White Paper One
Talent Buying & Event Management Since 1975
What Location Data Knows About Your Audience
The kinds of location data that exist, what each one can and cannot tell you, and where to buy it
The first of three white papers. For fair, festival, and event organizers. 2026 Edition.
TSE Entertainment, LLC
Austin, Texas 2026 Edition
This is the first of three white papers on location data for live events.
This white paper covers what location data exists, what each kind can tell you, and where to buy it. The second covers how to activate it: the advertising platforms, the campaigns that work, and how to budget them. The third covers what happens inside your gate, including event apps, proximity technology, and sponsor activation.
Each stands on its own. Read this one first if you have not bought anything yet.
Somewhere in a commercial database there is a record of the crowd that walked onto your grounds last season. You did not create it, you do not control it, and a competing promoter can buy access to it without your knowledge or permission.
That is the uncomfortable starting point for any honest discussion of location data in the live event business. It is also, handled correctly, one of the more useful tools available to an organizer trying to understand an audience.
This paper is written for fair boards, festival organizers, promoters, and the marketing staff who work for them. It assumes no background in digital advertising and defines every technical term in plain language on first use.
It does not recommend a vendor. It does not publish fee figures for artists or media, because those vary too widely by market and scale to be useful as printed numbers. And it does not treat location data as a substitute for the fundamentals of selling a show, which remain a compelling lineup, a fair price, and a reason to come.
Most confusion about this subject comes from vocabulary. The word “geofencing” gets applied to at least four different activities that share almost nothing beyond the fact that a boundary is drawn on a map. Before anything else is useful, those four things need separating.
This is the point where nearly every organizer forms the wrong picture, so it is worth stating plainly and in the negative first.
No push notification, text, or direct message is sent to anybody. Nothing buzzes and nothing appears on a lock screen because of the fence itself. No app of yours is involved at any stage. The person is not contacted directly and sees nothing that announces they were included in a historical audience.
What actually happens is this. Nearly every website and free app carries empty advertising slots. When somebody opens a news site, a weather app, a recipe page, or a game, those slots have to be filled in the fraction of a second before the page appears, and an automated auction decides what goes in them. Where the app, the device settings, and the advertising transaction make an identifier available, the buying system can compare it against the audience lists advertisers have loaded. If yours is a match and you win the auction, your advertisement appears on the page the person was already looking at.
So your advertisement turns up in the middle of a weather forecast they opened for their own reasons. That is the entire mechanism.
The instinct is to treat a device identifier as an address, like a phone number or an email address. It is not, and nothing can be sent to one.
It is closer to a name tag than an address. Where an app, a device and an advertising transaction make that identifier available, the buying system can compare it against an eligible audience list. You are not sending anything to the identifier itself. You are waiting for a match to occur somewhere in advertising inventory. Your audience is a guest list at the door rather than a set of addresses, and identifiers are not handed over at every door.
Three consequences follow, and each one explains something that otherwise looks like a flaw in the product.
That last point is the mechanical reason behind the argument in Section 9, and it is worth carrying forward. A device audience is not a list of people. It is a list of name tags, and name tags get thrown away.
One further thing follows. The boundary you drew was used once and then set aside. You fenced your grounds last August to work out who was standing there, and that job is finished. Nothing is live, nothing is being triggered, and the advertisement appears wherever the person happens to be rather than anywhere near your venue.
This is the first question nearly every organizer asks, and it deserves a direct answer rather than a deferral. The short version is that Facebook and Google can do one of the four things in Figure 1 and not the other three.
Neither is a neutral advertising system. Both are what the trade calls a walled garden: they sell only their own inventory, on their own terms. Both accept customer lists, each under its own rules. Google limits Customer Match to information collected in a first-party context, meaning directly from your own customers. Meta requires the advertiser to represent that it holds the necessary rights, permissions and lawful basis for the data it uploads. What neither offers is an open endpoint for a specialist vendor’s historical geofence audience. You cannot have a location vendor build a set of devices seen at a venue last August and push it into either platform as a targetable audience. That rules out the right-hand column of Figure 1 on both.
