Ecommerce OS · Telehealth7 min readUpdated November 2, 2026

Checkout A/B Testing for Telehealth Companies

What a telehealth company should A/B test, what it should not, the metric that decides a test, and how much traffic you need for a result.

Nick Bonnassies, CMO at Tagada · checked against 8 sources

Article summary

TL;DR

Test the commercial parts of the journey: the offer, how the price is shown, the checkout and the plans. Judge every test on patients who were approved by a clinician and paid, not on intakes or checkouts. Leave clinical questions and required disclosures out of testing, and keep patient answers out of your testing tool.

A variant that brings more intakes can bring fewer paying patients if those intakes are declined more often. Approved and paid is the primary metric.
Required terms are not test variables: ROSCA requires material terms before billing information and a simple way to stop charges.
Visa requires a reminder at least 7 days before a trial or introductory price ends, whatever variant the patient saw (April 2026).
From a 5% baseline, detecting a 10% relative lift takes about 31,000 visitors per variant at standard confidence: most brands should test big changes.
Forrester found 33% of healthcare organizations use A/B testing in their marketing mix, as cited by Kameleoon in 2024.

Want to go further? Ask

On this page

This guide is part of our playbook on the best ecommerce platform for telehealth companies. The general guide, for any business: what a custom ecommerce checkout must do. For the checkout itself, see checkout optimization for telehealth companies.

The metric that decides a telehealth test

In ecommerce, a completed checkout is revenue. In telehealth, it is a request: a clinician still decides, and the card is captured only after approval. So the winning variant is the one that brings more approved, paid patients per visitor.

Metric	Role	Why
Paid patients per visitor	Primary	Combines conversion, approval and capture
Intake completion	Diagnostic	Shows where people drop, not who pays
Checkout conversion	Diagnostic	Can rise while approvals fall
Approval rate	Guardrail	A variant can attract people who are not eligible
Capture success	Guardrail	Cards that fail after approval
Refunds and disputes at 30 days	Guardrail	A clearer offer should lower them, not raise them
Month-2 retention	Read later	Confirms the winner on the first refill

Example

Variant B lifts checkout conversion from 20% to 23% of intakes. Its approval rate falls from 80% to 68%, because a shorter page attracted more people outside the program. Paid patients per 100 intakes: A gives 16, B gives about 15.6. B "won" the checkout and lost the test.

Fig. 1B wins the checkout, A wins the test

Variant A

Checkout20% of intakes
thenClinician approval80% approvedApproved

16 paid patients per 100 intakes

Variant B: shorter page

Checkout23% of intakes
thenClinician approval68% approvedDeclined

About 15.6 paid patients per 100 intakes

Count paid patients, not checkouts.

Illustrative example.

What to test first

Start where the change is big and the step is commercial.

Step	Test	What it can change
Offer page	Price above the fold vs. after "How it works"	Who starts the intake
Entry point	Short eligibility quiz first vs. offer page first	Intake starts and completion
Plans	Monthly as default vs. a 3-month plan as default	Plan mix, refunds, retention
Billing message	"Charged only if a clinician approves" near the button vs. not	Checkout conversion, disputes
Payment methods	Apple Pay and Google Pay on top vs. card form first	Mobile checkout conversion
Order bump	One non-prescription add-on vs. none	Order value, checkout conversion
After approval	Confirmation page with next steps vs. plain receipt	Support tickets, first-refill retention

Plan pricing: bundles and multi-month plans for telehealth companies. Add-ons: order bumps for telehealth companies.

Fig. 2Test the commercial steps, never the clinical ones

Offer pagePrice above the fold vs. after 'How it works'Commercial: OK to test
Entry pointShort eligibility quiz first vs. offer page firstCommercial: OK to test
PlansMonthly as default vs. a 3-month plan as defaultCommercial: OK to test
Billing message'Charged only if a clinician approves' near the button vs. notCommercial: OK to test
Payment methodsApple Pay and Google Pay on top vs. card form firstCommercial: OK to test
Order bumpOne non-prescription add-on vs. noneCommercial: OK to test
After approvalConfirmation page with next steps vs. plain receiptCommercial: OK to test
Clinical questionsYour clinicians own themDo not test
Required disclosuresTerms, reminders and notices go to every patientDo not test

Seven steps to test; the intake and the disclosures stay out.

From the tables on this page.

What not to A/B test

Some parts of the journey are not yours to optimize for conversion.

Clinical questions and eligibility criteria
Your clinicians own them. Changing their order or wording is a clinical decision, not a growth experiment.
Consent language
Changes go through your counsel, then to everyone at once.
Required subscription terms
Price, frequency, renewal and how to cancel must be clear before billing details, with express consent1. Test their layout, never whether they appear.
Reminders and notices
Visa's trial and intro-price reminder2 and California's renewal notices3 go to every patient.
Cancellation
A variant that makes cancelling harder is the practice the FTC challenges in subscription cases.

