Reduce Customer Churn from Disengagement: Lessons from Behavioral Health

The Problem No Dashboard Catches Until It’s Too Late

“About 25% of people in behavioral health treatment disengage from treatment primarily due to not taking their medications and the subsequent erratic behavior that might result.” — Tony Sterns, Founder, Valera Health

That statistic is not a behavioral health statistic. It is a churn statistic. One in four users quietly fails to get value from their product, their behavior degrades, and they exit — not because they chose to leave, but because the system never caught the early signal.

Tony Sterns is the Founder of Valera Health and the researcher behind a clinically validated, FDA Class 2-designated medication adherence system. He won the American Medical Informatics Association national contest in 2019. What he built to solve psychiatric medication non-adherence translates directly to the mechanics of B2B SaaS disengagement: users who stop using a product don’t announce their exit. They drift. And the systems designed to re-engage them are often the very reason they stopped engaging in the first place.

The frameworks Sterns developed — Situational Smart Alerting, Soft Operant Conditioning, and the Care Team Integration Loop — are disengagement intervention architectures. The terminology is clinical. The logic is universal.

Key Takeaways

Disengagement is predictable and preventable when you design alert systems around actual failure events rather than scheduled prompts. Tony Sterns built a clinically validated adherence platform that reduced behavioral health patient dropout by replacing passive reminders with behaviorally precise triggers. The core insight: most re-engagement systems train users to ignore them because they fire at the wrong moment. Fixing the trigger architecture — not the message or the frequency — is what changes outcomes.

• Alert timing is the core variable: Reminders fired after a user has already completed an action create detraining — a behavioral conditioning loop that makes future alerts invisible. Alerts must fire only when a failure event actually occurs.
• Physical product design can actively harm retention: Countertop dispensers and dashboard notification centers remind users of their dependency rather than reinforcing autonomous behavior. The goal is to make positive behavior habitual, not to keep your product visually present.
• 25% disengagement is a system design failure, not a user failure: When 1-in-4 users disengage, the problem is structural. Sterns’ research shows the failure traces to how existing alert architectures are constructed, not to user intent.
• Soft operant conditioning works at scale: Bandura’s decision-point cycle, applied correctly, makes desired behavior habitual without requiring ongoing prompting. Users internalize the action and stop needing external triggers.
• Care team integration prevents silent churn: Real-time alerts to clinicians (or CSMs, in SaaS terms) when a user misses a dose eliminate the lag between failure event and intervention. Proactive outreach beats reactive damage control every time.
• Visual AI can detect disengagement signals before humans can: Early-stage side-effect monitoring through AI assessment catches deterioration weeks before clinical observation — directly analogous to predictive churn scoring in SaaS.
• Co-opetition is a market entry requirement in regulated spaces: Sterns explicitly collaborates with direct competitors because no single vendor can solve system-wide adherence problems alone. Enterprise B2B SaaS companies entering regulated verticals face identical dynamics.

Deep Dive: What Medication Non-Adherence Teaches Us About Disengagement

Why Traditional Re-Engagement Systems Make Churn Worse

Most re-engagement systems — onboarding nudges, win-back email sequences, in-app tooltips, push notifications — are built on a flawed behavioral model. They fire on a schedule, not in response to a failure event. And when they fire at the wrong moment, they don’t just fail to help. They actively condition users to ignore future alerts.

Tony Sterns identified this problem with clinical precision in the context of psychiatric medication. The mechanism he describes maps directly to SaaS product engagement.

“When they have that 7-day pill box and their digital watch, it was always going on after the fact. You’d already taken it. So you were sort of always training yourself to ignore the alarm.”

This is detraining: repeated exposure to irrelevant alerts weakens the behavioral response. In SaaS, the equivalent is the Monday morning product digest that goes to users whether or not they’ve been active, or the in-app tooltip that fires on a feature a user already uses fluently. Over time, the user’s brain classifies all alerts from your product as low-signal noise. The moment you actually need to reach them — when they’re genuinely at risk of churning — your message lands in a cognitive dead zone.

The fix is not better copywriting. It is better trigger architecture.

What Is Situational Smart Alerting — and How Does It Apply to Disengagement?