The reason is worth understanding, because it is not what most people assume. It is not that these platforms are technically incapable of handling device identifiers. In some contexts they clearly do. The barrier is in their terms, and the two are framed differently. Google’s rule turns on where the data came from: your own relationship with the customer. Meta’s puts the burden on you to hold the rights and lawful basis for whatever you upload. Under either framing, a vendor’s list of strangers observed standing in a field is not ordinary first-party customer data, whoever owns the field, and an organizer should not treat it as if it were.
Organizers reasonably push back on this. If I have a direct relationship with my ticket buyers, and my terms of sale say so, can I not just use their device identifiers? The answer is yes in principle and no in practice, for a reason that has nothing to do with consent.
Ticket sales do not produce device identifiers. They produce names, email addresses, telephone numbers and postal addresses. A device advertising identifier is an application-level object: it exists inside a mobile app and is visible to whichever software is running there. Your ticketing system never sees it. Consent language cannot create data you did not collect.
This turns out not to matter, because the data you do have is better. The major platforms accept customer lists built from email, telephone, name and address, which is exactly what ticketing produces. Those identifiers match more reliably than device identifiers, they do not reset when somebody replaces a phone, and no tracking permission prompt stands between you and them. Upload your buyer list, build similar audiences from it, suppress the people who have already purchased. That is first-party data doing the work this paper recommends, on the cheapest inventory available. One practical catch on Google: using a customer list to target, rather than only to exclude or observe, requires an account with at least ninety days of history and more than fifty thousand dollars in lifetime spend. Many fair offices will not meet that threshold, so check eligibility before planning around it.
The place you would legitimately hold device identifiers is your own event app, where users have granted tracking permission. At most events that covers a small fraction of the gate and produces a worse list than your ticket buyers.
There is a harder point underneath all of this, and it is worth sitting with. Owning the fairground does not give you a relationship with everybody who stood on it. The walk-up who paid cash, never bought online, never gave you an email address and never downloaded the app is not somebody you have a relationship with in any sense a platform recognizes, even though they were on your property and handed money to your staff. That gap is exactly what geoframing sells into, and exactly why buying it is a purchase of observed information about strangers rather than an extension of your own records. A vendor can build you a list from your own grounds. It is still not your first-party data.
It is also the strongest practical argument for the survey described in Section 3. An intercept at the gate is how that otherwise invisible person becomes somebody you actually know something about, on terms they agreed to, in a record you own.
The specific limits are worth knowing before a vendor describes them to you.
So if the question is whether Facebook can fence your grounds and reach the people who attended, the answer is no. Not with a bigger budget, not with a better agency, and not with a setting somebody has not found yet. The capability is absent rather than restricted.
Having said all of that, both platforms do the one thing this paper actually recommends, and they do it cheaply.
Postal code targeting is a standard setting on both, and it gives you something a radius does not: a shape that follows real boundaries rather than a circle drawn around a pin. Layer household characteristics on top and you are running the profile approach described in Section 9 on the largest and least expensive advertising inventory available. It works at neighborhood scale. It does not work at venue scale, and nothing on these platforms does.
That produces an inversion worth holding on to. The technique this paper warns against, device-based historical targeting, requires a specialist vendor, a premium price, and a state-by-state legal check. The approach it recommends runs on platforms your marketing coordinator already has a login for.
Campaign construction, creative, sequencing, and budget across all of these platforms are the subject of the second white paper. What matters here is knowing which of the four things in Figure 1 each platform can actually do, so that a vendor conversation starts from the right place.
Messaging people through an app while they are physically inside your grounds is a real capability. It is not this one. That version requires attendees to have installed your event app, to have granted it permission, and to be on site at the time. It sends an actual notification. It reaches only the people who downloaded the app, and it stops working the moment they leave. It gets called geofencing too, which is a large part of why this subject is confusing. It is the subject of the third white paper, and nothing in the first or second depends on it.
When a vendor pitches you, the first question is not what it costs. It is whether they are selling you information about a crowd or access to a crowd. Those are different purchases and they solve different problems.
A large number of ordinary smartphone applications collect location in the background. Weather apps, games, navigation tools, coupon apps, and many others. When a user taps “allow location access,” a meaningful share of those apps pass that location information along to data companies that purchase it.
