Context
Oliver Kharraz, founder and CEO of Zocdoc, explains why healthcare access still feels broken even though the industry has spent years building provider directories, EHRs, and patient portals. His argument is that the hardest part was never simply listing doctors. It was making provider information accurate enough, and scheduling infrastructure connected enough, that a patient could actually complete a booking. This matters to PMs because it is a strong lesson in infrastructure products, marketplace design, and what changes when AI becomes a new interface layer rather than the system doing the real work.
The Big Idea
AI will improve healthcare access only if the underlying scheduling and availability infrastructure is already reliable enough to turn intent into a real appointment.
Zocdoc's core claim is that better search, recommendations, and conversational agents are useful, but they do not solve access by themselves. The real product advantage comes from owning the hard transaction layer that connects patients, providers, insurers, and scheduling systems well enough to make booking actually work.
Key Insights
Directories fail without transactions
- What: Kharraz argues that static provider directories stay inaccurate because they behave like published reference books. By the time they are exposed to patients, the data is already drifting.
- Why it matters: A product that only displays information has weak feedback loops. A product that completes transactions gets continuous validation from real usage.
- Example: He contrasts a directory with an appointment booking flow. Every successful or failed booking teaches Zocdoc whether the doctor, schedule, and insurance mapping were correct. In his framing, the shift is from an "encyclopedia" model to a Wikipedia model.
Scheduling accuracy took years
The main barrier was not a lack of consumer demand. It was the operational mess underneath. Zocdoc built integrations for 175-plus EHRs, but Kharraz says the integration itself was only the start because each practice used its scheduling system differently. His story about an office manager using a Post-it note instead of updating the EHR captures the core product lesson: source systems are often only partially true. If a PM assumes the system of record fully reflects reality, the product will fail in production even if the integration technically works.
Consumer alignment changed the design
Kharraz believes Zocdoc solved more of this problem because it stayed anchored on the patient rather than on a single institution. His view is that many healthcare products assume one company should "own the front door" to care. He rejects that premise. Patients move across urgent care, specialists, chronic care, and rare-disease journeys based on need, so the product has to work across many entry points rather than trap the user inside one system.
For PMs, the broader lesson is that market structure should shape product architecture. If user behavior is naturally multi-home, a closed experience may feel strategically appealing but will fight the actual job users are trying to get done.
Loyalty follows good care
One reason healthcare companies justify closed ecosystems is the fear that open discovery will reduce continuity. Kharraz argues Zocdoc's data shows the opposite.
- Patients do compare multiple providers, read reviews, and spend real time choosing.
- Once patients choose a provider and stay within the same specialty, Zocdoc sees a 75 to 80% rebooking rate with that same provider.
- His implication is that loyalty does not need to be forced through product lock-in. If the care experience is good, loyalty emerges after the first visit.
That is a useful PM distinction between acquisition friction and retention value. Products often add friction to protect retention when they should improve acquisition and let quality drive repeat behavior.
AI creates more front doors
Kharraz sees AI as a demand multiplier, not the full solution. AI agents can help users express symptoms, preferences, and insurance needs more naturally. They can also make recommendations feel more personalized and more actionable. But once the user decides to book, the agent still needs reliable infrastructure behind it.
His strategic point is that AI does not consolidate demand into one winner. It may do the opposite. More agents, more interfaces, and more surfaces mean more "front doors" into healthcare. That is why Zocdoc is opening its scheduling layer to AI platforms, payers, review sites, search sites, and other consumer platforms instead of insisting every booking must happen inside the Zocdoc app.
Last-minute capacity is the unlock
One of the strongest product insights in the episode is that access problems are partly a utilization problem, not only a supply problem.
- National average wait time to see a primary care doctor was about 17 days when Zocdoc started.
- Kharraz says that average is now 31 days.
- On Zocdoc, patients can typically see a doctor in three days or less.
His explanation is that doctors have cancellations and unused slots that traditional workflows fail to surface. Zocdoc captures that last-minute capacity and turns it into bookable supply. He compares this to air travel: airlines once flew at about 60% load factors, then online travel distribution pushed utilization above 90%. In the wild, he says doctors still run at around 70% load factors, which means access can improve a lot before the industry even adds more clinical supply.
