Multi-location healthcare marketing attribution starts from an arithmetic problem before it becomes a compliance one: a group running twelve sites is not running one market twelve times, it is running twelve separate local markets, each with its own competitive intensity, its own provider capacity, and its own demand curve - funded out of a single regional or system-wide budget that has to be re-divided among all twelve on some ongoing basis. The question a group actually needs answered is not "is marketing working," which barely means anything applied to twelve unrelated places at once. It is "which of these twelve places should get more of the budget this month, and which should get less" - a question a blended, group-wide number cannot answer no matter how precisely it is calculated.
Why a group-level cost per patient is close to meaningless
Averaging cost per new patient across twelve sites produces a number that describes none of them. A site sitting in a market with little competition can post a low cost per patient purely because the auction it bids into is thin, regardless of how well its marketing is actually run. A site in a saturated metro can post a high cost per patient while running the identical campaign strategy, purely because the competitive floor in that market is higher. Blend the two into a single group figure and the number that results is closer to a statement about how competitive the group's collection of markets happens to be than it is a statement about how well any campaign is performing.
The distortion runs the other direction too. A genuinely underperforming site sitting in a cheap, low-competition market can still post a respectable blended number, simply because its market floor is low enough to make mediocre execution look adequate next to it. A group relying on the blended figure to decide where marketing needs attention will consistently miss this site, because nothing about the group average isolates a market's baseline cost from what a specific campaign actually contributed on top of it. The only way to separate the two is to look at each site against its own local market rather than against a number that has already absorbed eleven other markets into itself.
A group of twelve does not even need one site to be an outlier for the average to mislead. Twelve sites each performing exactly at their own market's expected level will still produce a blended figure that matches none of them individually, simply because twelve different market baselines averaged together describe a market that does not exist anywhere the group actually operates. The average is not a measurement error correcting itself out over enough sites. It is a description of a market nobody is competing in, calculated from twelve markets somebody actually is.
Capacity as a constraint that inverts the usual goal
Marketing optimization runs on one default instinct almost everywhere: find what is converting well and put more budget behind it. That instinct is actively wrong the moment a location it would apply to is running near its provider capacity ceiling. A site whose campaigns are converting inquiries into bookings at an excellent rate, sitting at a location with little room left on the schedule, does not need more of that budget - it needs less, redirected somewhere the resulting bookings can actually be seen. Pushing more spend at a site because its numbers look strong, without checking what capacity that site has left, produces exactly the outcome the campaign was supposed to prevent: more people calling in, a longer hold before anyone can see them, and a growing number of callers who give up and book somewhere else entirely.
This means the marketing decision at a capacity-constrained site is the mirror image of the decision everywhere else in the funnel. Everywhere budget is unconstrained, success is the signal to spend more. At a site nearing its ceiling, success is the signal to spend less there and more at a site with real room, and getting that backwards is not a minor inefficiency - it is actively worse than doing nothing, since the wasted spend also degrades the experience of the people it did reach. A group that reviews capacity only at budget-planning time, rather than on a rolling basis, will keep making this exact mistake between reviews, because a site can cross from having room to being full well before the next scheduled check catches up to it.
Nobody sets out to fund a queue on purpose. The mistake happens because the signal that should trigger a pullback - a site's own capacity tightening - lives in a scheduling system the campaign dashboard has no visibility into, while the signal that keeps the current budget in place - a strong-looking conversion rate - sits directly in front of whoever is reviewing the account. Absent a deliberate capacity check running on its own schedule, the easier signal wins by default, and a location keeps getting rewarded with more spend for the exact success that should have triggered less.
Provider-level demand variation within a single location
Per-location measurement is real progress over a group-wide blend, and it is still one level short of where the actual constraint lives. A single location's stated capacity is itself an average across every provider working there, and providers at the same address routinely carry very different loads. A location reporting comfortable overall capacity can be hiding one specialist booked out for months alongside a newer provider at the same address with real openings that nobody's marketing is directing anyone toward, because the campaign was built to fill "the location" rather than to fill the specific provider with room.
This matters most exactly where a group is trying to ramp a new provider or build a newer specialty's patient base at an established site. Generic location-level marketing keeps sending inquiries into whichever provider intake happens to route them to, which usually means the established provider with the existing reputation and referral base, since that is who a caller asks for by default. The new provider's real capacity sits unused, invisible in a location-level number that only reports the site as a whole rather than broken out person by person. Reaching that unused capacity takes campaigns and, where the front desk supports it, call routing built around the specific provider who has room, not just the site that houses them.
What a location scorecard contains
Turning any of this into a decision someone can actually act on requires an artifact built for that purpose, reviewed on a schedule short enough to matter: a location scorecard. At minimum it shows, for each site, the cost per booked and attended appointment measured against that site's own history rather than the group average; remaining capacity, ideally broken down by provider where more than one works at the address; the local competitive read that explains whether a given cost figure reflects execution or simply reflects how expensive that specific market is to compete in; and the trend on all three over the recent weeks, since a snapshot alone cannot show whether a site is heading toward its ceiling or away from it.
The scorecard's job is narrower than a full marketing report. It exists specifically to answer the reallocation question - where does the next incremental dollar produce a new patient, and where would it just add to a queue nobody can see into from a group-wide view. A regional lead working from twelve individual scorecards can move budget toward real remaining capacity and away from a site quietly nearing its ceiling; a regional lead working from one blended number is making that same decision blind, and will keep funding whichever site happened to convert well the last time anyone checked, whether or not it still has anywhere to put the result.
Building it once and treating it as finished defeats the purpose as surely as never building it at all. A scorecard reviewed on the same cadence turnover and seasonal demand actually move on stays accurate; one pulled together for a single planning cycle and left untouched describes a group of twelve sites as they stood months earlier, which is close to describing a different group entirely once enough turnover and seasonal shift has passed underneath it.
Multi-location healthcare marketing attribution, in practice
None of this replaces attribution built to where a health system's revenue is actually recorded - it depends on it, since a scorecard is only as accurate as the booked-appointment data feeding each site's row. What it adds is the layer above that connection: the discipline of reading twelve sites as twelve separate constrained systems rather than one campaign running at scale, and reallocating a single budget across them on the same rolling basis capacity itself changes. The intake and routing work behind reaching a specific provider's unused capacity, rather than whichever provider a caller happens to be routed to by default, is conversion rate optimization work as much as it is a marketing decision - the two stop being separable once capacity is measured at the level where it actually lives.
A multi-location healthcare marketing program that cannot show this breakdown is still making budget decisions on twelve markets' worth of noise collapsed into one number. The group is not one place with one problem. It is twelve places with twelve problems, sharing a single budget that has to be re-split as often as any one of the twelve changes.