Attribution Windows for Healthcare: Choosing Yours

Choose your attribution window by measuring how long your patients actually take between first click and booked appointment, then set the window to cover the bulk of that distribution rather than accepting a platform default. Healthcare consideration cycles routinely run weeks, so the common 7-day click window systematically undercounts clinic conversions and pushes budget toward whichever campaigns happen to convert fastest. Curve is the HIPAA-compliant tracking, attribution, and analytics platform for healthcare, and it holds the click-to-outcome record you need to measure that lag in the first place, with a signed BAA on every plan.

What an attribution window actually is

An attribution window is a time limit. It says: if a conversion happens within this many days of an ad interaction, credit the ad. Outside it, do not.

There are usually two of them. A click-through window covers people who clicked the ad. A view-through window covers people who saw it and did not click. They are set separately and they behave very differently, because a click is direct evidence of interest and a view is not.

The window is a reporting and optimization decision, not a fact about the world. Widening it does not create conversions. It changes which of the conversions you already had get credited to which ads, and, crucially, it changes what the platform's bidding algorithm learns from.

Why healthcare defaults are wrong out of the box

Platform defaults were tuned for the median advertiser, and the median advertiser sells things people buy quickly. A pair of shoes, a subscription box, a software trial. Same-week purchase is normal, so a 7-day click window captures most of the truth.

Healthcare does not work like that. Consider what actually sits between the ad and the appointment:

• The decision is consequential. A surgical consult, a fertility program, a course of orthodontics, or a weight management program is a life decision, not an impulse purchase.
• Money is complicated. Insurance verification, deductible checks, financing applications, and HSA timing all add days or weeks.
• Other people are involved. A spouse, a parent, a referring physician. Consultation with other humans takes calendar time.
• Scheduling is a constraint. Even a decided patient has to find a slot that works, and specialist availability can push the appointment weeks out.
• The valuable event is late. The form fill is not the conversion that matters. The attended appointment, the accepted treatment plan, or the started program is, and it is downstream of everything above.

Put those together and a clinic's real click-to-booking lag can span a month or more, with a long tail beyond it. A 7-day window sees the front of that distribution and nothing else.

What a too-short window does to your budget

The damage is not just that the number looks small. A short window actively distorts allocation in a predictable direction.

Fast-converting campaigns look better than they are

Branded search converts fastest, because those people already decided. Bottom-of-funnel terms convert fast. Retargeting converts fast. All of them get fully credited under a short window.

The campaigns that introduce new patients to your practice convert slowly, because their job is to start a decision rather than close one. Under a short window they look expensive and underperforming, so they get cut. Six weeks later the fast-converting campaigns run out of people to close, because nobody filled the top of the funnel.

The optimization algorithm learns the wrong thing

This is the part that gets underestimated. Platform bidding optimizes toward the conversions it is told about. If your window hides the slow conversions, the algorithm concludes that the audiences and placements producing them are unproductive, and it stops showing your ads to those people.

The result compounds. Reporting bias becomes delivery bias becomes an actual change in who sees your ads.

Cost per acquisition looks worse than reality

If a third of your bookings are credited to nothing, your reported cost per booking is materially inflated. Practice owners make real decisions on that number, including deciding that paid search does not work for the practice.

How to measure your actual lag

Do not guess and do not copy a number from an article, including this one. Measure it, using your own data.

1. Get a click-to-outcome dataset. For each booked or attended appointment over a decent period, you need the date of the originating ad click and the date of the outcome. This requires click IDs captured at landing and outcomes returned from your CRM or practice management system.
2. Use at least six months. Seasonality is strong in healthcare. January weight management, back-to-school pediatrics and orthodontics, year-end deductible surges in Q4. A short sample will mislead you.
3. Plot the distribution, not the average. The average lag is close to meaningless here because the distribution is skewed with a long tail. What you want is the shape: how many days covers most of your conversions.
4. Segment by service line. A same-day urgent care visit and an elective cosmetic procedure do not share a cycle. If your practice spans both, one window will not serve both well.
5. Segment by campaign type. Branded and non-branded behave differently enough that they are worth looking at separately.
6. Pick a window that covers the bulk of the distribution. Not the extreme tail. Chasing the last few percent of very late conversions adds noise and slows the feedback loop the algorithm needs.

The output of this exercise is a defensible sentence: "for this service line, most bookings occur within N days of the click, so we report on an N-day window." That sentence survives scrutiny in a way that "we use the default" does not.

Setting click and view windows differently

Treat them as separate decisions, because they carry different amounts of evidence.

Click windows should reflect your measured cycle. A click is a person who deliberately came to your site. Crediting a booking to that click weeks later is reasonable, because you have direct evidence of the interaction.

View windows should be short, or off. A view is much weaker evidence, and a long view-through window will credit your ads for conversions that would have happened anyway. It is the fastest way to build a report that overstates paid performance. Many clinics run view-through at the shortest available setting or exclude it from decision-making entirely, while still keeping it visible for context.

Consider a shorter window for optimization than for reporting. The bidding algorithm needs a reasonably fast feedback loop to learn. Your board report needs completeness. Those are compatible goals if you are explicit about which number you are looking at and why.

The constraint nobody mentions: you cannot report on data you did not keep

A longer window only helps if the attribution chain survives that long. In practice this is where most clinics fail before they ever get to choose a number.

Browser storage does not reliably last 30 days. Safari's Intelligent Tracking Prevention caps client-side cookie lifetimes aggressively, and other browsers have tightened as well. A conversion that happens five weeks after the click will frequently have no client-side record of that click left to match against.

