Growing in Regulated Markets: SurpassMD’s Vivek Girotra on Privacy, AI, and Better Signals
Signal engineering. Creative as targeting. AI guardrails. Better measurement. Agents doing the work.
If you thought growth was already complicated, try growing in regulated markets.
Fintech. Healthcare. Insurance. Financial services.
You can’t always target the audience you want. You can’t always send the events you want. You can’t make whatever claim you want in an ad. And sometimes, the thing you ultimately care about happens way too far down the funnel for Meta or Google to optimize toward it.
So… how do you grow?
That’s exactly what we dug into on this episode of App Masters with Vivek Girotra, co-founder and CEO of SurpassMD, a metabolic health telehealth company providing cash-pay GLP-1 care today, with a longer-term vision spanning peptides and longevity care. Independent licensed clinicians review patient intakes, and licensed pharmacies dispense prescribed medication.
Vivek has spent 15+ years working across growth, gaming, fintech, and now healthcare.
And he’s learned a thing or two about growing businesses where the normal playbook doesn’t always work.
His biggest takeaway?
When the obvious growth levers disappear, you have to get much smarter about the signals you send back to the platforms.
Here are some of his biggest growth moves.
1. Turn Customer Data Into Better Signals
Let’s start with one of the most interesting concepts from the conversation:
Signal engineering.
Or, as Vivek puts it, using the information you already have to create better feedback signals for ad platforms.
Here’s the problem.
In regulated industries, platforms increasingly limit how much you can target.
In fintech, for example, you may not be able to use certain lookalike audiences, age or gender targeting, ZIP codes, or highly specific interests.
So how do you tell Meta:
“Hey, these are the people I actually want more of?”
You teach the algorithm through your conversion signals.
Vivek gave an example from his time at Abound, a fintech company serving Indian expats in the U.S.
The funnel looked something like this:
Install → Sign Up → KYC → Bank Link → First Transfer
The problem?
Optimizing for sign-ups was too shallow.
Optimizing for first transfer was too deep.
There were too many high-friction steps between the two.
So Vivek’s team created an intermediate signal.
After someone signed up, they enriched the data internally and determined whether the user was likely to belong to their core audience.
If a new user met their criteria, they sent a qualified signal back to the ad platform.
Now the algorithm wasn’t simply learning:
“Find me people who sign up.”
It was getting a much better indication of:
“Find me people who look like the users who eventually become valuable customers.”
That’s a massive difference.
The three rules for a good signal
Vivek highlighted three things a useful optimization signal needs:
- Volume
You need enough events for the platform to actually learn from them.
- Recency
Ideally, the signal happens relatively early — often within the first 24–48 hours.
- Correlation
Most importantly, the event needs to be a good proxy for the outcome you ultimately care about.
The lesson?
Don’t automatically optimize for the easiest event to generate.
Find the event that gives the algorithm the best prediction of future value.
2. Sometimes You Need to Delay the Signal
This is where things get even more interesting.
Vivek and I had talked about this years ago around subscription trials.
Imagine you have a 7-day free trial.
The obvious optimization event is:
“Trial Started.”
But what if a huge percentage of those users cancel almost immediately?
In one example Vivek shared from a subscription app, around 25–30% of trial users were cancelling within the first 12 hours.
So instead of immediately telling Meta:
“This person started a trial. Find me more people like them.”
The team waited.
They created a better signal that essentially said:
“This person started a trial and made it past the point where most low-quality users dropped off.”
That small change made the signal much more valuable.
The broader lesson:
Your first conversion event isn’t always your best conversion event.
Sometimes the smartest thing you can do is wait.
You might have:
Install → Trial → 12 hours → Subscription
Instead of sending the platform the trial event immediately, you could potentially create a more qualified event after that critical drop-off period.
Of course, the exact implementation depends on the product, platform, privacy requirements, and data volume.
But the principle is powerful:
Don’t just measure what happened. Measure what predicts what happens next.
3. Creative Is the New Targeting
If you’ve been following mobile advertising, you’ve probably heard this before:
Creative is the new targeting.
But Vivek takes it a step further.
Creative variety doesn’t mean:
- Change the background
- Move the CTA
- Try a different aspect ratio
- Swap a headline
That’s not real creative diversity.
