The GLP-1 Clinic Weekly Report: Five Questions

A GLP-1 clinic's weekly marketing report should be five questions, answered every Monday from the clinic's own data, and the first one should count patients who started treatment rather than leads. Most weight loss clinic marketing metrics stop at the quiz or the booked consult because that is where the ad platforms stop looking, and that is why the report lies. Curve holds the analytics, the goals and funnels, and the campaign reporting for the clinic, and Curve Analyst answers questions against that data in plain language. The med spa version of this routine follows the same shape.

TL;DR

• Answer the same five questions every week: starts by channel and campaign with cost per start, where the quiz-to-consult-to-start funnel leaked, which pages booked, what moved versus last week, and whether conversions still reach Google and Meta.
• Count patients started and monthly refills. A lead who never books teaches the ad platforms the wrong lesson.
• Demand for semaglutide and tirzepatide moves with drug availability, pricing changes, and news cycles. Compare to the prior week before you touch a bid.
• Restricted ad categories mean the pixel may be sending the platforms less than you think. The reconciliation view shows what Curve sent next to what Meta or Google credited.
• Curve Analyst answers all five from your account's data, with charts, and tells you when a number is empty, stale, or still syncing.

Which channel and campaign started the most patients, and at what cost?

Why this matters more for a GLP-1 clinic

The economics live in the refill. A patient who completes a consult and never starts treatment is worth nothing to the clinic, while a patient who starts and stays on monthly refills pays for the ads that brought them and for the ads that missed. So the top line of the report is cost per patient started, by the channel and campaign that produced each one.

Platform dashboards cannot give you this. Google Ads and Meta see only the events you were allowed to send them, credited under their own rules, and under a restricted health category that is often a thin slice of what happened. Lead counts also flatter the wrong campaigns: a broad Meta campaign can fill the quiz with people who will never be eligible and look like your best performer until you follow them to the consult.

How to ask

Open the campaign reporting screen with last week selected, open Analyst, and ask for patients started by channel and campaign with cost per start. Analyst keeps the date range and filters of that screen, so you will not get an all-time number by accident. It reports Curve-attributed conversions under the model you chose, with spend from the platforms directly, and says which it is quoting.

What a good answer looks like

A ranked table of campaigns with starts, spend, and cost per start, and a plain statement of whether the treatment-started goal fired at all last week. If that goal is empty, Analyst says so instead of quietly substituting consults, and you want to know that before the number lands in front of the owner.

What to do

Move budget toward the campaigns with the lowest cost per start rather than the lowest cost per lead, and expect the two rankings to disagree. If you have not decided what a sensible cost per started patient is, the guide to GLP-1 clinic cost per acquisition covers how to think about that number.

Where did the quiz-to-consult-to-start funnel lose people?

Why this matters more for a GLP-1 clinic

Almost every GLP-1 clinic sells through a quiz or eligibility intake, and the quiz is where most of the funnel disappears. The usual path is ad, landing page, quiz or intake, consult booked, consult completed, treatment started, then monthly refill. Each step has a different owner. A loss before quiz start is a page and offer problem. A loss inside the quiz is question design or eligibility. A loss between consult booked and consult completed is scheduling and reminders, invisible to the ad platforms because they stopped watching at the booking. A loss between consult completed and treatment started is often supply or price, and it swings weekly as availability of semaglutide and tirzepatide changes.

How to ask

Configure the funnel once in Curve with those steps as goals, then ask Analyst which step between quiz start and consult lost the most people last week. It answers with a step chart and the count at each stage. Filter the same funnel to one channel; a quiz that converts from search and collapses from Meta means the Meta audience arrives expecting something the page does not say.

What a good answer looks like

The step with the largest loss named outright, the counts at each step, and a period comparison so you know whether the leak is new. If a step has no data because the goal never fired, Analyst treats that as unknown rather than zero. That is the difference between "nobody completed the quiz" and "the quiz completion event is not reaching Curve," and those two findings call for opposite responses.

What to do

Fix the step with the largest loss that you actually control this week. Quiz abandonment usually means the quiz is too long or asks for identifying details too early; the guide to GLP-1 landing pages that convert without collecting PHI covers how to ask for less and still qualify the visitor. Consult no-shows are a reminder problem. A loss at treatment start needs a conversation with the clinical and pharmacy side, and no bid change will fix it.

Which pages and offers pulled traffic that actually booked?

Why this matters more for a GLP-1 clinic

GLP-1 clinics run several offers at once: a semaglutide page, a tirzepatide page, a virtual consult page, a membership or program page. Traffic to those pages is a vanity number under restricted ads, because the platforms will happily send people who read and leave. The page that matters is the one whose visitors go on to book, and that ranking shifts as availability and news change which drug people search for.

How to ask

Ask Analyst for the pages with the highest consult booking rate last week, then ask which sources and campaigns drove traffic to the top ones. Because Curve keeps the referring hostname as the source, you can also ask what chatgpt.com or perplexity.ai sent you and whether any of it booked. At the channel level those visits currently count as referral, so ask by source to separate them.

What a good answer looks like

A table of pages with visitors, consults booked, and booking rate, plus a note on which page has too little traffic to trust yet. The offers that pull traffic and the offers that book are usually different pages, and a good answer makes that gap visible.