Tagada follows Visa and Mastercard rules and the laws that apply to payments.

How much traffic you need

Most telehealth brands do not have the traffic for small tests. The standard two-proportion calculation (95% confidence, 80% power) gives:

Baseline conversion	Lift to detect	Visitors needed per variant (approx.)
5%	+10% relative (5.0% to 5.5%)	31,000
5%	+20% relative (5.0% to 6.0%)	8,100
5%	+40% relative (5.0% to 7.0%)	2,200

Fig. 3Visitors needed per variant, from a 5% baseline

+10% relative lift31000 visitors
+20% relative lift8100 visitors
+40% relative lift2200 visitors

Source: Two-proportion calculation, 95% confidence, 80% power

Halve the lift, and the traffic roughly quadruples.

Standard two-proportion calculation, 95% confidence, 80% power, from the table above.

Three rules follow:

Test big changes (a new entry point, a new default plan), not button colors.
Run full weeks, and long enough for clinician decisions on the last visitors to come in before you read the primary metric.
Decide the stop rule before you start, and do not stop the moment one variant looks ahead.

Testing tools and patient data

A testing tool sees what happens on the pages it runs on. On an intake page, that can include answers.

Run page tests on the offer page and checkout, not inside the medical intake.
Check what the tool records: session replays, heatmaps and form analytics can capture typed text. The FTC's 2023 analysis says pixels can collect "information users have typed within a form"4.
Ask your counsel about contracts: a vendor that handles protected health information for a covered entity is a business associate under HIPAA5.
Keep variant names neutral, with no condition or treatment in experiment names or URLs.

Watch out

Session recording turned on "site-wide" also records the intake. Exclude those pages explicitly.

Tracking setup: server-side tracking for telehealth companies.

How to run a test, step by step

Write the hypothesis: "Showing 'charged only if approved' near the button raises paid patients per visitor."
Pick the primary metric and guardrails from the table above.
Check the sample size for the lift you expect; drop the test if you cannot reach it in 4 to 6 weeks.
Split traffic evenly and keep everything else unchanged, including ads.
Wait for clinician decisions on the last visitors before reading results.
Ship the winner, then check month-2 retention on both cohorts.

Read the primary metric only once clinicians have decided on the last visitors.

Reading results in money: LTV and CAC for a telehealth company.

How Tagada helps

Tests on the offer and the checkout

Tagada runs tests on the checkout itself, with payments on TagadaPay: "Native 3DS2, device intelligence, and rules engine: approve real customers, block the rest."8

A/B tests on offers and checkout, with order bumps, one-click upsells, Apple Pay and Google Pay
Server-side events to Meta (Conversions API)
Routing across TagadaPay and processors such as Stripe, NMI and Airwallex, with a cascade to another processor when one is down or times out

Building the whole telehealth brand? With Tagada Rx, the card is held and captured only once a doctor approves7, so each variant can be read on approved, paid patients. US patients only, by selection.

Start free

Updated on November 2, 2026

Nick Bonnassies

CMO at Tagada

CMO at Tagada. Before Tagada, Nick built direct-to-consumer brands that generated over $115M in sales, raised over $25M in crowdfunding and sold two of them. He also works as a fractional CMO on marketing operations and AI systems.

Frequently asked questions

What should a telehealth company A/B test first?

The biggest commercial levers: where the price appears, quiz first or offer first, which plan is the default, and the billing message near the pay button. Small visual changes rarely reach significance at typical telehealth traffic. Judge each test on approved, paid patients per visitor.

What metric should a telehealth A/B test use?

Paid patients per visitor: visitors who completed intake, were approved by a clinician and had their card captured. Intake and checkout rates are useful diagnostics, but a variant can raise them while attracting patients who are declined more often.

How long should a checkout A/B test run?

Long enough to reach the sample size for the lift you want to detect, in full weeks, plus the time clinicians need to decide on the last visitors. From a 5% baseline, a 20% relative lift needs about 8,100 visitors per variant.

Can I A/B test my intake questions?

Not the clinical content. The questions and eligibility criteria belong to your clinicians, and consent language to your counsel. You can test what surrounds the intake: the offer page, the entry point, progress indicators and the checkout.

Does my A/B testing tool need a BAA?

If it receives protected health information for a covered entity, it is a business associate under HIPAA, and your counsel will likely require one. The simpler path is to run tests only on pages without health information and exclude intake pages from recording tools.

Sources

Checked November 2, 2026

Tagada

Judge every test on approved, paid patients

Funnels, one-click upsells, subscription billing and payment routing, on one customer record.

Start free See pricing

01Create your free account
02Connect your store or build a checkout
03Launch funnels, upsells and subscriptions