You need different ideas for different people.
Think about an online weight-loss program.
“People who want to lose weight” is not a useful creative brief.
Instead, you might have:
The busy parent
They’ve got work, kids, and almost no time for themselves.
Or:
The person who’s tried everything
They’ve tried diets, workouts, apps, and programs — and they’re frustrated that nothing has stuck.
Or:
The health-conscious 40-something
They’re less focused on appearance and more focused on staying healthy, mobile, and active for the next 20 years.
Same product.
Completely different motivations.
That’s where creative gets powerful.
Vivek’s recommendation?
Get almost “comically granular” with your ICP.
Don’t just define demographics.
Define:
- Their situation
- Their motivation
- Their pain
- Their aspirations
- Their objections
- Their desired outcome
Then build creative around those insights.
4. Stop Listening to What People Say. Watch What They Do.
One of my favorite parts of this conversation was Vivek’s point about stated vs. revealed preferences.
Customers aren’t always great at telling you what they want.
Their behavior is often a much better source of truth.
Vivek gave a great example:
The Kindle isn’t necessarily selling you a more convenient way to read.
It’s selling you the identity of being someone who reads.
That’s an important distinction.
Because the thing your customer buys isn’t always the thing your customer really wants.
At SurpassMD, cash-pay prices are only part of the story. Vivek treats the entire patient journey as the product, putting real thought into what patients need at each step: clear information before they begin, a smooth path through clinical review and fulfillment, and help when questions come up.
5. Build AI Guardrails Before You Scale
Now we get into the really interesting stuff.
Vivek is currently building SurpassMD, backed by top angel investors in Silicon Valley, in a category where the cost of getting marketing wrong is significantly higher.
You can’t just launch an ad and hope for the best.
There are:
- Platform policies
- Healthcare regulations
- Privacy requirements
- Advertising restrictions
- Prescription drug rules
- Testimonial restrictions
- Data boundaries
- Operational requirements
And one bad claim can get an ad rejected — or create a much bigger problem.
So Vivek built an AI-powered review framework around three layers.
Layer 1: Platform Guardrails
Does Meta, Google, or another channel allow this?
Layer 2: Category Guardrails
Is the claim actually permitted for this particular industry or product?
Layer 3: Operating Reality
Can the company actually deliver what the ad promises?
For SurpassMD, those checks also include the requirements tied to its LegitScript certification, alongside platform policy and healthcare advertising rules.
For example:
If your ad says:
“Same-day shipping.”
But your operation doesn’t offer same-day shipping…
You have a problem.
AI can act as a first-pass review.
Before an ad goes live, have an AI system review the copy, claims, creative, and messaging against your predefined rules.
A human still makes the final call.
It helps the human catch problems before they become expensive problems.
6. Your Measurement Stack Has to Match How People Actually Buy
Another major problem in modern app growth?
The customer journey isn’t always:
Ad → App Store → Install → Purchase
Especially in fintech and healthcare.
Someone might:
See an ad → Visit your website → Download your app → Complete something on the web → Return to the app
Or:
Start on web → Download app → Finish onboarding → Complete checkout on web
Good luck measuring that with a simplistic install-to-purchase setup.
Vivek’s advice is to build a measurement ecosystem that can actually connect these journeys.
That means thinking about:
- Web-to-app
- App-to-web
- Deep linking
- Attribution
- First-party data
- SDKs
- Lifecycle events
- Privacy boundaries
And healthcare adds another layer.
You have to be extremely careful about PHI — protected health information.
Patient information cannot simply flow into your advertising systems.
So Vivek argues for separating the patient lifecycle from the marketing lifecycle and creating controlled data boundaries between them.
The goal isn’t just better attribution.
It’s better attribution without breaking the rules.
7. Welcome to the Golden Age of Marketing
And then we get to the favorite part.
AI agents.
Vivek believes we’re entering what he calls a golden age of marketing.
Why?
Because things that previously required:
- A developer
- An analyst
- A researcher
- Multiple tools
- Hours of work
can increasingly be handled by an AI agent.
Here’s a crazy example.
You could tell an agent:
Go through five competitors.
Review their publicly accessible onboarding steps.
Compare them with ours.
Identify where they’re reducing friction.
Tell me where we’re better.