The result is an enormous pool of timestamped records. Each one says, in effect, that a particular anonymous device identifier was at a particular set of coordinates at a particular moment. There are billions of these records, covering populated places across the United States, though the density varies and rural areas are represented more thinly than metropolitan ones. That thinness matters directly to a county fair, which usually sits in exactly the territory where coverage is weakest.
Nothing was installed at your fairground to make this happen. No agreement was signed. The records came from your attendees’ own phones, by way of applications that have no connection to your event.
Because the vendor already holds location records covering places across the country, they had your audience’s data before you called them. When you request an audience, you are not commissioning collection. You are asking them to filter records that already exist: show every device that was inside this shape between these hours on these dates.
That is a database query. It does not touch your ticketing system, your Wi-Fi, your app, or your staff. It requires no permission from you and generates no notification to you. The same is true of every other event in the country, which is the mechanism behind Section 8.
The pool is never everyone who attended. It is only the people whose phones happened to be sharing location with a participating application at the time, and that share has been declining since mobile operating systems began prompting users to approve or deny tracking.
This produces the single most common and most expensive misunderstanding in this field. A vendor reports an audience of fifty thousand devices. The organizer hears fifty thousand attendees. Those are not the same number, and the gap between them is not disclosed unless you ask.
When a vendor quotes an audience size, ask two questions in writing. First: what percentage of actual gate does this figure typically represent for an event of this type? Second: how has that percentage moved over the past three years? A vendor who cannot answer either question is not one you should be buying measurement from. A vendor who answers both honestly is worth paying more for. This matters twice over. It affects what you pay for media, and it affects what you can credibly promise a sponsor.
Some organizers, on hearing the above, decide they want no part of it. That is a defensible position and worth stating rather than hiding, because your attendees may reach the same conclusion.
It is worth noting what is and is not happening. The data is anonymized in the sense that it carries a device identifier rather than a name. It is also, in practice, precise enough that a device that spends every night at one address and every weekday at another is not difficult to characterize. Legislators have noticed this, which is the subject of Section 5.
People talk about location data as though it were one thing you either buy or do not buy. It is five things. They come from different places, cost different amounts, answer different questions, and carry different risks.
Getting this wrong is the most expensive error in the field, and it is usually made at the first purchase. An organizer buys a device list expecting audience insight, or buys an analytics subscription expecting an advertising audience, and concludes the whole category is oversold.
A panel of phones is observed over time. A platform draws a boundary around your grounds, counts how many panel devices entered it, and scales that count up to estimate real attendance. It then reports where those devices spend their nights, which becomes a map of where your audience lives.
The resolution is finer than most organizers expect. Leading platforms report the trade area not as a radius or a list of counties but as the specific census block groups producing visits, a block group being a few hundred to a few thousand people. The same platforms can report where visitors went directly before and after your event. Route-of-travel analysis, where you need it, is generally a separate mobility product.
Everything is reported in groups. You see that a few thousand visits originated from a given block group, not that a given person attended. Protections are typically built in so that small groups cannot be isolated.
What it answers. Who comes, from which block groups, how far, how often, where they went before and after, and what else they do. Whether any of that has changed since last year.
What it cannot do. Reach anybody. There is no advertising here at all.
Smartphones can expose an anonymous number that advertising systems use to recognize a device over time, subject to the permissions the owner has granted. Vendors may hold historical records tying some of those numbers to places and times. Query a location and a time window and the result can become an advertising audience.
This is the raw material of geoframing, and it is the kind of location data that can let you re-reach people whose devices were observed at a specific event. How many of them you actually reach depends on match rates and permissions, which Section 4 takes up.
What it answers. Nothing. It is not an analytical product. It is a list.
What it cannot do. Reach far into the past, capture the majority of your gate, or survive a phone upgrade. It is also the kind now being restricted by state law.
This one is not a separate purchase so much as a layer on the one above it. A device identifier on its own reaches only that phone. The linkage step connects it to a home: vendors observe where a device settles at night and treat that as the household, then match it to that home’s internet connection and, in some cases, its postal address. Without a device identifier there is nothing to link, so everything that constrains kind two constrains this as well. The twelve-month wall, the commercial floor, and the state restrictions all pass straight through.
This is what turns a phone list into something you can use on a television or in a mailbox. It is also the least accurate link in the chain, and Section 4 gives it the scrutiny it deserves.