Fast availability is not low quality
Kharraz pushes back on a common user assumption that quick appointments must mean lower-quality doctors. He says highly sought-after physicians can still have last-minute openings because their patients often have busy lives and cancel. His example was a dean at a New York academic medical center who thought being booked three months out meant Zocdoc was unnecessary. After staying on the platform for another 30 days, that physician still booked around 40 patients through last-minute openings that otherwise would have gone unused.
The product lesson is that users often misread marketplace signals. PMs should test whether apparent scarcity actually reflects demand, or whether it reflects poor inventory visibility.
Pricing lowered provider friction
Zocdoc's business model evolved because a flat subscription did not match value delivered. Some doctors got many bookings, others got few, and everyone paid the same amount. The company moved to charging providers for discovery bookings, meaning appointments generated from a general search where Zocdoc matched patient to provider.
Kharraz says Zocdoc does not charge when a patient is already looking for a specific doctor and Zocdoc is only facilitating scheduling. He frames that as software that should be free. The PM lesson is that pricing can be a product design choice. Charging for differentiated value while removing cost from commodity workflow can increase supply-side participation and reduce onboarding friction.
Mental Models & Frameworks
Encyclopedia versus Wikipedia data
- Encyclopedia model: publish a provider directory periodically and accept that it starts aging immediately.
- Wikipedia model: let every real booking act as a validation event that improves the underlying data.
- When to use it: use this model whenever a product depends on facts that change often and where a transaction can confirm or deny the facts.
- Why it matters: the best way to improve data quality may be to design more feedback into the product, not to launch a bigger data-cleaning project.
Front doors should multiply
Kharraz's view is that healthcare should not have one canonical entry point. Patients start from insurers, search engines, AI agents, review sites, provider homepages, and marketplace apps. A PM can use this as a platform strategy test: if users naturally begin the journey in many places, the winning product may be the infrastructure that makes all those doors work, not the destination that tries to own every visit.
Utilization beats pure supply
This is a useful marketplace lens for access problems.
- Question: is the market truly short on supply, or is available supply hidden, fragmented, or hard to route correctly?
- Mechanism: expose cancellations, open slots, and underused inventory in a form that users can trust and book.
- Example: Zocdoc claims average access of three days or less while the broader system averages 31 days for primary care, largely by surfacing capacity that already exists.
Patient choice, provider loyalty
This model separates two phases that many products blur together:
- Discovery phase: maximize patient choice, transparency, and ease of comparison.
- Care phase: let provider quality create repeat usage and loyalty.
Use this when a team is debating whether to restrict discovery in order to protect downstream retention. Kharraz's argument is that friction at discovery often hurts more than it helps.
Decision Principles
Principle: Build the transaction layer
- When: a product category is full of directories, search, or recommendations that users still do not trust.
- Why: a transaction creates a stronger feedback loop than a listing page. If the booking, purchase, or handoff succeeds, the data gets verified. If it fails, the product learns exactly where reality and system state diverged.
Principle: Follow user entry points
- When: users start journeys across many external channels instead of one owned surface.
- Why: trying to force a single front door can conflict with actual behavior. Platform reach may matter more than owning the full UI when demand is inherently distributed.
Principle: Price only unique value
- When: a marketplace or SaaS product helps some users discover value and helps others only complete a workflow.
- Why: charging for commodity workflow can slow adoption. Zocdoc's choice to charge for discovery bookings, but not for scheduling a patient who already chose a specific doctor, is a good example of pricing based on differentiated value.
Trade-offs & Nuance
Open access versus coordination
Julie Yoo presses on a real tension: if no single company owns the front door, care can feel more accessible, but coordination may become harder. Kharraz's response is that patients already behave this way, so product strategy should start from actual behavior rather than from an idealized closed system. Still, the tension remains. A more open discovery layer helps access, while longitudinal care still requires coordination across providers and settings.
Choice versus routing control
Kharraz says Zocdoc historically gives agency to the patient by showing available options and letting the user choose. In an AI-agent world, that choice may be delegated to software acting on the patient's stated preferences. The trade-off is clear: more automation can reduce effort, but it also shifts decision power into ranking and routing logic that users may not see.
Broad coverage versus qualification
Zocdoc wants broad provider coverage across 250 specialties, including highly specialized care such as oncology. But Kharraz also says the marketplace is more selective than a typical state board and removes providers who do not meet its standards. The tension is that larger supply improves access, while tighter curation improves trust. A marketplace PM needs to manage both at the same time.
