The first ad test: one variable, one window, one stop rule

How much should a first ad test cost?

Nothing official sets the number. The Small Business Administration's startup guidance lists advertising and marketing as its own cost category and sorts costs into one-time and recurring, but it publishes no percentage of revenue and no dollar figure; its stated method is to look online and ask mentors, vendors and service providers what comparable businesses pay 1Ref 1U.S. Small Business Administration (2026).Plan your business — Calculate your startup costs.The absence of any percentage-of-revenue or dollar rule of thumb for sizing an advertising line, SBA's own comparison-shopping method for pricing a startup cost line, and its treatment of advertising and marketing as a distinct cost category split into one-time and recurring.. So the size of a first test comes from the practice's own numbers, starting with what it can lose.

In practice that is a number you can hand over and still make every other decision the same way. If losing it would push out the startup budget, delay the first hire, or change what you pay yourself in March, it is too large for a first test, whatever the platform suggests you bid.

Before any of it, read the advertising rule of the board that licenses you. Those rules are written state by state and they differ: what a clinician may claim, whether patient testimonials are permitted at all, how a specialty may be described. The board that licenses you is the one that controls, and its regulation is a document you can read in an evening.

A test tells you what a new patient costs to acquire. Whether that cost is survivable depends on what a patient is worth to the practice across the whole relationship, and that is a separate calculation with its own inputs. The order is patient value first, ad budget second.

One variable: what the test is allowed to change

One thing changes and the rest holds still for the length of the window: one audience, one message, one destination, one intake path. Two changes at once give a result that belongs to neither of them, and a practice this size never runs enough volume to pull them apart afterwards. Pick the variable that is cheapest to change again, because you will be changing it again.

The fixed half is where a first test usually comes apart. The phone gets answered differently in week three. The intake form loses a field. A second ad goes up because the first one looked slow, and each of those is defensible on its own. Together they turn a month of spending into a story that can be told either way.

Attribution at this scale is one question and one field. Ask every new patient how they found you and record the answer somewhere it can be counted later. A platform's own report counts what happened on the platform; the intake answer counts what reached the practice, and the two numbers rarely agree.

One window: long enough for your own booking lag

The window has to outlast the gap between someone seeing the ad and sitting in a chair. That gap is a number your own scheduling data already holds, and no published benchmark can supply it. Pull the last thirty new patients, take the date of first contact and the date of the first paid visit, and look at the longest gap you see. The window is that gap plus the weeks it takes to accumulate bookings worth counting.

Small numbers move for reasons that have nothing to do with the ad. Four new patients in a month becomes three when one of them reschedules into the next, a swing of a quarter produced by a calendar. A longer window does not remove that. It makes the swing a smaller share of the total.

No federal small-business publication sets a minimum spend, or a minimum number of bookings, at which a first test's result stops being noise, and the thresholds that circulate come from firms selling the service. The working rule here is a convention, offered as one: run the window at least twice through your booking lag, and read the result as a direction to check rather than a number to plan against.

One stop rule, fixed before the money moves

A stop rule is two numbers and a date, written down before the first dollar goes out: the total spend that ends the test whatever the results look like, the minimum number of booked and paid visits that would justify a second round, and the day you sit down and compare them. Set afterwards, it becomes a decision made by whoever most wants to keep going, and that is usually you.

Then check that stopping is possible. Most of the tooling around a first test bills monthly and renews itself: the landing page, the call tracking, the scheduling widget, the retainer. For anything bought over the internet, federal law sets a floor under those charges, requiring material terms disclosed clearly and conspicuously before billing information is collected, express informed consent before the first charge, and a simple mechanism to stop future charges 2Ref 2United States Congress (Restore Online Shoppers' Confidence Act) (2010).15 U.S. Code § 8403 — Negative option marketing on the Internet.The three federal baseline requirements for a recurring charge sold online: clear and conspicuous disclosure of material terms before billing information is collected, express informed consent before the first charge, and a simple mechanism to stop future charges, plus the statute's own scoping to a transaction effected on the Internet.. The statute's own text scopes that to a transaction effected on the Internet, so a contract signed on paper is a different question with a different answer.

Find the cancel path in every tool before the test starts, and write the date you would use it beside the stop rule.