What it answers. Roughly where an attending household lives.
What it cannot do. Exist at all without kind two. It also cannot be relied on at the individual level: the error rate is meaningful and rarely disclosed.
Not location data in itself, but the lens that makes location data mean something. Commercial systems sort every United States household into behavioral categories using demographics, spending, media habits, and movement. One widely used system classifies households into eighty segments grouped into seventeen families.
Unlike household linkage, this one is not derived from anything else on the list. It is an independent classification that exists whether or not anybody ever visits anywhere, and you can buy it and apply it to a list of postal codes with no location data in sight.
What it does need is a subject. On its own it tells you what kinds of households exist and where they live, which says nothing about your event until you attach it to something that identifies your audience. There are two things it can attach to, and this is worth knowing before you price anything. The obvious one is aggregated visitation: who came, described in segments. The other is your own first-party data. Append segments to the postal codes on your ticket orders and you get the same kind of description of your audience, at higher accuracy for the portion of the gate those records cover.
That second path matters to any organizer who cannot justify an analytics subscription. It produces a usable audience profile from data you already hold, for the cost of the segmentation alone. Section 11 sets out the order to work in.
Attach this to visitation data and a list of counties becomes a description of people. It is the difference between knowing your audience comes from three counties and knowing what kind of household they keep.
What it answers. What sort of people these are, in terms an organizer can act on and a sponsor will pay for.
What it cannot do. Identify anyone. It describes categories, not people, which is precisely why it is durable.
Ticket purchases with postal codes, gate scans, Wi-Fi registrations, app installs, contest entries, email signups, and surveys. Most organizers hold more of this than they realize and analyze less of it than they should.
It is the only category on this list collected directly through your own relationship with attendees and held under your control. It is considerably more durable than purchased device identifiers and costs nothing per use once collected, though it remains subject to your own privacy notices, your consent basis, your retention and deletion practices, platform terms, and applicable law. Its weakness is that its coverage is whatever you built: a transaction record reaches only the people who transacted, which at a fair with a substantial cash gate is a fraction of the crowd.
What it answers. Many of the same questions the other four answer, often at higher accuracy for the portion of your audience you actually know. Surveys add the one thing no location dataset can supply directly, which is why people came.
What it cannot do. Cover more of your gate than you actually collect from or sample. Its reach is whatever you built, not whatever happened.
Every other kind of location data in this section describes behavior. Where people came from, how far they traveled, how long they stayed, what else they do. None of it explains why any of that happened.
A survey does. It is the only instrument on this list that can tell you somebody drove two hours for one act, that they nearly did not come because of parking, that they would have stayed a second day if there had been a reason to, or that they heard about you from a neighbor rather than from anything you paid for. That last answer alone can change a media plan more than a trade area map will.
Surveys also reach the people every other method misses. An intercept survey at the gate captures the walk-up who paid cash, never bought online, never downloaded the app, and appears nowhere in your records. That group is invisible to your ticketing system and underrepresented in device panels, and at a fair with a substantial cash gate it can be a large share of your audience.
The cost is discipline rather than money. A survey answers only the questions you thought to ask, it reaches only the people willing to stop, and people who agree to be surveyed at an event are not a random sample of the people at it. Treat the results as directional, ask the same core questions every year so the comparison means something, and keep it short enough that people finish it.
There is a category of location data better than any of the five above, and an organizer will never be offered it. It is worth understanding anyway, because it explains why this industry runs on the data it does, and because it shows where the rest of this is heading.
A mobile network necessarily holds location information about the handsets connected to it, because devices communicate with the network periodically in order to be reachable. No application is involved, no permission is granted, and no panel is sampled. The coverage is the carrier’s own subscriber base rather than the shrinking share whose phones happen to be sharing location with a participating app. It is less precise than satellite positioning, and in rural areas considerably less precise, but on coverage nothing else is close.
Carriers did sell access to it, through the same aggregator arrangement described in Section 2. Federal regulators found that each of the major carriers sold access to customer location information to aggregators who resold it downstream, and that the carriers attempted to push the consent obligation onto those downstream recipients, which in many cases meant no valid consent was ever obtained. The data reached bail bond companies and bounty hunters.