What to do

Send paid traffic to the page that books, even when a different page gets more organic attention. For telehealth clinics, the consult page is often the weakest link between ad and calendar; the guide to marketing virtual weight loss consultations covers what that page has to carry.

What changed versus last week, and was it ads or the market?

Why this matters more for a GLP-1 clinic

GLP-1 demand is news-driven in a way most healthcare categories are not. A supply announcement, a pricing change, or a celebrity interview can move search interest for a week and then let it fall back. If you read a bad week as an ads problem and cut budget, you miss the recovery. If you read a good week as a creative win and scale it, you overpay for the fall-off. The weekly comparison separates the two.

How to ask

Ask Analyst what changed versus the prior week in spend, conversions, and cost per start, by platform. Then ask whether the change is concentrated in one campaign or spread across all of them. A market swing moves everything at once, including organic and direct traffic. An ads problem shows up in one platform, campaign, or ad group while the rest hold.

What a good answer looks like

The change stated with both periods, a line chart of the trend, and the concentration question answered directly. Analyst also surfaces automatically computed insights on the campaign data. When the current week is still syncing from a platform, it says so before the number, and that caveat belongs in the report rather than a partial week presented as a drop.

What to do

If the market moved, hold bids, watch the funnel, and make sure your pages say something true about availability so the quiz does not fill with people you cannot serve. If the ads moved, go to the campaign that changed and check question five before you blame the creative. A conversion that stopped arriving at the platform looks exactly like a campaign that stopped working.

Is tracking still delivering conversions to Google and Meta?

Why this matters more for a GLP-1 clinic

This is the question that makes the other four trustworthy, and the one GLP-1 clinics have most reason to worry about. Meta may classify a weight loss clinic as a restricted health and wellness advertiser, which limits what the pixel can send. Clinics respond by moving conversions to server-side tracking, which is the right move, and which also puts those conversions on a path the platform's own diagnostics cannot see. Curve receives the event first, strips what should not leave, and forwards the conversion. If that forwarding stalls, the platform goes blind, cost per start climbs, and nothing in its dashboard explains why.

How to ask

Ask Analyst what Curve sent to Meta last week next to what Meta credited, then the same for Google. This is the reconciliation view. The two numbers will not match, and a gap is usually the platform's attribution window still being open rather than under-reporting. What you are looking for is a change in the shape of the gap: sent steady while credited drops, or sent falling to nothing.

What a good answer looks like

Two columns per platform, sent and credited, with the prior week alongside. Sent stable, credited lagging: normal. Sent at zero for a platform that had volume the week before is the finding, and Analyst states it as zero rather than leaving the cell blank, because it treats zero and unknown as different answers.

What to do

Analyst is read-only and does not yet report on tracking configuration or connector health, so when sent drops, the next stop is the destination settings in the Curve dashboard rather than another question. Check that the destination is connected, that the consult and treatment-started goals are mapped to it, and that the events fire on the site. Then ask the reconciliation question again next Monday to confirm sent recovered.

What to ask Curve Analyst

Type these three prompts in order on Monday morning, with last week selected on the screen you open Analyst from. Each returns a chart or table you can drop into the owner's report.

• Show me patients started by channel and campaign last week, with cost per start.
• Which funnel step between quiz start and consult booked lost the most people last week, and how does that compare to the week before?
• What did Curve send to Meta last week versus what Meta credited?

The Curve Analyst announcement covers how the feature works, and the overview of Curve Analyst lists what it can and cannot answer yet.

Frequently asked questions

Why count patients started instead of leads or booked consults?

Because leads and consults are costs, and a started patient is revenue. Optimizing ads toward quiz completions trains Google and Meta to find people who like quizzes, a different population from people who will pay for monthly refills. Reporting on starts keeps the feedback loop pointed at the outcome that keeps the clinic open.

Can Curve Analyst show me a specific patient's journey or quiz answers?

No. Analyst does not look up individuals and does not answer person-level questions. It reports aggregates: how many people reached each funnel step, which campaigns produced starts, which pages booked. One patient's path is a job for the clinic's own systems.

Is asking an AI about our marketing data a HIPAA problem?

It depends on where the AI runs and who has signed what. Curve Analyst runs on Claude through Amazon Bedrock, inside infrastructure covered by Curve's BAA with AWS, and Curve signs a BAA with every customer on every plan. Your data is not exported to a separate AI product and is not used to train models. Pasting the same data into a consumer chatbot is a different situation: consumer ChatGPT and consumer Claude.ai are not covered by a BAA, and you should confirm current terms with any vendor first.

Why do Curve's conversion numbers differ from what Meta or Google report?

Curve reports conversions credited by its own attribution engine under the model you chose. The platforms report conversions credited under their own rules and windows, from only the events they received. The two will not match, and Analyst tells you which one it is quoting. The reconciliation view shows both so you stop arguing about which is right.

What can Analyst not answer yet?

It does not access session recordings or heatmaps, does not write SQL, does not take actions, and does not report on tracking configuration or connector health. That last one is planned. Until then, question five tells you sent dropped, and the Curve dashboard tells you why.

If you would rather answer these five questions on Monday morning than rebuild a spreadsheet every week, book a demo and bring your current report. We will run the same five questions against a live account.