they can sort out what kind of tracking is acceptable.
This creates a challenge for healthcare marketers who need to generate demand but now have lost their ability to track and report on their campaigns.
Building a new way to measure marketing impact
To deal with this challenge, a collaborative team at Brunner has started building a system to replace media performance reporting. We have been working closely with our healthcare clients to identify the right key performance indicators to create a new model for measuring marketing performance.
For this work, we are using machine learning and artificial intelligence including something very new — explainable artificial intelligence, sometimes referred to as XAI. An additional key piece of our new system is a knowledge base built from a well-established marketing mix model.
Here’s how our team is approaching it:
- Since we can no longer rely on digital conversions, we focus on aggregate values that comply with HIPAA, such as call volume or insurance leads.
- We set up models to generate new results regularly, focusing on the most recent campaign performance.
- We collaborate closely with Brunner’s media teams, who are running campaigns on various channels.
- We encourage our media teams to question the model scores and suggest new ways to evaluate the performance of ad units.
The collaboration between data science, media, and strategy teams is crucial for improving the quality of the marketing mix models and the usefulness of the outputs.
Keys to success in using XAI
If there is one thing you take away from this article, it should be this: use XAI with caution and only when you have an established knowledge base for marketing’s bottom-line impact on the business. This understanding can come from a marketing mix model, previous marketing experiments, or a combination of both. But without this knowledge base, it’s difficult to evaluate the AI/XAI models you are using, and you may end up with unreliable results.
In our work with Brunner clients, we incorporate insights from the marketing mix model directly into our processes that deal with AI/XAI outputs. This establishes guardrails to keep the results and interpretations of AI/XAI models in check. You may have heard how generative AI can “hallucinate.” Similarly, other varieties of AI can generate misleading results. We have to steer clear of those.
There are pros and cons to each situation. One weakness of marketing mix models is their lack of granular insights, such as those at the ad or creative level. That limitation exists for very good reasons, but we see AI as the best tool to resolve it. That is because one of the great traits of AI is its ability to accommodate highly granular data as inputs. However, when you use AI, you are typically left with a black box: something that produces helpful results but does not give you insight into its internal workings. That is where XAI comes in. XAI, while still a developing field, is a collection of approaches that probe these black boxes to measure the relative importance of the original inputs, which are easier for an analyst to interpret.
A good example comes from image classification. An XAI model would explain which pixels from an image were most important in determining that the image depicts, for example, a frog. Now apply that same concept to advertising. Just as XAI might help us identify the pixels that let us know that the image shows a frog, XAI might help us understand which ad units, or creative executions, are most effective at improving some healthcare marketing key performance indicators. We also just need to have that knowledge base in place to keep us from misinterpreting something.
Forging the future of measuring marketing impact
At Brunner, we’re excited to collaborate with our healthcare clients to build a new approach to measuring marketing performance in a way that complies with the OCR rules. We understand how important it is for marketing teams to know how each campaign and channel is performing and where we can identify opportunities to improve ROI.
If you’d like to learn more about new ways to measure marketing performance and impact, don’t hesitate to reach out. We’d be happy to chat.
Eric Perz is vice president, data science at Brunner. His team specializes in machine learning applications for media performance measurement, including marketing mix models, diminishing returns analysis, and other optimization efforts that maximize marketing efficiency. Eric has worked in agency environments for more than 10 years including the last five at Brunner, where he has supported clients across most industries. He has a Master of Engineering Management from Dartmouth College and a B.S. in Mechanical Engineering from Purdue University. | See Eric’s LinkedIn profile.