In April 2024 the four largest carriers were fined a combined total approaching two hundred million dollars under the section of federal communications law that governs customer network information. The litigation is not finished. AT&T and Verizon carried a constitutional challenge to the Supreme Court, which in June 2026 upheld the Commission’s in-house forfeiture process on narrow grounds, holding that such an order is not a final judgment and becomes collectible only if the Justice Department sues to enforce it. Sprint and T-Mobile lost separately in the D.C. Circuit and have a petition pending before the Supreme Court as this paper goes to press, raising the different question of whether the location information falls within the statute at all. One carrier states that it discontinued its location-based services program more than six years ago.
The commercial market for carrier location data, for marketing purposes, is effectively closed. What remains available legitimately is heavily aggregated mobility data sold for transport and planning work, and a carrier advertising to its own subscribers within its own inventory. If a vendor offers you carrier data or carrier-grade accuracy, ask precisely what they mean, because the device-level version of it is the thing that drew the fines.
The carrier story is the arc that device identifiers are somewhere in the middle of. Sold through aggregators, exposed by journalism, investigated, fined, and withdrawn from the commercial market. Section 5 asks you to plan on the rules narrowing. This is the same story, already much further down that regulatory path, and it is a better guide than any forecast.
Yes, by three routes with three different standards, and organizers ask about this often enough to be worth stating plainly.
Historical cell site location information of the kind at issue in Carpenter generally requires a warrant. The Supreme Court decided that in 2018, and decided it narrowly: it expressly left real time cell site information and tower dumps for another day.
Platform records can be compelled by what is called a geofence warrant, which orders a provider to identify the devices its own location dataset indicates were present within a defined area during a defined window. The geometry is similar to historical geofencing, place plus time, but the source data and the legal mechanism are different. In June 2026 the Supreme Court held for the first time that acquiring a provider’s location history this way is a Fourth Amendment search, reasoning that it implicates greater privacy interests than the cell site data at issue in the 2018 case. The Court sent back the separate question of whether that particular warrant satisfied probable cause and particularity. Some providers had already engineered around the problem by moving location history onto the device, and at least one now reports that it is no longer capable of answering such a request.
Commercially purchased data is the unsettled one. Neither ruling addressed whether the government buying brokered location data on the open market is constitutionally equivalent to compelling a company to produce it. Both concerned compelled production. The practical effect is that a question about probable cause becomes a question about budget, and federal agencies have made use of it.
For an organizer the point is not the constitutional argument. It is that commercial location records of the same general kind you would buy to analyze your own event can also be purchased by government agencies, and that a public gathering is exactly the sort of location these requests target. That is worth knowing before you tell an attendee, or a reporter, that the data is anonymous.
Aggregated visitation data plus household segmentation. One tells you where your audience comes from, the other tells you who they are, and together they answer the questions that change decisions. Neither one delivers an advertisement. That is the subject of the second white paper, and it is a smaller problem than it appears once this pairing has told you where to point.
Six properties determine whether a given kind of location data is worth buying for a given purpose. Vendors discuss the first and rarely volunteer the rest. The last one disqualifies most events outright.
Event-focused historical geofencing products currently cap the lookback at roughly one year, and at least one imposes a floor as well, declining to query anything more recent than thirty days because the data needs time to settle. Aggregated visitation platforms can retain history over a much longer horizon.
The practical consequence is the answer to a question organizers ask constantly. With a one-year device product you cannot combine three past years of your own event into one device audience, because two of those years sit outside the window. You can combine three years of visitation analysis wherever the platform retains that history. Section 9 develops what to do with that.
No kind of location data sees your whole gate. Device-based products see the share of attendees whose phones were passing location to a participating application at that moment, and that share has been falling since operating systems began asking users for explicit permission. Aggregated platforms observe a panel and scale it upward, which introduces estimation error but does not pretend to completeness.
Either way, the number a vendor quotes is not attendance. Insist on knowing the relationship between the two before you build a plan or a sponsorship promise on it.
This is the property that decides, for most readers of this paper, whether the rest of the discussion is academic. Vendors will not run a device-based campaign below a floor, because very small audiences deliver unpredictably and cannot be reported on meaningfully. One vendor serving the event market publishes a floor of fifty thousand unique devices.
