Key Takeaways:

• Sweetch believes good participatory health solutions 1) make the patient feel like a person rather than just a patient and 2) empower patients to better manage their own care (jump to)
• Just-In-Time Adaptive Intervention (JITAI) and Emotional Intelligence (EI) work together to “crack” human behavior and find the best time to send a tailored notification (jump to)
• Sweetch collects data across devices to build a patient “persona” that can help “crack” the patient’s behavior patterns (jump to)
• Sweetch provides value to its pharma and medical device partners by increasing retention, increasing adherence, and decreasing customer acquisition costs (jump to)
• Dr. Bahagon’s advice to aspiring entrepreneurs: find what makes your eyes brighten in the morning, and have an acute grasp of the “true need” it addresses (jump to)

Thank you so much for joining us, Yossi. It’s a pleasure to be chatting with you. For the readers who aren’t yet familiar with Sweetch, could you give us a brief introduction about yourself, your background, and your previous experience with startups? I'm sure our readers are really excited to meet you.

Thank you, Olivia. Nice to meet you and nice to meet you all. I'm a family physician by trade. I'm still practicing, less frequently, but enough to stay connected to care in the clinic. Most of my time during the last 15-20 years was spent in the digital health space. I led the digitalization of the largest health maintenance organization in Israel called Clalit and a worldwide benchmark for nationwide deployments of digital health solutions. I also co-founded Qure Ventures, the first fund in Israel that was purely dedicated to digital health. All in all, digital health is not only what I do for a living, it's my passion, it's my life mission.

I am especially excited about patient-focused solutions, or the way I prefer to describe it, participatory health solutions: solutions that make the patient

1. Feel like a person rather than a patient
2. Empower patients with chronic conditions to better manage their own care

Would you mind sharing a bit about the company, the product, and your ultimate vision for it?

One of the biggest hurdles for an intervention is patient compliance. A pharma company, or a medical device company can spend several years in R&D. But at the end of the day, if the patient doesn't take their treatment as prescribed, billions of dollars that were invested in R&D are rendered useless. And I want to emphasize that when I'm saying compliance, it's not only about taking medications, I’m referring to a range of things that they should do as part of managing their condition, whether it's physical activity, weight loss, reporting on outcomes, coming to clinic visits, and so on.

At Sweetch, we view the patient, not as a walking disease, rather as a person with life involving various aspects they should be able to manage, one being chronic disease. Our goal is to increase the likelihood of the patient doing what he or she needs to do.

We encourage patients to take ownership by combining what we call advanced data science or AI, with behavioral models powered by Emotional Intelligence (EI) . Through our unique platform, we create a very highly personalized system that interacts with the patients based on each individual’s life habits. For example, instead of telling you, “Hey, Olivia, you walked 4000 steps today, and you have 6000 steps more to go!” Sweetch would send, “Hi, Olivia, you have 45 minutes before your next meeting, it’s raining outside. So, maybe grab an umbrella and go to the nearest Starbucks, it’s only nine minutes away. And this way, you will achieve your 90-minute activity goal.”

Everything in that message was created automatically based on seamlessly understanding what we call your “life print,” which seeks to understand questions like how do you live your life? When do you wake up in the morning? What are your favorite coffee shops? And so on. This “life print” profile is created through data we collect from your smartphone and other connected devices, which enables us to reach the patient with the right message at the right time in the right tone of voice, in the right context. It also learns from the patient's real-world behavior by seeing which messages worked and which were ignored, to dynamically optimize the messages that each individual gets on an ongoing basis.

‍

What were some of the biggest challenges designing the product for users that you approached? And how did you tackle these?

Behavioral change is very complicated. We are very dynamic, complicated creatures, so what worked for you today might not “activate” you next week, maybe because you had a busy week or were in a bad mood.

Many times we are asked, “Okay, so what's your secret sauce?” My answer is that our secret sauce is that we don't have a secret sauce. If you try to crack human behavior using a secret sauce, you will find it eventually stops working because we get used to the secret sauce, and then we start ignoring it. So, we take a very dynamic approach with our users. It's like a safe with an ever changing code: each time we want to “influence” you, we need to figure out the code. We “crack” this code by looking and listening very, very carefully to the patients’ lifestyle and habits, to find the window of opportunity to send the right message at the right time, in the right place and the right context.

This methodology is called JITAI (Just In Time Adaptive Intervention). Every time the system is about to send a message it takes into account two parameters:

1. What's the user's current context?
2. How did this user react to similar messages in the past?

If the system recognizes that sending a message now, at that place, in that context, only has a 20% chance of being acted on, it will hold off, to avoid frustrating the patient and causing alert fatigue.

What did the first iteration or minimum viable product look like? And how did data and feedback inform following iterations?

JITAI sounds like a great model, but we are all familiar with great models that don't work in reality because the problem is really, really complex. Before commercializing our product, we conducted a clinical trial led by Johns Hopkins. This pilot study took Sweetch’s fully automated intervention and tested it on people with early-stage diabetes to test our theory.

The study results were beyond our expectations; we achieved clinically and statistically significant outcomes on both clinical statistical levels, like A1C reduction (biomarker for diabetes), significant increase in physical activities, significant reduction in weight. Based on this study, John Hopkins and Sweetch got an NIH grant to conduct a much more robust study in a randomized control trial that started treating patients a few months ago to compare Sweetch with the gold standard, which is a human-based CDC-approved diabetes prevention program. This study is ongoing and will take a few years before we can see results.

After the pilot study we started rolling out on a commercial level and saw similar levels of success as the clinical trial. And as you know, AI systems improve as more people join the system; AI is not a static creature. We have recently launched our version 2.0 which features upgrades to the user experience and improvements in the algorithm.

In the trial and in general, how do you account for patient compliance and encourage patients to stick to your recommended routine? For example, if the patient is not following recommendations, can the app detect that and alert a physician to follow up with them?

Yes, it does. Just like how healthcare is a system and not just patients living in a silo, Sweetch is not a direct-to-consumer solution; it is typically recommended to patients by what we call a “trusted authority”. This can be the patient's physician, nurse, diabetes educator, health coach, or patient support program. Once this trusted authority recommends Sweetch to the patient, they also have access to a management dashboard, to monitor how the patient population is behaving, and ask the system to give them alerts on patients that had adherence or compliance issues. This is another added value from the system because now physicians, nurses, and medical team members can move from “just in case” follow up to “just in time” follow up based on real world patient behavior data.

Many times, when we train the medical team members, we say to them that when they describe Sweetch to their patients, the best way to describe it is “a tool that is a clinically validated solution that will help you be on top of your condition, that will also help me better treat you.” It takes less than 10 seconds to say this sentence, but it has a huge impact on patient trust because now it's not just another app in the App Store. It's a clinically validated solution that is part of their care cycle.

Does the risk of non-compliance on the patient's part affect which patients may be eligible to use Sweetch? Put another way, are all patients eligible to use Sweetch under physician guidance, or is it up to the provider's discretion to determine whether they're trustworthy enough to comply?

It's on a case-by-case basis. Meaning there are cases where we aim for mass distribution, give it to everyone. One of the things that we learned along the years, is also to differentiate between people who are more likely and less likely to be activated by Sweetch.

We also have tools to guide our customers, primarily pharma companies and medical device companies, on which users Sweetch will be more effective for, and which Sweetch will be less effective for.

Could you elaborate a bit more on which members are more likely to use Sweetch? For example, what does your ideal customer look like?

So, this is part of the things that I won’t be able to elaborate on because it's part of our secret sauce; we have an IP pending on this exact type of question. I apologize for not getting into details here, but there are characteristics, and even when we notice that a specific patient type is less likely to respond, there are ways to intervene.

Part of the idea is when you look at the population of patients, there is no one-size-fits-all. If you asked me, “What would be more effective, a human-based intervention, a pure tech-based intervention, or a hybrid?” I would tell you that there are some who would be frustrated if they have to speak with a human, some who would be frustrated if they only use an app, and so on. So, the goal is to determine and say, “Hey, this patient is more suited to a pure digital intervention. And this other patient may need a more high-touch approach.” You will still use Sweetch, but Sweetch will be a tool in the hands of the care provider to provide more efficient high-touch interactions.

In this digital behavior change intervention market, what would you say Sweetch’s biggest competitive advantage is?

If you would have asked me this question two, three years ago I would have said, “Hey, we have this, they don't have that, and we have this,” like a comparison table. Every company has this though, and you already get the gist of it.

For us, I think we’ve put the proof in the pudding. We are working today with top global pharma and medical device companies. One recent example is Bayer’s G4A accelerator where more than 400 companies from over 65 countries compete. We were the only company in our batch to go into the cardiovascular space and leave with a commercial agreement.

AI is an extremely powerful tool when you try to predict an outcome while feeding certain variables; however, emotions are very complex. Can you explain how you’ve defined the problem of emotional intelligence (EI) for AI to solve.

Until now, we spoke about Sweetch’s ability to turn data into actions. What we will be speaking about now is the strategic perspective of the data that we collect, or what we call “digital phenotyping.” If you look at the data world today, there is data locked in the EHR, clinical data, genomic data, biometric data such as glucose levels or blood pressure, and much more. We have added a new layer called behavioral data. When I say new layer of data, it’s going to the granularity of when the user arrived home, left home, idled at home, traveled home, left work, entered work, entered a restaurant, entered a pharmacy, entered a grocery store, arrived at the café, started walking, started running, and so on. We call these patient moments, and we create patient personas based on these moments. These personas can be characterized as, “This person is a night owl,” or an early bird, a long commuter, and so on.

We also like to get daily summaries. What does a typical Monday or Tuesday look like? And how does this compare to a weekend? This analysis enables us to create behavioral predictive analytics. For example, “Good evening, Helen, it’s getting close to your bedtime.” How do we know it’s near her bedtime? Well, it’s because Helen goes to sleep during a certain window during the week. We can then utilize this window to give a recommendation relevant to this window, whether it’s taking medication or taking a photo of your feet in the case of a diabetic patient.