Where the spend lands on the tax return

It depends on whether the practice was operating when the money went out. Advertising run before a practice opens is a start-up cost on the IRS's own example list, which names advertising alongside travel, surveys and training 3Ref 3Internal Revenue Service (2024).Publication 583 (12/2024), Starting a Business and Keeping Records.Advertising appearing on the IRS's own example list of pre-opening start-up costs, the election to deduct up to $5,000 of start-up costs in the first year, the dollar-for-dollar reduction of that allowance once total start-up or organizational costs exceed $50,000, the pointer to Form 4562 for amortizing the excess, and the absence of a defined point at which pre-opening advertising becomes a current expense.. Once the practice is operating, the same spend is an ordinary and necessary business expense under the general rule a Schedule C filer works from 4Ref 4Internal Revenue Service (2025).Publication 334, Tax Guide for Small Business (For Individuals Who Use Schedule C).The general ordinary-and-necessary test a Schedule C filer applies to a business expense once the practice is operating.. The two reach the return by different routes.

Publication 583 does not walk through the amortization of anything above the election; it sends the reader to another form 3Ref 3Internal Revenue Service (2024).Publication 583 (12/2024), Starting a Business and Keeping Records.Advertising appearing on the IRS's own example list of pre-opening start-up costs, the election to deduct up to $5,000 of start-up costs in the first year, the dollar-for-dollar reduction of that allowance once total start-up or organizational costs exceed $50,000, the pointer to Form 4562 for amortizing the excess, and the absence of a defined point at which pre-opening advertising becomes a current expense.. That form is Form 4562, where the item sits in Part VI, Amortization, on line 42, column (a), under the label Startup and organizational costs 6Ref 6Internal Revenue Service (2025).2025 Instructions for Form 4562, Depreciation and Amortization.The exact form, part, line and column where start-up costs above the election are amortized, and the label the 2025 instructions give that item..

Timing matters more than it looks for a test that straddles a year end. Under the cash method an expense is generally deducted in the tax year it is actually paid 5Ref 5Internal Revenue Service (2022).Publication 538 (01/2022), Accounting Periods and Methods.The cash-method rule that an expense is deducted in the tax year it is paid, which splits a test window crossing a year end, and the rule that accrual is required only where merchandise is an income-producing factor, so a service practice with no inventory can stay on the cash method., so a window opened in November and closed in January splits its deduction across two returns by payment date. A service practice carrying no inventory can generally stay on that method, since accrual is required where the production, purchase or sale of merchandise is an income-producing factor for the business 5Ref 5Internal Revenue Service (2022).Publication 538 (01/2022), Accounting Periods and Methods.The cash-method rule that an expense is deducted in the tax year it is paid, which splits a test window crossing a year end, and the rule that accrual is required only where merchandise is an income-producing factor, so a service practice with no inventory can stay on the cash method..

Where pre-opening advertising stops being a start-up cost and becomes a current expense is a line Publication 583 never draws 3Ref 3Internal Revenue Service (2024).Publication 583 (12/2024), Starting a Business and Keeping Records.Advertising appearing on the IRS's own example list of pre-opening start-up costs, the election to deduct up to $5,000 of start-up costs in the first year, the dollar-for-dollar reduction of that allowance once total start-up or organizational costs exceed $50,000, the pointer to Form 4562 for amortizing the excess, and the absence of a defined point at which pre-opening advertising becomes a current expense.. Give your preparer the date the practice began operating and the date on each ad invoice. That is the math to run with a CPA, not an election to make from a web page.

What one finished window leaves you

A cost per new patient computed from your own records, and a short written note of what you would change next. The cost is total spend, tooling included, divided by the new patients who booked, paid, and named the ad at intake inside the window. It will read worse than any figure quoted in a sales call, because it counts the parts a sales call leaves out.

Keep the pieces underneath the ratio: spend by week, bookings by week, the intake answers, and a dated line for anything that changed mid-window. A second test reads them, and so may a preparer.

One question is worth settling before an intake form goes live rather than after. Ad tooling that collects a name and a reason for calling is holding patient information, and whether HIPAA reaches the practice at all runs through the covered-entity test. A form that emails you the answers and a vendor that records the calls are different problems with the same first step.

Whatever the first window returns, the second test costs less to design than the first did, because the booking lag, the intake question and the stop rule are already written down.