GLP-1 Quiz Funnels: Where Patients Drop Off

The eligibility quiz is where most GLP-1 marketing budgets quietly die, and most clinics cannot see which question is doing the killing. The quiz usually lives in a form tool that reports submissions and nothing else, or on a domain the analytics never connected, so the funnel report shows a cliff between "landing page" and "consult booked" with no detail in between. A clinic that can see each step separately can fix the one that leaks. A clinic that cannot is guessing, usually at the ad creative, when the problem is question four. Curve tracks those steps as configured events, keeps the health answers out of the ad platforms, and lets you ask Curve Analyst which step is losing the most people.

TL;DR

• The quiz is a funnel, not a form. Track quiz start, each question group, contact details, consult booked, consult completed, and treatment started as separate steps.
• Where people leave tells you why. First question means the ad set the wrong expectation. Weight and condition questions mean eligibility doubt or trust. Contact details mean price or privacy fear. Booked but never completed means scheduling and reminders.
• Split every quiz funnel by device. The blended number hides which one is broken.
• Quiz answers are health information. They do not go to Google, Meta, TikTok, or any tool that has not signed a BAA. Step counts go to your analytics; answers do not.
• Curve funnels render step-by-step drop-off from your configured events, Analyst shows it as a step chart, and both split by device and traffic source.

Why the eligibility quiz is where GLP-1 ad spend dies

Every GLP-1 clinic runs the same shape of funnel: ad, landing page, "see if you qualify," a quiz about goals, weight range, conditions, and state, then contact details, then a consult. Somewhere in that sequence a large share of the people you paid for leave, and the platform dashboard cannot say where because the platform only sees the click and, if you are lucky, the final booking.

The reason clinics cannot see the middle is structural. Quizzes get built in whatever tool the agency or the front desk already knew: a form builder, a survey tool, a quiz plugin. Those tools collect answers well and tell your analytics nothing about a visitor who reached question three and stopped. Many run on a separate domain, so the visitor who clicked your ad and the visitor who submitted the quiz look like two unrelated people. So the report says clicks up, bookings flat, and the meeting turns into a debate about creative, because creative is the only thing anyone can see.

What a GLP-1 quiz funnel looks like step by step

Write the steps out before you configure anything. Most clinics find their "quiz" is eight steps once they are honest about it.

1. Ad click. The person arrives with whatever promise the ad made, and holds every later step against it.
2. Landing page view. Offer, price framing if any, and the "see if you qualify" button.
3. Quiz start. The first question loads. This is distinct from the page view and is the first place people vanish.
4. Mid-quiz questions. Goals, weight range or BMI bracket, medical history, medications, state of residence. Track these as a few grouped steps (goals, health screen, location), not one per question.
5. Contact details. Name, email, phone. The moment the quiz stops being anonymous.
6. Consult booked. A time is chosen on a scheduler, often on yet another domain.
7. Consult completed. The person showed up.
8. Treatment started. The prescription was written and the first order happened.

Your ad platforms should see, at most, that a lead or a booking happened. Your analytics should see all eight steps as counts. Nobody outside your clinic and your BAA-covered vendors should see what anyone answered. Be careful naming events: "quiz_step_2" is a fine event name, "answered_diabetes_yes" is not, because the name itself has become health information.

The drop-off patterns and what each one usually means

Once the steps exist, the shape of the loss tells you what to fix.

Drop at the first question: the ad and the page disagree

People who click, land, press start, and leave on question one were sold something the quiz does not appear to deliver. The ad promised "GLP-1 from home this week" and the first question asks for an insurance provider. The ad showed a price and the page hides it. The fix lives in the ad set and landing page copy. Compare this drop by campaign and creative before touching anything else, because it almost always varies by source.

Drop at the weight and condition questions: eligibility doubt or trust

Someone reaches the question about weight range, conditions, or current medications, and closes the tab. Either they suspect they will not qualify and would rather not be told, or they are not sure who will see the answer. If the ad implies everyone qualifies, the BMI question feels like a trap. And an unbranded form on a strange domain asking about diabetes is a reasonable thing to abandon. Branding the quiz domain, stating what happens to answers, and putting the clinical reason for each question in one short sentence near the field usually moves this step more than any design change.

Drop at contact details: price or privacy fear

The visitor answered every health question and then declined to give a phone number. They did the hard part and balked at the easy part, which means the easy part is not easy. Either they expect the phone number to trigger a sales call and a price they have not seen, or they realise the health answers are about to be attached to their name. Show the consult price before asking for a phone number, and say who will contact them and how.

Drop between booked and completed: scheduling and reminders

This is the loss most clinics never measure because it happens after the "conversion." A consult was booked, the ad platform got credit, the agency report went green, and nobody showed up. The causes are dull: slots too far out, a confirmation that went to spam, no reminder the day before, no reschedule path. It belongs in the same funnel chart because it is where cost per booked consult turns into cost per patient.

Why you should split the quiz funnel by device

An eight-step form behaves differently on a phone. Progress indicators wrap, the state dropdown is clumsy, the keyboard covers the next button, and the scheduler on the booking step may not render inside an in-app browser from a social feed. Paid social traffic is overwhelmingly mobile, so a quiz that works on the desktop where it was built can be quietly failing for most of the people who see the ad.