Situational Smart Alerting only fires a notification when a user has actually failed to complete a required action within a defined time window — not on a schedule, not as a courtesy reminder, and never after the action has already been completed. The alert is triggered by the absence of a success event, making it both contextually accurate and behaviorally reinforcing.

The Situational Smart Alerting framework Sterns built operates in five steps:

1. Establish a time window for when the target action should occur
2. Monitor whether the user completes the action in the first half of the window without intervention
3. If the action is completed, record it automatically — no alert fires
4. If the user misses the action, trigger a single, contextually accurate reminder
5. Notify the care team (or account team) only when a genuine failure event occurs

“We came up with this idea we now call situational smart alerting. Which is that we need we get a window around the time of taking and if you take it in the first half, then when you take it, our device knows that you’ve taken it and there’s no alert.”

In chronic disease management platforms and in B2B SaaS alike, this architecture has a measurable effect on alert fatigue. Users who receive alerts only when they’ve genuinely missed something treat those alerts as high-signal. Users who receive scheduled reminders regardless of behavior learn to treat every alert as noise. The difference in long-term engagement is not marginal — it is categorical.

This is the foundation of any serious medication adherence software or psychiatric care workflow automation strategy. It is equally applicable to customer success automation, product adoption campaigns, and renewal risk management in SaaS.

How Soft Operant Conditioning Creates Durable Behavioral Habits

Soft operant conditioning — grounded in Bandura’s decision-point cycle — makes desired user behavior habitual by having users make repeated, voluntary decisions to take a specific action from the same location at the same time. Over enough repetitions, the behavior becomes internalized. The user no longer needs an external prompt.

Sterns describes the experience from his own life with characteristic clarity:

“It’s actually a form of soft operant conditioning. And so myself when I took a daily medication for my psoriasis… I would find myself in the kitchen, you know, at 6:00 and wonder I’d already eaten dinner. Why am I back in the kitchen? And all of a sudden the light on the pod would come on and I would go, ‘Oh, I’m a trained cat here to get my dinner.’”

The behavioral implication for mental health digital therapeutics — and for any SaaS product trying to build usage habits — is significant. The goal is not to create dependency on your notification system. The goal is to make the behavior so habitual that the notification becomes redundant. Products that achieve this have dramatically higher long-term retention because the user’s behavior is no longer contingent on your engagement infrastructure.

The Soft Operant Conditioning for Medication Habits framework removes physical reminders from the patient’s environment entirely:

1. Eliminate visible pill dispensers from the patient’s space — avoid the “temple of pills” effect
2. Route alerts through ambient devices (smartphone, smartwatch, smart speaker)
3. The patient makes a daily voluntary decision to go to the medication location
4. Repeated activation at the same time and location reinforces the decision-making pattern
5. The behavior becomes habitual and internalized without ongoing device dependency

The “temple of pills” critique is a precise diagnosis of a failure mode that exists across product categories:

“The engineers have created these, you know, Keurig-like devices on the counter that I call it the temple of pills and that the temple of pills is calling you over remind you that you’re sick and to take your medication.”

In SaaS terms: a dashboard full of red warning indicators, a customer success portal that surfaces every instance of underutilization, or an onboarding checklist that stays permanently visible — these are all temples of pills. They remind users of what they haven’t done rather than reinforcing the positive behaviors they have developed. Healthcare compliance architecture that prioritizes visible accountability over internalized habit formation will reliably underperform systems that do the opposite.

How Care Team Integration Closes the Loop on Silent Churn

Silent churn — users who stop engaging without ever raising a flag — is the most expensive churn variant in both healthcare and B2B SaaS. By the time a clinician notices a patient has been non-adherent for weeks, the behavioral consequences have already cascaded. By the time a CSM notices a renewal account has dropped to 20% product utilization, the buying decision has often already been made.

Sterns’ Care Team Integration Loop solves this with real-time failure event routing:

“If you skip the medication, you can have initiate a discussion with the care. And if you’re forgetting to take the medication, which happens, you know, occasionally, you’re distracted or something, then it reminds you to take it. Only when you miss the medication do you get on what we call our daily intelligent action list. And this is another differentiator for us is that we’re connected to the care team.”

