Healthcare Ops · Governed Messaging · MVP

To scale clinic communication, I made it harder to send a message.

As clinics grew, patient messaging turned into a quiet operational fire — inconsistent, costly, and risky. Everyone wanted a bulk-blast tool. I built the opposite: a governed system where every message needs a category and doctors get just 2 SMS a month. The constraint is what made it scale safely. No-shows dropped from 22% to 15%.

See the impact View the work
−32%
No-show rate (22→15%)
−40%
SMS cost
−70%
Setup time
−83%
Compliance flags
Scroll
clinic.app/broadcast/new New broadcast Every message follows the same governed path. 1Category 2Group 3Channel 4Timing Step 1 · Pick a category (required — no free-form blasts) Doctor Advice Important Alert Health Tips Out of Office Festival Offer & Discount General Message preview DOCTOR ADVICE Reminder: take your T2DM medication with breakfast. Walk 20 mins daily. Next check-up in 2 weeks. SMS this month 1/2 Hard limit protects cost & trust Targeted group T2DM · low adherence 312 patients match RELEVANT · LOW NOISE

The broadcast setup — category first, a live SMS counter, and smart targeting built in.

01 — The Trap

Growth turned patient messaging into a quiet operational fire.

Everyone was sending. No one was governing.

As clinics scaled, doctors and staff had to handle a flood of patient communication — schedule changes, reminders, follow-ups, closures, festival wishes, offers, health advice. It was all ad-hoc, manual, and uncontrolled. Each person messaged in their own tone, on their own timing, with their own wording. That variance looked harmless. It wasn't.

The organization thought it needed a bulk messaging tool. But more sending power aimed at an ungoverned process just scales the chaos. Hiring more staff scales cost, not consistency. Free-text tools amplify legal risk. Unlimited sends breed spam. Every incremental fix made the real problem worse.
02 — The Insight

This wasn't a messaging feature. It was a trust-system failure hidden inside growth.

Discovery made the root cause clear: 8 doctor interviews, 5 with clinic assistants, workflow walkthroughs across 3 clinics, and an analysis of 1,200 appointments over six months. The findings pointed in one direction — the problem was structural, not a missing feature.

The synthesis that reframed the brief: 90% of the risk came from unstructured process, not from a lack of capability. Governance had to come before features. The fix wasn't a more powerful send button — it was a system that made bad sends impossible.

Doctors wanted speed and safety, not options

75% of doctors prioritized speed over flexibility — "I need quick sends, under two minutes." They didn't want a rich composer. They wanted safe defaults and almost no decisions. Every choice you ask a time-pressed doctor to make is a chance to get tone, timing or compliance wrong.

The counterintuitive move: constrain to scale

The instinct was to add capability. Research said subtract it. 80% of assistants made repetitive errors for lack of templates; 30% of sends were already excessive. So the design bet was to limit — locked categories, reusable templates, a hard SMS cap — because constraint is what makes communication safe at scale. Patients proved the point: relevant, timed messages got 65% higher response than generic blasts.

The shift: from uncontrolled variance to governed structure BEFORE · AD-HOC "clinic closed tmrw" "PLS COME FOR CHECKUP!!" "offer 20% this week only" "reminder reminder" "happy diwali + book now" "u missed appt again" Inconsistent tone · no limits · 12 compliance flags/mo AFTER · GOVERNED DOCTOR ADVICE Medication reminder · T2DM group IMPORTANT ALERT Clinic closure · all patients HEALTH TIPS Seasonal care · push only FESTIVAL Greeting · scheduled Every message categorized · limited · targeted · 2 flags/mo
FIG 01 The core reframe, made visible: the same communication need, moved from uncontrolled variance to a governed structure.
Two personas, one design mandate: protect them from over-communication Dr The Doctor PRIMARY · extremely time-poor "I don't want to worry about wording or timing." GOALImprove adherence without extra effort CONSTRAINTAlmost no time per task NEEDSafe defaults, minimal decisions → Drove: category-first defaults & under-2-min setup CA The Clinic Assistant SECONDARY · executes the sends "I need structure so I don't make mistakes." GOALManage communication efficiently PAINRewriting the same messages repeatedly NEEDReusable, approved templates → Drove: the template library & reuse system
FIG 02 Personas as decision drivers — each frustration maps directly to a feature I built, not a poster for its own sake.
8 DOCTOR INTERVIEWS·5 ASSISTANT SESSIONS·3 CLINIC WALKTHROUGHS·1,200 APPOINTMENTS ANALYZED·500+ MESSAGES AUDITED·3 ROUNDS USABILITY· 8 DOCTOR INTERVIEWS·5 ASSISTANT SESSIONS·3 CLINIC WALKTHROUGHS·1,200 APPOINTMENTS ANALYZED·500+ MESSAGES AUDITED·3 ROUNDS USABILITY·
03 — The Build

A governance-first architecture: structure before send.

The whole system rests on one rule: no message exists without a category. You can't reach a send button by accident. Every broadcast walks a strict, sequential path — category, then group, then channel, then timing — and each step quietly enforces compliance and relevance before the next unlocks.

1

Category-first entry

Eight locked types — Doctor Advice, Important Alert, Health Tips, Festival, Offer, Out of Office, Seasonal, General. Pick one before anything else. No free-form mass blast is possible.

2

Smart patient groups

Segment by ailment (e.g. T2DM), status and adherence level — so the right message reaches only the patients it's relevant to, cutting noise and lifting response.

3

The 2-SMS hard limit

Two SMS per doctor per month, full stop, with a live counter and a graceful limit-hit state. Push is unlimited. This one constraint controls cost, prevents spam, and forces intentional sends.

4

Templates & reuse

Reusable, approved, named templates so assistants stop rewriting the same message — eliminating the repetitive errors that wasted a fifth of their day.

Card sort → the category taxonomy Doctors & staff grouped 30+ real message types. Clusters became the eight locked categories. RAW MESSAGE TYPES (SORTED BY USERS) medication reminder dosage change clinic closed emergency notice diwali wishes seasonal flu tip 20% discount new service hydration advice follow-up due …30+ types, no consistent grouping RESULTING CATEGORIES Doctor Advice Important Alert Health Tips Festival Offer & Discount Out of Office Seasonal Greetings General RESULTING INFORMATION ARCHITECTURE Broadcast 1 Category 2 Group 3 Channel 4 Timing
FIG 03 The card sort that grounded the system. User clusters became the eight categories, which became the locked four-step IA.
One locked path · every message, every time 1 Category mandatory 2 Patient group smart segments 3 Channel SMS (2/mo) · push ∞ 4 Timing now or scheduled
FIG 04 The governed flow. Each step gates the next, so compliance and relevance are structural, not optional.
Designing the constraint to feel like guidance, not a wall 0 OF 2 USED 2 left Plenty of SMS left this month. Send freely. 1 OF 2 USED 1 left Last SMS this month — make it count, or use push. 2 OF 2 USED Limit reached No abandonment — we offer a path: Send via push instead →
FIG 05 The limit states. Round 3 testing added the counter, warnings and a push fallback — which eliminated abandonment when the cap was hit.
Three rounds: low-fi → mid-fi → high-fi Each round tested with doctors and resolved one specific problem. ROUND 1 · LOW-FI category-first entry Goal: validate category-before-send ROUND 2 · MID-FI Fix: complex filters → checkboxes/presets Result: 30% faster targeting (n=5 doctors) ROUND 3 · HIGH-FI DOCTOR ADVICE Medication reminder · T2DM SMS 1/2limit warning Send ✓ Added: counters, warnings, success toasts Result: zero abandonment on limit hit
FIG 06 The fidelity progression. Round 2 simplified targeting by 30%; Round 3's counters and fallback killed abandonment at the cap.
Design system: safety over flexibility Principles — safety over flexibility · defaults over decisions · clarity over customization. MESSAGE CARD DOCTOR ADVICE T2DM · 312 patients Consistent, scannable, always categorized. CATEGORY TAGS DOCTOR ADVICE ALERT HEALTH TIP FESTIVAL OFFER Color-coded by intent — tone is enforced by design. DELIVERY INDICATORS Sent · delivered Scheduled Pending review Blocked · over limit Status is never a mystery for a busy clinician. USAGE COUNTER 1/2 SMS used this month The limit, always visible — scarcity made tangible.
FIG 07 The component library. Each element encodes a principle — categories enforce tone, counters make the limit tangible, statuses remove guesswork.

The doctor's experience, before and after

BeforeAfter
Ad-hoc reminders relayed through staffPredefined templates, set up in seconds
Manual texting, error-proneSmart targeting to precise patient groups
Frequent missed follow-upsScheduled, automated outreach
Inconsistent, unlimited messagingControlled frequency via hard limits
04 — The Proof

The numbers moved — across every dimension that mattered.

Straight talk: these are measured over 3 months post-MVP, against a pre-project baseline of 1,200 appointments, with A/B testing (control vs. MVP) to confirm the adherence gains weren't coincidence. This is the rare case where the proof is operational and clean, not projected.

Measured, 3 months post-MVP

−32%
No-show rate, 22% → 15%. The headline outcome — and the one tied directly to relevant, timely outreach.
−40%
SMS cost, $150 → $90 per doctor. The 2-SMS cap paid for itself immediately.
−83%
Compliance flags, 12 → 2 a month. Locked categories made most violations structurally impossible.

And the operational dividend

−70%
Setup time per broadcast: 5–7 min → under 2 min
−40%
Staff coordination: 10 hrs/week → 6 hrs
+65%
Patient response to relevant vs. generic sends
Five metrics, one direction Darker bar = before · gradient bar = after No-show rate 22% 15% SMS cost / doctor $150 $90 Staff coordination 10 hrs 6 hrs Setup time 5–7 min <2 min Compliance flags / mo 12 before 2 after Every missed appointment costs a clinic ~$150–200 A 7-point no-show drop compounds fast across a full appointment book.
FIG 08 The full before/after. Constraint didn't cost engagement — it improved every operational and trust metric at once.
The result I'd stake the project on is the compliance drop — 12 flags to 2. It proves the core thesis: when you make bad sends structurally impossible, you don't need to police behaviour. The architecture does it for you.
05 — The Call I Had to Make

I had to convince stakeholders that limiting the product was the right move.

"

The team wanted unlimited bulk messaging. I argued for a hard cap — and had to prove that less sending meant more trust.

This was the hard part, and it wasn't a screen. Stakeholders were skeptical of limits — a 2-SMS cap sounds like crippling your own product. The instinct in a growth context is always "give users more power." But the research was unambiguous: 30% of existing sends were already excessive, compliance flags were piling up, and patients were tuning out. An unlimited tool would have scaled every one of those problems.

So I made the case with data, not opinion. I showed that relevant, limited messages outperformed blasts by 65%, that each compliance flag carried real regulatory and cost exposure, and that the cap would protect the engagement the team was trying to grow. The reframe landed: we weren't limiting communication, we were protecting its value. That turned skepticism into buy-in, and the constraint became the MVP's defining strength.

What I'd do differently

Introduce financial-literacy-style guidance for doctors earlier, so the value of the limit is obvious from day one rather than something they discover. And explore AI-assisted predictive targeting in the next iteration — early modelling suggests another 10–15% adherence lift by sending the right message to the right patient at the right moment.

The Role

Domain — Healthcare operations · doctor appointment & broadcast messaging
Role — UX Lead: product strategy, governance architecture
Timeline — 3 weeks, discovery to MVP release · system redesign, not feature design
Team — Me (UX Lead), 1 product designer, 1 product manager, with engineering & compliance stakeholders
What I owned — UX strategy, the patient-communication governance model, research with doctors and staff, information architecture, flow design, and the adoption & safety guardrails. UI detailing went to the designer; testing to the PM.