At the beginning, the system starts sending messages without knowing how this specific individual will react to it. Some of these messages receive reactions while others are ignored. At the beginning, there is not enough consistency in the ability of the system to activate the patient in a coherent way. However, through reinforcement learning, after four to six weeks, the system learns the patient’s behavior and at which times messages are more likely to activate patients and when messages are most likely to be ignored. Based on this, Sweetch can send messages to patients at the right time.

We also focus on the tone in which we approach our patients. In general, there are four categories of voice tones, with hundreds of sub-categorizations. Different patients are activated by different tones of voices, but it’s more complicated than that. For example, if I say to you, “Hey Rishi, today, you’re number one in your age group.” This is a very powerful message to encourage people. However, after running into this message for a few days, it will lose its effectiveness. So, it’s not just about the message you receive, but it’s about the overall music the system plays on your emotional piano. And every patient is a different piano.

How is Sweetch able to interact with other apps on the phone’s OS and what measures are put into place to ensure patient privacy?

I will start from a more technical answer. We are GDPR, HIPAA, Australian privacy, and Canadian privacy compliant. Before we enter any country, we make sure that we are compliant with the local regulations. On the privacy and security level beyond that, there are various measures we place to protect these. Once again, I think the proof is in the pudding. We work with several multinational companies and you can imagine how sensitive they are to patients’ privacy. We went through hoops to ensure that all the demographic data is completely separated from the behavioral data. The connection between the two is heavily encrypted. So even if the behavioral database is compromised, which has never happened, it still does not allow access to the patient’s personal data. Moreover, just because we deliver recommendations based on the patient’s availability. It doesn’t mean that we know with whom you meet, just that the time is blocked. The fact that we are recommended to the patient by a trusted authority provides an additional layer of trust. That’s part of why we prefer to go through long sales cycles of B2B and sign agreements with multinational companies rather than going direct to consumers.

Many retail and social media companies collect similar data which is used to convince users to buy a product or service. We are trying to do good with that same data. The concern will always be there regarding privacy, despite being HIPAA- and GDPR-compliant, but our hope is that because Sweetch is recommended by a professional authority they trust, and the perceived value is great enough – they will embrace using the app. That’s what our current engagement data shows. Big time.

Can you explain how Sweetch generates their revenue?

Our model is B2B2C, meaning patients are not paying unless we decide that it is part of the engagement model. The business model is based on our big pharma and medical device manufacturer partners. So why is it worth it for these companies to pay for our product?

There are three things that Sweetch affects:

1. Higher retention: We increase the time that people use the solution, whether it’s a medication or device.
2. Increased adherence: Frequency of use increases over the same time period
3. Lower customer acquisition costs: With Sweetch, the offering is more attractively positioned

Our model is based on two levels, like any SaaS company, one is license and two is patients per month. We like to emphasize that patients per month is specifically referring to active patients. An active patient is someone who is losing weight, refilling a prescription, or renewing their device. The definition of active users changes per use case, but in order to make it a true win-win, we work with our partners to make KPIs and base our model on these.

Sweetch places a very strong emphasis on passion in its company culture. How do you believe that this passion-integrated culture is reflected in your product and how does it help with achieving your OKRs and KPIs?

Wow, that’s a great one, and I will go personal here. I have spent many years working in the digital space and have succeeded in most of the projects I was involved in while learning from those that didn’t succeed. I’m in a position today where we could have had this discussion from the beach and for many years I was an investor. For me, Sweetch is not just another opportunity to make a profit, we want to impact hundreds of millions of patient's lives and this is the language that is ingrained in the company culture. When I interview new employees, I want to hear this value loud and clear, it’s not just about seeking KPIs and increasing revenue cycles.

I’m happy you asked this question because the values I’m speaking about are not only essential to enjoy work but also the essence for succeeding. It’s great for someone to write amazing code but if we don’t align on the passion and impact that you wish to have, Sweetch is probably not the place for you. We put a lot of emphasis, not only on our website, but also in the company’s culture. We recently had a meeting with the entire company to present our 2023 strategy, emphasizing transparency and togetherness. The strategy is to first build on values and then on business KPIs. Building a startup is a roller coaster and sometimes a lot can happen in just one day.

Do you have any final thoughts about Sweetch or any advice you’d like to impart to future entrepreneurs?

One thing is something we’ve already touched upon. Make sure that what makes your eyes brighten when you wake up in the morning is not your dreams about an exit but dreams about how you can have an impact. Another is that when you start a company, make sure you really understand the need. Many times entrepreneurs are excited about the technology they built but end up not bringing value as they didn’t touch a true need. It may be a need, but not a need that many people are willing to pay for.

‍

Comments and opinions expressed by interviewees are their own and do not represent or reflect the opinions, policies, or positions of DeciBio Consulting or have its endorsement. Note: DeciBio Consulting, its employees or owners, or our guests may hold assets discussed in this article/episode. This article/blog/episode does not provide investment advice, and is intended for informational and entertainment purposes only. You should do your own research and make your own independent decisions when considering any financial transactions.