The phrase daily intelligent action list is doing specific work here. It is not an alert dump. It is a curated, failure-triggered intervention queue — surfacing only the accounts (or patients) that actually need human attention that day. This is precisely what high-performing telemedicine medication management platforms and enterprise SaaS customer success teams do differently from average ones: they separate signal from noise at the infrastructure level, not through manual triage.

The framework steps:

1. Adherence data feeds to care team dashboard in real-time
2. Missed doses (or missed product actions) trigger the daily intelligent action list
3. Care team initiates proactive check-in or treatment adjustment
4. Visual AI for patient monitoring provides additional clinical context on side effects and deterioration
5. Care team uses aggregated data to inform treatment decisions based on medication effectiveness and tolerability

The visual AI component is particularly noteworthy for teams thinking about clinical outcomes measurement and predictive churn scoring:

“We’ve developed a set of visual AI assessments that allow for you to just answer a few questions about your day and essentially determine level of anxiety, depression, or whether you have side effects. One in particular for the antipsychotic medications that we’re focused on for people with schizophrenia and bipolar disorder… tardive dyskinesia, which is an abnormal movement disorder… You want to see that earlier than is typically understood to be seen.”

Detecting deterioration earlier than clinical observation is possible through behavioral signal analysis. In SaaS, the equivalent is identifying churn-predictive usage patterns — feature abandonment sequences, session length decline, support ticket sentiment — before a renewal conversation surfaces the risk explicitly. The architecture that makes early detection possible in Sterns’ system is the same architecture that powers effective predictive churn models in chronic disease management platforms and enterprise SaaS alike.

Navigating FDA Class 2 Designation for Hardware-Software Medical Devices

Building regulated hardware-software medical devices requires a credentialing strategy that most software founders are completely unprepared for. Sterns’ path to FDA Class 2 medical device designation for Valera Health’s combined app, IoT pill dispenser, and cloud infrastructure involved 513G pre-submission discussions, predicate identification, and assembling a regulatory team with credentials most early-stage companies don’t have access to.

“We’re in pursuit of a class two designation which is our full system, our app, our device, our IoT pill dispenser, and the cloud that we utilize with these AI components as well. The class two designation comes from the device being built in a quality manner… doesn’t require any additional validation the way an internal device with a class three device.”

The FDA pre-market submission process for a Class 2 combination device is meaningfully less burdensome than Class 3, which requires rigorous post-market clinical validation. But it still demands FDA regulatory expertise, HIPAA-compliant cloud infrastructure, and clinical credibility that must be assembled into a team before the first regulatory submission — not after.

Sterns is unambiguous about the team requirement:

“First, think carefully before you take on a problem that involves hardware. Absolutely. But also wrap yourself with a great team… Having a team that you can count on to do these hard things is what you need to be successful. It’s too hard a lift.”

His lead regulatory advisor, Dr. Fred Ma, brings 25 years of FDA regulatory experience and a background as a neurologist and brain surgeon. The 20-year working relationship between Sterns and Ma is not incidental — it is a core competitive asset in a regulated hardware-software integration environment where credibility is a prerequisite for market access.

Why Co-opetition Is a Non-Negotiable GTM Strategy in Healthcare

Founders entering healthcare, insurance, or any heavily regulated vertical frequently underestimate the degree to which healthcare vendor co-opetition determines market access speed. No single platform can solve system-wide adherence problems independently. The care ecosystem requires interoperability across EHR vendors, payer systems, pharmacy benefit managers, and competing care delivery platforms.

“You really need to work with all the other healthcare companies because you can’t it’s all about the word in the textbooks is co-opetition, right? So you even your enemies are your friends, so to speak. Your closest competition, you need to be friends with because you need to work together.”

For medical device go-to-market strategy, this is not a soft suggestion. It is a hard constraint. Interoperability partnerships, shared data standards, and clinical credibility built through collaboration with established players are often faster paths to market than head-on competition — particularly when the market is fragmented across regional health systems, payer networks, and specialty clinics.