Give me your five biggest recommendations.
And the agent can actually go do the work.
That’s very different from asking ChatGPT:
“Give me five ideas to improve my onboarding.”
You’re moving from AI as an advisor to AI as an operator.
And this gets even more interesting when you connect the agent to your real data.
Your:
- Analytics
- App data
- Website data
- Session recordings
- Heatmaps
- Conversion funnels
Now the AI isn’t just giving you opinions.
It’s analyzing what your actual users are doing.
And that’s where Vivek sees the biggest opportunity.
8. Don’t Overthink Prompt Engineering
Vivek shared a surprisingly simple AI tip:
Just talk to it.
His approach to prompt engineering?
Voice-to-text.
He starts with detailed voice notes.
He explains what he’s trying to accomplish, gives the AI context, and lets the model figure out the structure.
And honestly…
This might be one of the best pieces of AI advice we’ve heard recently.
You don’t need to become a prompt engineer.
You need to become good at explaining problems.
The better you can describe:
What you’re trying to do → what you know → what you’ve tried → what you’re worried about → what success looks like
…the better AI can help.
9. When Should You Start Spending on Brand?
We also got into a question every growth team eventually faces:
When do you stop focusing entirely on performance and start investing in brand?
Vivek shared a simple three-stage framework.
Stage 1: Explore
You’re trying to find product-market fit.
Don’t overinvest in broad brand campaigns before finding a working offer and channel.
Find a cohort.
Find a channel.
Figure out your unit economics.
Get from 0 → 1.
Stage 2: Expand
You found something that works.
Now go from:
1 → 10.
If Facebook works, test Google.
Try Apple Ads.
Expand channels.
Improve your funnel.
Hire help.
Scale what you’ve proven.
Stage 3: Extract
Eventually, your direct-response channels start getting saturated.
Your low-hanging fruit is disappearing.
That’s when brand starts becoming more important.
Maybe that’s CTV.
Maybe it’s YouTube.
Maybe it’s creators.
Maybe it’s broader brand campaigns.
The key is that brand doesn’t have to replace performance.
Think of it more like a step function:
Build brand → capture demand → build more brand → capture more demand.
10. The Biggest Lesson: Better Growth Comes From Better Feedback Loops
If you take nothing else from this episode, take this.
Growth isn’t just:
Run ads → get users → increase budget.
The best growth teams build feedback loops.
You:
Create → Measure → Learn → Improve → Feed better signals → Create again.
And now AI can accelerate almost every step.
Vivek’s five-part framework is a great way to think about it:
1. Engineer better signals
Give ad platforms better information about who actually becomes valuable.
2. Build a creative engine
Treat creative as targeting.
3. Create AI guardrails
Catch compliance and messaging problems before they become expensive.
4. Build the right measurement stack
Measure the actual customer journey — not the journey you wish customers took.
5. Put AI agents to work
Don’t just ask AI for ideas.
Give it a job.
Biggest Takeaway
One of the biggest themes we’ve been talking about at App Masters is that the old growth playbook is changing.
Algorithms are getting smarter.
Targeting is getting broader.
Privacy restrictions are increasing.
Creative is becoming more important.
And AI is moving from something you chat with to something that can actually do the work.
Vivek’s experience across gaming, fintech, and healthcare makes this conversation especially interesting because the more regulated the industry gets, the more creative your growth strategy has to become.
And that’s the irony:
More restrictions don’t necessarily kill growth.
They force you to become better at it.
So if you’re working in fintech, healthcare, subscriptions, or any other regulated category, don’t just ask:
“What can I target?”
Start asking:
“What signals can I create?”
“What does my best customer actually look like?”
“What does their behavior tell me?”
“What can AI analyze for me?”
And most importantly:
“How can I build a better feedback loop?”
That’s where the next generation of growth is going to come from.
Watch the Full Episode
Want to hear Vivek break down these strategies himself, including signal engineering, Meta optimization, creative targeting, AI agents, compliance, measurement, and building SurpassMD?
Watch the full conversation here:
And if you’re an app founder trying to figure out what’s actually working right now, subscribe to the App Masters YouTube channel and join the App Founders Community.
Because the best growth strategies aren’t sitting in a textbook.
They’re being tested right now.