Destination versus infrastructure
Zocdoc still has a consumer app, but Kharraz now frames the company more as healthcare access infrastructure than as a single destination site. The upside is broader distribution through payers, AI platforms, provider websites, and other channels. The downside is less direct control over user experience, ranking, and brand touchpoints. This is a classic platform trade-off between reach and control.
Practical Application
Audit your hidden workflow gaps
Pick one critical workflow in your product and map where the system of record can drift from reality.
- Review not only the official fields and integrations.
- Interview frontline users about off-system workarounds such as spreadsheets, sticky notes, side chats, or verbal handoffs.
- Treat every unofficial workaround as a product requirement, not as user error.
Kharraz's Post-it note story is a reminder that a product fails at the edges first.
Find transaction-based feedback loops
If your product depends on accurate data, ask how user actions could continuously verify that data.
- A booking can validate availability.
- A completed onboarding step can validate identity data.
- A successful handoff can validate routing rules.
The goal is to stop treating data quality as a periodic clean-up project and start making it a byproduct of product usage.
Surface underused capacity first
Before proposing a big supply-side expansion, test whether the main problem is actually poor visibility into existing capacity.
- Measure cancellations, no-shows, and empty slots.
- Expose that inventory in a way users can act on quickly.
- Track whether faster matching reduces wait times without adding new supply.
This is especially useful in marketplaces, staffing products, scheduling tools, and service businesses.
Revisit your pricing boundary
List the moments where your product creates real incremental value versus moments where it only facilitates a workflow the user already intended to complete.
Then ask whether pricing matches that distinction. Zocdoc's split between discovery bookings and specific-doctor booking is a strong example of how a pricing model can lower adoption friction while preserving monetization where the platform actually creates demand.
Design for many front doors
If AI agents, partners, or aggregators are becoming meaningful traffic sources, do not treat them as edge cases.
- Define which parts of the experience should be embeddable or API-first.
- Decide what external surfaces need access to your availability, inventory, or workflow engine.
- Preserve trust rules and quality controls even when the UI is not yours.
This matters any time the interface layer is fragmenting faster than the core system.
Questions to Consider
Real data or static listing
Is your product still acting like a static directory that slowly drifts out of date, or does each real user transaction help verify and improve the underlying data?
Hidden supply or true shortage
Where might your team be treating an access problem as a pure supply shortage when the bigger issue is hidden capacity, poor routing, or low visibility into last-minute availability?
Closed flow or open network
Are you designing as if one company should own the full user journey, even though users actually start from many different channels, partners, or AI assistants?
Pricing aligned with value
Does your pricing model charge for the unique value your product creates, or are you also charging for low-differentiation workflow steps that may only create adoption friction?
Bottom Line
Zocdoc's lesson is that fixing access requires more than better search or smarter AI. The real breakthrough is building a transaction layer that makes provider data accurate, turns hidden capacity into bookable supply, and works no matter where the patient starts the journey.
Tools & Products
| Tool / Product | What it does | Why it was mentioned |
|---|---|---|
| Zocdoc | Doctor discovery and scheduling platform | Kharraz explains how it evolved from a consumer destination into healthcare access infrastructure that other platforms can use. |
| Gemini | AI platform and agent interface | Zocdoc announced a partnership so users can move from AI-assisted discovery to actual appointment booking. |
| EHR systems | Clinical and practice software systems | Zocdoc built integrations into more than 175 EHRs because provider availability and practice data live there. |
| Blue Shield of California member directory | Payer-facing provider directory | Kharraz uses it as an example of embedding Zocdoc booking directly inside an insurer's member experience. |
People to Follow
Oliver Kharraz
Founder and CEO of Zocdoc. He is relevant for PMs interested in healthcare marketplaces, infrastructure strategy, and the operational realities behind making healthcare scheduling actually work.
Julie Yoo
General Partner at a16z and the episode host. She brings an operator-investor view on healthcare infrastructure, provider directories, scheduling, and the trade-offs between consumer choice and care coordination.
Notable Quotes
"When you put out a provider directory, frequently you're in the business of publishing the encyclopedia Britannica." (Oliver Kharraz)
"Once you have the transaction, it's much more like Wikipedia." (Oliver Kharraz)
"Healthcare isn't like a building where you lure the patient in and you slam the door shut behind them. It's a lot more like a city." (Oliver Kharraz)
"We don't need every booking to happen on Zocdoc.com or our app. We just want to make sure that whatever door the patient is knocking on, that door actually opens." (Oliver Kharraz)