This is a structural argument for server-side measurement, not a preference. If the click ID was captured at landing and stored server-side, the association survives regardless of what the browser does to its cookies six weeks later. If it lived only in a browser cookie, the long window is theoretical. Our piece on why client-side pixels fail in healthcare covers the mechanics.

The same applies to the booking handoff. If your patient leaves your site for a separate scheduling tool and attribution does not travel with them, the window length is irrelevant because the chain broke on day one.

How Curve supports longer windows

Curve is HIPAA-compliant ad tracking, attribution, and analytics for healthcare, and long consideration cycles are one of the specific problems it is built around.

The Curve script installs in place of the Meta Pixel and Google tag. Click IDs including gclid, fbclid, and msclkid are captured on arrival and stored server-side in Curve's US-hosted infrastructure, alongside UTM parameters. That capture at landing is the foundation. Everything downstream depends on it.

• Server-side click ID storage. The click-to-conversion association does not depend on a browser cookie surviving for weeks, so a booking that lands 40 days later can still be matched to its originating click.
• Bridge tokens across the handoff. When a patient clicks out to IntakeQ, Calendly, or Jane App, a bridge token carries attribution across the domain boundary so the eventual booking is not orphaned.
• Incoming webhooks for outcomes. Your CRM or scheduling system posts booked and attended results back, matched on email, click ID, or bridge token. Protected core attribution and contact fields cannot be overwritten by incoming data.
• Offline conversion uploads. For outcomes that live in a practice management system with no live integration, bulk upload with automatic click ID matching gets late conversions back into the record, up to 10,000 rows per file.
• Attribution models in analytics. First-touch, last-touch, U-shaped, and assisted attribution with a 90-day lookback, so you can see how much of your reporting depends on the credit rule you picked.
• Per-destination field mapping and hashed identifiers. Late conversions forward to Meta CAPI, Google Ads Enhanced Conversions, Microsoft, and other destinations with only explicitly mapped fields, SHA-256 hashed, under neutral event aliases. Extending your measurement window does not extend your disclosure.

A signed BAA is included on every plan. For the mechanics of pushing late outcomes into Google, see server-side Enhanced Conversions setup without PHI leakage.

Practical guidance by clinic type

These are starting hypotheses to test against your own data, not prescriptions.

Urgent care and walk-in. Genuinely short cycles. Defaults are close to right, and a long window mostly adds noise.

Dental and orthodontics. General dentistry is moderate. Orthodontic and implant cases are long, often involving a consult, a financing conversation, and a scheduling gap. Segment them.

Med spa and aesthetics. Highly variable by treatment. Injectables can be quick, surgical and high-ticket body procedures are not.

Behavioral and mental health. Long and non-linear. People start, stop, and restart the search. Assisted attribution matters more here than a single credit rule.

Fertility, bariatrics, and elective surgery. The longest cycles you will encounter. Multi-month consideration is normal, and short windows are actively misleading.

Weight management and GLP-1 programs. Long research phases with heavy comparison shopping, plus a policy environment that constrains what you can say in the ad. See GLP-1 landing pages that convert without collecting PHI.

Frequently asked questions

What attribution window should a clinic use?

Whatever your measured click-to-booking distribution supports, which for most clinics is longer than the 7-day default and shorter than the extreme tail. Measure it from your own click-to-outcome data rather than adopting a number from a benchmark article.

Does widening the window inflate my results?

Widening a click window does not create conversions, it recredits ones that already happened. Widening a view-through window is different and can genuinely overstate paid performance, because it claims credit for people who never clicked.

Why did my conversion count change when I changed the window?

Because the window governs which conversions get attributed, historical reporting shifts when you change it. Note the change date in your reporting so a methodology change is never mistaken for a performance change.

Can I use different windows for different campaigns?

Platform capabilities vary and change, so verify current options before planning around them. What you can always do is analyze differently by segment in your own measurement layer, which is where the service-line differences are clearest anyway.

Do long windows create a HIPAA problem?

The window itself does not. The risk is in what you send to ad platforms and how long you retain identifiable data, not in how many days you count. Keep detailed data inside BAA-covered systems, forward only explicitly mapped hashed identifiers under neutral event names, and set a retention policy deliberately.

What breaks first when I try to measure a 30-day cycle?

Client-side storage, usually. Browser cookie lifetimes are capped well below the cycle length in several major browsers, so late conversions arrive with no matchable client-side record. Server-side click ID capture at landing is the fix.

Should the optimization window match the reporting window?

Not necessarily. Bidding algorithms benefit from a faster feedback loop, while executive reporting benefits from completeness. Running a shorter optimization view and a longer reporting view is defensible as long as everyone knows which number they are looking at.

Where to start

Pull six to twelve months of click-to-booking pairs and plot the lag. That single chart will tell you more about your attribution settings than any benchmark, and it turns your window from a default you inherited into a decision you can defend.

Curve gives clinics the record that makes the chart possible: click IDs captured at landing and stored server-side, bridge tokens that carry attribution across booking handoffs, webhooks and offline uploads that bring late outcomes back, multi-model attribution with a 90-day lookback, and per-destination field mapping with hashed identifiers so a longer window never means a wider disclosure. A signed BAA comes with every plan. Run our free compliance scanner against your site, or visit curvecompliance.com to size your real consideration cycle.

Reviewed August 2026. Ad platform attribution settings and healthcare advertising policies change frequently. Verify current requirements before implementation.