A blended completion rate that looks acceptable can be the average of a healthy desktop funnel and a collapsed mobile one. Split by device before concluding anything about a step, then split by source, because the device mix follows the traffic source. A mid-quiz drop that only exists on mobile from paid social is a rendering problem. A drop that exists on every device from every source is a question problem.

The privacy rule: quiz answers stay out of ad platforms

Everything a person types into a GLP-1 eligibility quiz is health information. Weight, BMI bracket, diagnoses, medications, whether they have tried semaglutide before. Once it is tied to a name, an email, a phone number, or a persistent identifier, you are handling protected health information, and the tools that touch it need a BAA. The FTC has also pursued health data disclosures to ad platforms on unfairness grounds, so the exposure is not limited to HIPAA.

• Ad platforms receive the fact that a step was reached, never the answer. "Consult booked" is fine. A parameter carrying the weight range is not.
• URLs are data too. A quiz tool that puts the answer in the next step's URL, plus a pixel that sends page URLs by default, has just shipped the answer to the platform.
• Any tool that stores or processes the answers needs a BAA with you: the form builder, the scheduler, the CRM, the reminder email tool, any analytics that sees form contents.
• Retargeting off quiz progress deserves its own caution. An audience of "people who reached the health screen" is an audience of people who told you about their health. Our article on GLP-1 retargeting and medication exposure walks through how that goes wrong.

None of this stops you measuring the funnel. Step counts are not health information. Our piece on where intake forms leak health information covers how form tools and pixels interact, and the companion on GLP-1 landing pages that convert without capturing PHI covers the page in front of the quiz.

How Curve funnels and Analyst show the step-by-step view

Curve receives events on its own US-hosted infrastructure first, server-side, strips what should not leave, and forwards only the allowed conversions to Google Ads, Meta, TikTok, Microsoft, LinkedIn, and the rest, under a BAA signed with every customer on every plan. That is what lets you track every quiz step while the ad platforms only learn that a booking happened.

Each quiz step becomes a configured event, and a funnel is an ordered list of those events. Curve reports how many times each fired, compares periods, and shows the drop-off between them. When the quiz or the scheduler lives on another domain, Curve Forms and bridge tokens keep the later steps attributable to the original ad click, so the person who booked on the scheduler domain is the same funnel row as the person who clicked the ad. That is the part most setups lose.

Asking instead of building

Curve Analyst is the AI chat inside the Curve dashboard. You ask a plain-language question, it queries your account's own data, and it answers with text and a rendered chart. Funnels come back as step charts. Ask for the quiz funnel over the last thirty days and you get each step with its count and the loss between steps, and you can immediately ask it to split by device or by source without rebuilding anything. It keeps the date range and filters of the screen you opened it from.

The honesty rules matter here. Every number Analyst quotes comes from a query against your account; it never estimates. If a step has no events, it says the step is empty rather than inventing a zero, and it treats "unknown" and "zero" as different, which is the difference between "nobody reached step three" and "step three is not being tracked."

It is read-only: it cannot change tracking, edit destinations, or take actions. It does not look up individuals or answer person-level questions, so there is no path from "who dropped at the health screen" to a name. It does not access session recordings or heatmaps and does not yet report on tracking configuration, which is planned. It runs on Claude through Amazon Bedrock inside infrastructure covered by Curve's BAA with AWS; your data is not exported to a separate AI product and is not used to train models. For a broader walkthrough, see what Curve AI Analyst is, and for the same approach at the consult stage, see diagnosing a telehealth consult funnel with AI.

What to ask Curve Analyst

Start with the whole funnel, then narrow. These three prompts, in this order, usually tell you within minutes whether the problem is the ad, the quiz, the booking step, or the phone.

• Walk me through the quiz funnel for the last 30 days and name the step losing the most people.
• Split that funnel by device.
• Compare quiz completion for paid social vs paid search traffic.

The first answer gives you the step chart and the biggest loss, the second tells you whether it is a mobile problem, and the third tells you whether it is tied to one channel's ads. The Curve Analyst announcement has more examples.

Frequently asked questions

Our quiz is built in a form tool on its own domain. Can we still see step-by-step drop-off?

Yes, provided each step fires an event your analytics can receive and the cross-domain hop is bridged so later steps stay tied to the original visit. Curve Forms and bridge tokens exist for that hop. If the form tool cannot fire step events at all, the practical answer is usually to rebuild the quiz in a tool that can.

Is it a HIPAA problem to send "quiz completed" to Meta or Google?

The fact that a step was reached is a count, not a health answer. The problems begin when the event carries the answers, when the event name encodes an answer, when the page URL carries an answer, or when the identifiers sent alongside it let the platform tie a health context to a person. Curve's server-side path exists to strip those before forwarding. Whether a given event is safe to send depends on what it carries, and that review is worth doing with counsel.

Which drop-off step should we fix first?

The one with the largest absolute loss, after splitting by device, unless that step is the first question. A first-question drop is cheaper to fix at the ad and landing page level than anywhere in the quiz, and fixing it raises the volume flowing into every later step. After that, the contact step, because showing the price earlier costs nothing to test.

Can Curve Analyst tell us which patients abandoned the quiz?

No. Analyst does not look up individuals or answer person-level questions, and we built it that way deliberately. It reports counts and rates by step, device, source, campaign, and period. Following up with people who abandoned belongs in your CRM under your consent terms.

If your GLP-1 funnel report has a cliff between the landing page and the consult, the fix starts with seeing the steps in between. Book a demo and we will walk through your quiz, step by step, and show you where it is leaking.