The IoT medical device ecosystem Sterns operates in exemplifies this dynamic. Contract manufacturing for medical devices requires supplier relationships. HIPAA-compliant cloud infrastructure requires platform partnerships. Clinical validation requires hospital system relationships that competitors may already have established. Building these networks requires treating potential competitors as potential partners first.

About Tony Sterns

Tony Sterns is the Founder of Valera Health and the architect of a clinically validated, FDA Class 2-designated medication adherence system targeting behavioral health patients with schizophrenia, bipolar disorder, and related chronic psychiatric conditions. His perspective matters because he has built and validated — not theorized about — a system that reduces disengagement in one of the highest-stakes, most complex user retention environments that exists.

Sterns won the American Medical Informatics Association national contest through the MedStarter organization in December 2019. He has spent the subsequent years assembling a regulatory, clinical, and engineering team — including Dr. Fred Ma, a neurologist and brain surgeon with 25 years of FDA regulatory experience — to navigate FDA pre-market submission, regulated hardware-software integration, and the clinical partnership requirements of psychiatric care delivery. His research-grounded approach to behavioral science, particularly his application of Bandura’s decision-point cycle to medication adherence, gives his disengagement frameworks a clinical specificity that most SaaS retention frameworks lack.

Ready to Stop Disengagement Before It Becomes Permanent Churn?

The frameworks Tony Sterns built for behavioral health — Situational Smart Alerting, Soft Operant Conditioning, and Care Team Integration — are disengagement intervention architectures. The problem they solve is identical to the one facing every B2B SaaS company watching users quietly drift toward non-renewal: alerts fire at the wrong moment, product design reinforces passivity instead of habit, and the care team finds out too late. If your current retention stack is built on scheduled nudges and reactive CSM outreach, you are running the same system Sterns diagnosed as fundamentally broken. The fix starts with trigger architecture — and it scales from there.

Talk to a Growth Strategist →

Frequently Asked Questions

How does situational smart alerting prevent medication reminder fatigue?

Situational smart alerting only triggers a reminder when a patient actually misses a dose within a defined time window. Traditional systems fire alerts after the action is already complete, training users to ignore them over time — a process called detraining. By restricting alerts to genuine missed doses, the system keeps reminders contextually relevant and actionable, preventing the habituation that renders conventional pill box and digital watch alarms ineffective for long-term adherence.

Why do traditional pill dispensers fail to improve long-term medication adherence?

Traditional pill dispensers create two compounding problems. First, they alert patients after a dose has already been taken, conditioning users to ignore alarms. Second, countertop dispensers serve as constant visual reminders of illness — what Tony Sterns calls the “temple of pills” — which undermines patient autonomy and mental health. Effective adherence systems remove physical pill reminders from the patient’s environment and route alerts through ambient devices only when a dose is actually missed.

What percentage of mental health patients quit treatment due to medication issues?

According to Tony Sterns, approximately 25% of people in behavioral health treatment disengage from treatment primarily due to not taking their medications and the erratic behavior that results from non-adherence. This disengagement rate represents a structural failure in psychiatric care delivery — one that existing pill boxes, digital watch reminders, and countertop dispensers have consistently failed to solve because they rely on alert patterns that train patients to ignore reminders over time.

How can AI detect medication side effects before clinicians observe them?

Visual AI assessments can analyze patient responses to daily self-report questions about anxiety, depression, and movement to detect early indicators of medication side effects — including tardive dyskinesia, an involuntary movement disorder associated with antipsychotic medications used for schizophrenia and bipolar disorder. Because the system collects daily behavioral and symptom data, it can identify deterioration patterns weeks before they become clinically observable, enabling earlier treatment adjustment and reducing the risk of full patient disengagement.

How does care team integration improve behavioral health outcomes?

When a patient misses a dose, a real-time alert feeds to a daily intelligent action list for the care team — surfacing only patients who need intervention that day rather than creating alert fatigue across the full caseload. Clinicians can immediately initiate a check-in or treatment adjustment rather than discovering non-adherence weeks later during a scheduled appointment. Combined with visual AI side-effect monitoring, the care team has both the trigger and the clinical context needed to intervene before disengagement becomes irreversible dropout.