The TPM Model: Why Structure Changed Everything at Ada Health
How embedding clinical product thinking from day one unlocked safer, faster product delivery
This is Clinical Product Thinking đ§ , your weekly newsletter featuring practical tips, frameworks and strategies from the frontlines of clinical product.
Good afternoon friends, this is issue No. 021. This week, weâre chatting with Lorenzo Espinosa who led B2B product at Ada Health, on the clinical product model that worked for them.
Ada has spent years building an AI-powered symptom assessment used by millions of patients across healthcare systems, insurers and providers. What interested me wasnât just what they built, but how they structured clinical product to move fast without compromising safety.
Introducing the TPM model
Lorenzo explains that at Ada, product work was organised around standing cross-functional triads:
Tech (engineering leadership)
Product (product management)
Medical (clinical / clinical product)
In some teams, a designer was also part of the core group. In others, a clinical product person was split across two squads, depending on demand.
Crucially, this wasnât just a delivery-level construct. The same TechâProductâMedical setup existed at the leadership level, creating a consistent decision-making structure from strategy through to execution.
That meant no meaningful decision happened without all three perspectives in the room.
Mutual education, not alignment theatre
This way of working didnât click overnight.
It took time for teams to find their footing, but once they did, the value of true interdisciplinary work became obvious. A big part of that was mutual education.
Each function had to be explicit about what it optimised for:
Tech prioritised platform stability, quality and scalability
Product pushed for differentiation, speed and user-centred outcomes
Medical focused on safety, regulatory compliance and avoiding harm
Those incentives are in tension, and that tension wasnât a problem to be smoothed over. It was a feature.
Before this, many of those priorities were implicit. Decisions felt slow not because people disagreed, but because the trade-offs were never named.
With TPM, trade-offs became explicit. No single function could optimise unilaterally. Every major decision required negotiation across safety, value and feasibility.
Counterintuitively, that led to better decisions earlier, when changes were cheap and options were still open.
Pragmatism as a muscle
One of the biggest shifts TPM created was when clinical showed up.
Instead of product and engineering ideating in isolation, only to bring clinical in halfway through and discover fundamental issues, clinicians were involved from the very start.
Designers, engineers, product managers and clinical product worked together from inception. That alone prevented a huge amount of late-stage backtracking.
But structure wasnât enough. The deeper shift was in how teams approached delivery.
Early on, clinical feedback often landed in an all-or-nothing frame:
âThis cannot be done until everything is fully validated, verified and tested across all user demographics.â
That instinct is completely understandable. Itâs driven by a deep responsibility for patient safety.
But framed that way, teams often had to wait until a feature was âfully doneâ before launching anything at all, which, in practice, meant shipping very slowly.
As Lorenzo described, you simply canât operate that way and still build product.
Over time, Ada deliberately reframed the role of clinical product.
From approval to shaping the path
Instead of asking clinicians to approve or reject a fully formed solution, teams involved them in defining the safest sequence of delivery from the outset.
This wasnât about lowering the bar on safety.
It was about doing the right safety thinking earlier, so teams could agree what was acceptable to release now, and what could responsibly come later.
The practical question shifted from:
Can we ship this at all?
to:
What does this need to include to be clinically viable as a first release, given the risks and context?
That single reframing, from binary approval to phased responsibility, unlocked progress without compromising patient safety.
A concrete example: EHR integrations
Consider ingesting data from an Electronic Health Record (EHR).
All-or-nothing approach:
Donât launch until you can ingest the full dataset.
TPM approach:
Phase 1: Ingest the minimum clinically necessary data (defined by clinical product). Launch and learn.
Phase 2: Expand based on real-world usage.
Phase 3: Full payload integration.
Because clinical product was in the room from day one, âminimum clinically necessaryâ was defined before a single line of code was written.
Thatâs the difference structure makes.
Why this matters
The TPM model worked at Ada because it:
Embedded clinical product from day one
Made trade-offs explicit and negotiable
Built pragmatism into the system, not individual heroics
This wasnât about faster delivery at any cost.
It was about creating a structure where safety, value and feasibility could be negotiated early, when changes are cheap and options are still open.
Thatâs the real power of clinical product done well.
Tune in for Part 2 next week on the three mindset shifts CPMs need to excel in the TPM model, plus practical ways to apply them in your own teams.
Clinical Product Dinner â¨
đ 4th February - How Clinicians Can Turn Expertise Into Category-Defining Healthtech Products
With Dr Arun Notaney, Founder of GP Automate, exploring how lived clinical experience can be transformed into category-defining healthtech products, and how frontline frustration becomes scalable, NHS-ready solutions. đ Get your ticket here.
Thatâs the public post for this week. See you next time! đ
đ¤ Work with me | đ Attend an event | | âď¸ Send a message
Written by Dr.Louise Rix, Head of Clinical Product, doctor and ex-VC. Passionate about all things healthcare, healthtech and clinical product (âŚobviously). Based in London. You can find me on Linkedin.
Made with đ for better, safer HealthTech.
[New] Want to Go Deeper?
The two patterns that quietly derail TPM and access to new Clinical Product Thinking events in Q1.
The TPM model: the two failure modes that quietly kill it, and how to avoid them
If you take one thing from Adaâs TPM model, let it be this:
