You are probably optimizing the wrong step
Most GLP-1 teams have a conversion problem they cannot see clearly.
They know revenue. They know ad spend. They know how many patients started this month. What they usually cannot answer in one sentence is: "Of 1,000 people who clicked our ad yesterday, exactly how many made it to each step, and where did the largest group fall out?"
Without that, optimization becomes guesswork. A team spends a sprint redesigning the landing page hero when the real leak was a photo-upload step that loses 40% of qualified patients. The landing page felt visible. The upload step was buried.
The fix is not another tactic. It is instrumentation.
When you can see every step, the biggest leak is usually obvious, and it is usually not the step you were arguing about.
This post is the diagnostic. How to map the funnel, fire the right events, do the math, find the leak that matters, and rank fixes by money instead of opinion.
The real GLP-1 funnel has more steps than you think
Most teams model four or five steps: click, intake, checkout, started.
The real funnel has more, and the steps teams skip are often where the money leaks.
| # | Stage | Event to fire | Why it matters |
|---|---|---|---|
| 1 | Ad or organic click | landing_view | Top of funnel, source attribution |
| 2 | Eligibility quiz start | quiz_start | First micro-commitment |
| 3 | Eligibility result | quiz_complete | Qualified vs not |
| 4 | Intake start | intake_start | Real intent |
| 5 | Intake sensitive section | intake_sensitive_view | Privacy and trust friction |
| 6 | Photo or lab upload | upload_complete | High-friction, often-skipped leak |
| 7 | Intake complete | intake_complete | Ready for checkout |
| 8 | Checkout view | checkout_view | Price encounter |
| 9 | Payment attempt | payment_attempt | Intent to pay |
| 10 | Payment success | payment_success | The sale, but not the outcome |
| 11 | Provider approval | rx_approved | Where async programs stall |
| 12 | First dose or shipment | first_fulfillment | The real conversion event |
Two things stand out.
First, "payment success" is not the end. In GLP-1, the patient has not really converted until medication is approved and shipped. A program with great checkout conversion and slow provider review can still hemorrhage refunds between step 10 and step 12.
Second, the upload step (6) and the sensitive-intake step (5) are invisible in most dashboards, yet they are common high-drop points. You cannot fix a leak you do not measure.
For the broader stack that connects these systems, see Telehealth Growth Stack: How to Connect Ads, Intake, CRM, Billing, and EHR.
Step one: fire a clean event at every stage
You cannot analyze what you do not capture. The foundation is a consistent event at each stage with consistent properties.
Every event should carry:
- a stable anonymous ID and, once known, the patient ID
- a timestamp
- the traffic source, medium, and campaign
- the device type
- the program and product interest (for example GLP-1, injection vs oral)
- the stage name
A few rules that prevent months of bad data:
- Use one naming convention and never improvise.
intake_start, notIntake Startedin one place andstart_intakein another. - Fire events server-side where possible. Client-only tracking loses 10 to 30% of events to ad blockers and mobile reliability, and mobile is where most GLP-1 traffic lives.
- Stamp the source on the first event and carry it through. Otherwise you cannot break conversion down by channel.
- De-duplicate. A double-fired
payment_successwill quietly inflate your best number.
If you use a customer data platform, this is exactly what it is for. If you do not, a disciplined event schema in your own database works.
Step two: build the funnel table
Once events flow, build the simplest possible table: how many unique people reached each stage, in a fixed window, for a fixed cohort.
Here is a worked example. Numbers are illustrative.
| Stage | People | Step conversion | Drop-off | Cumulative |
|---|---|---|---|---|
| Landing view | 10,000 | - | - | 100% |
| Quiz start | 6,000 | 60% | 4,000 | 60% |
| Quiz complete | 4,800 | 80% | 1,200 | 48% |
| Intake start | 3,840 | 80% | 960 | 38.4% |
| Sensitive section | 3,072 | 80% | 768 | 30.7% |
| Upload complete | 1,843 | 60% | 1,229 | 18.4% |
| Intake complete | 1,659 | 90% | 184 | 16.6% |
| Checkout view | 1,492 | 90% | 167 | 14.9% |
| Payment attempt | 1,045 | 70% | 447 | 10.5% |
| Payment success | 982 | 94% | 63 | 9.8% |
| Rx approved | 884 | 90% | 98 | 8.8% |
| First fulfillment | 836 | 95% | 48 | 8.4% |
Read the "step conversion" column, not the cumulative one. The cumulative number always shrinks, which makes the last steps look worst even when they are healthy.
In this example the worst step conversions are:
- Upload complete: 60% (1,229 people lost)
- Quiz start: 60% (4,000 people lost, but this is cold traffic)
- Payment attempt: 70% (447 people lost)
The upload step loses 1,229 highly qualified people who already completed sensitive intake. That is the leak. Not the hero image.
Step three: weight the leak by value, not volume
The biggest drop-off by count is not always the biggest opportunity. You have to weight by how qualified and how valuable the lost people are.
A useful way to think about it: estimate the revenue recovered if you improved each step by a realistic amount.
| Stage | People lost | Lead quality at this stage | Realistic recoverable | Value if recovered |
|---|---|---|---|---|
| Quiz start | 4,000 | Low (cold clicks) | Small % | Low per head |
| Upload complete | 1,229 | High (finished sensitive intake) | Large % | High per head |
| Payment attempt | 447 | Very high (chose to pay) | Medium % | Highest per head |
The math usually points to two places:
- Late-funnel, high-intent steps where even a small percentage recovery is pure money (payment attempt)
- Mid-funnel, high-friction steps where a large group of qualified people is stuck (upload)
Cold top-of-funnel drop-off matters too, but it is a traffic-quality and messaging problem, often cheaper to fix by changing the ad and audience than by reworking the funnel.
A simple prioritization formula:
opportunity = people_lost_at_step
x realistic_recovery_rate
x downstream_conversion_to_paid
x average_revenue_per_patient
Rank every step by opportunity. Fix the top of that list first. This single calculation ends most internal debates because it replaces opinion with a dollar figure.
Step four: diagnose the leak before you fix it
Once you know the step, find out why people leave it. The funnel tells you where, not why.
Tools for the "why":
- Session replays on the failing step (with PHI masking)
- Field-level analytics inside intake (which question is the last one touched before exit)
- Device and browser breakdowns (a leak that is 80% mobile is a mobile problem)
- Time-on-step (a 3-minute median on a 20-second step signals confusion)
- Exit surveys on the failing step, kept to one question
- Support and chat logs filtered to that step
For the upload example, replays might show the patient does not have a lab handy, or the mobile uploader fails on large photos, or the instructions are unclear. Each of those is a different fix.
Diagnosing before fixing is what separates a real improvement from a redesign that moves the leak somewhere else.
For the intake-specific version of this, see Weight Loss Intake Form Scorecard: 12 UX Checks Before Launch.
Step five: fix, test, and confirm with an experiment
A fix you cannot measure is a guess with extra steps.
- Change one thing at a time on the failing step.
- Run an A/B test where volume allows, not a before-and-after, which confounds with seasonality and traffic mix.
- Measure the step conversion AND the downstream paid conversion. A change that lifts intake completion but lowers paid conversion may be letting in unqualified patients.
- Watch refund rate of the new cohort. Pushing more people through is only good if they stay.
- Confirm the leak actually moved, then re-rank and attack the next one.
For the testing discipline, see A/B Testing Telehealth Landing Pages and Intake and A/B Test Intake Forms Without Breaking Clinical Ops.
The segmentation that reveals hidden leaks
A single funnel hides leaks that only show up in segments. Always break the funnel down by:
| Segment | What it reveals |
|---|---|
| Device (mobile vs desktop) | Mobile-specific friction, usually the biggest hidden leak |
| Traffic source and campaign | Which channels send people who actually convert |
| Product interest (injection vs oral) | Whether one product path is leakier |
| New vs returning | Whether you are losing first-timers or repeat shoppers |
| Geography and state | State-specific availability or licensing drop-offs |
| Time of day and day of week | Whether provider review lag correlates with abandonment |
A funnel that looks healthy in aggregate can hide a mobile checkout that converts at half the desktop rate. Since most GLP-1 paid traffic is mobile, that single segment cut often finds the real money.
For the source-to-care view, see Telemedicine Marketing Funnel: From Paid Click to Started Care.
The post-payment funnel most teams ignore
Front-funnel conversion gets all the attention. In GLP-1, the post-payment funnel is where a surprising amount of revenue leaks back out.
After payment_success, you still have to clear:
- provider review and approval
- pharmacy acceptance and fill
- shipment
- the patient actually starting
If provider review takes days, a meaningful share of paid patients request a refund before they ever get medication. That shows up as a high refund rate and a gap between payment_success and first_fulfillment.
Instrument this part of the funnel with the same rigor as the front. Track:
- time from payment to provider approval
- time from approval to shipment
- percent of paid patients who reach first fulfillment
- refund rate segmented by time-to-approval
If refunds cluster among patients who waited longest, the highest-ROI "conversion" fix is not in marketing at all. It is provider capacity and review speed.
For that operational angle, see Telehealth Fulfillment Metrics: What to Track Between Prescription, Shipment, and First Fill and Provider Capacity Planning in Telehealth.
A dashboard leadership should actually look at
Most funnel dashboards are too detailed to drive decisions. Build one view that answers the only questions that matter weekly.
| Metric | Question it answers |
|---|---|
| Step conversion by stage | Where is the biggest leak right now? |
| Opportunity ranking (dollar-weighted) | Which fix is worth the most? |
| Mobile vs desktop paid conversion | Is mobile the hidden leak? |
| Paid conversion by source | Which channels actually work? |
| Payment to first-fulfillment rate | Are we losing paid patients post-checkout? |
| Refund rate by time-to-approval | Is provider speed costing sales? |
| Experiment status | What are we testing and did it move? |
For the broader leadership dashboard, see Weekly Telehealth Ops Dashboard: 12 Metrics Leadership Should Review.
Instrumentation checklist
Use this to stand up the measurement system.
Events
- Every stage fires an event - All 12 stages, including upload and provider approval.
- Consistent naming - One convention, documented, enforced.
- Server-side where possible - Resilient to ad blockers and mobile.
- Source stamped on first event - Carried through to paid and fulfilled.
- De-duplication - No double-counted successes.
- PHI handled correctly - Masked in replays, compliant in analytics.
Analysis
- Funnel table built - Unique people per stage, fixed cohort and window.
- Step conversion calculated - Not just cumulative.
- Opportunity ranking - Dollar-weighted leak prioritization.
- Segmented funnels - Device, source, product, geography.
- Post-payment funnel - Payment to first fulfillment tracked.
Action
- Top leak diagnosed - The "why," not just the "where."
- One-variable test - A/B where volume allows.
- Downstream guardrail - Paid conversion and refund rate watched.
- Re-rank after each fix - Attack the next leak.
- Leadership dashboard live - One view, weekly.
Common mistakes
A few traps that waste quarters.
Reading cumulative instead of step conversion
The last steps always look worst cumulatively. Judge each step by its own conversion rate.
Optimizing the visible step
The hero image is visible. The upload step is not. Instrument before you argue.
Ignoring mobile segmentation
Aggregate funnels hide mobile-specific leaks, and most GLP-1 traffic is mobile.
Stopping at payment
Payment is not the conversion in GLP-1. First dose is. Measure the post-payment funnel.
Before-and-after instead of A/B
Seasonality and traffic mix will fool you. Use controlled tests.
Chasing volume over quality
A step "fix" that increases throughput but raises refunds is a loss disguised as a win.
No dollar weighting
Without weighting leaks by value, teams fix the biggest count instead of the biggest opportunity.
How this connects to everything else
This post is the diagnostic layer under every other conversion tactic.
Once you can see the funnel:
- abandonment recovery becomes targeted, because you know the drop-off point. See Abandoned GLP-1 Checkouts: A Recovery Flow That Wins Back Drop-Offs Without Feeling Pushy.
- intake redesign becomes targeted, because you know which question loses people.
- checkout work becomes targeted, because you know whether the leak is price, payment, or trust.
- provider capacity becomes a conversion lever, because you can see refunds clustering by review time.
Tactics without instrumentation is guessing. Instrumentation without tactics is a dashboard nobody acts on. You need both, in that order.
Final takeaways
The team that wins at GLP-1 conversion is not the one with the most tactics. It is the one that can see the funnel clearly and fix the step that actually costs the most.
To get there:
- model the real funnel, all twelve stages, including upload and provider approval
- fire a clean, consistent, server-side event at every step
- read step conversion, not cumulative, to find the worst stages
- weight each leak by recoverable dollars, not raw count
- diagnose why before you fix
- confirm fixes with controlled tests and downstream guardrails
- segment by device and source to find hidden leaks
- measure the post-payment funnel, because first dose is the real conversion
Find the leak that matters. Fix it. Re-rank. Repeat.
That loop beats any single tactic, because it tells you which tactic to use and when.