Read that number carefully, because it is counted in devices and not in attendance. A device count is a fraction of gate, as this section has already established, so an event needs meaningfully more than fifty thousand people through the gate to produce fifty thousand queryable devices. Nobody will volunteer the multiplier. Ask for it.
The practical consequence is blunt. For a standalone county fair, a community festival, or a single-day event, device-based targeting is not expensive. It is unavailable. The vendor will decline the business.
Three things change that answer. An organization running several events inside the lookback window can bundle them into one audience, which is the vendor’s own suggested remedy. An organizer willing to include comparable events they do not own can reach scale that way, at the cost of the audience no longer being their crowd. And the floor is one company’s commercial policy rather than a law of physics, so it is worth shopping, though some floor will apply wherever you go.
Large single events, portfolios, or nobody. That is the honest summary, and it is worth establishing before a vendor spends an hour describing a product you cannot buy. Note that it disposes of household linkage at the same time. Linkage is built from device identifiers, so an event that cannot clear the floor for the second kind cannot reach the third either, and with it goes the streaming television and direct mail that linkage makes possible.
Household linkage deserves particular scrutiny, because it is what enables the most impressive-sounding capabilities and it is the least reliable step in the chain.
One measurement firm has reported that the connections between an internet address and a postal address hold up only a small fraction of the time, and estimated that a substantial share of money spent in the open streaming television market goes to inaccurate identity data. That finding was published by a company selling a competing approach, so weigh it accordingly, but the direction is corroborated by the identity industry’s own commentary on falling match rates.
The rule that follows is simple. The further a tactic sits from the phone itself, the more waste is priced into it. Observing that a device was at your fairground is reasonably solid. Concluding which house it went home to is much shakier, and anything built on top of that conclusion inherits the error.
One vendor in this market advertises that roughly ninety percent of people have location services turned on and that no additional hardware is required. That figure is difficult to reconcile with what identity resolution companies report about declining match rates following the introduction of tracking permission prompts. It is a useful example of the genre. When a capture or match figure sounds too clean, ask how it was measured, on what population, and in what year.
This section sits in Part One rather than in an appendix because the law now determines what is available before budget does. An organizer who plans a program and then checks the law will waste the planning.
State privacy legislation has moved quickly, and it has moved most directly against precise device-level location and the identity links built on top of it. Those are the second and third kinds described in Section 3.
Aggregated visitation, segmentation, and your own first-party data carry lower exposure at the organizer-facing end, which is not a coincidence and is worth understanding. They are not outside privacy law altogether. The location data feeding an aggregated panel is still regulated upstream, and your own first-party data carries its own consent and retention obligations. The difference is one of degree, and it is a large one.
Virginia prohibited the sale of precise geolocation data effective July 1, 2026, defining precise as identifying an individual within roughly 1,750 feet. Oregon’s prohibition on the sale of precise geolocation took effect January 1, 2026. Maryland’s Online Data Privacy Act took effect October 1, 2025 and goes further than either: it prohibits the sale of sensitive data outright, and permits processing it only where strictly necessary, regardless of whether consent was given.
California has additionally prohibited certain fencing around in-person health care entities and has opened an investigative sweep into the location data industry. The practical consequence is that vendors have begun switching individual products off state by state rather than defending them.
Some of what you will be told is state law is actually vendor policy. Connecticut, Washington, Nevada and New York restrict geofencing around health care facilities by statute, at distances such as 1,750 feet in Connecticut and 2,000 feet in Washington. Vendors then layer their own rules on top. One publishes a half-mile minimum radius around medical facilities in all four states and will not measure visits to them at all. The statutes set the legal floor; the half mile is that vendor’s risk posture.
The distinction matters when you shop. A restriction that comes from a statute applies to every vendor. A restriction that comes from one company’s risk posture may not apply to the next one you call, and is worth asking about specifically.
Two vendors serving the event market have published state-level restrictions on their own products. One has withdrawn historical audience building and offline visit measurement in Virginia, Connecticut, and Washington while continuing to offer real-time fencing there. Another cannot run historical geofencing in Virginia or Oregon and cannot offer geofencing of any kind in Connecticut.
Texas deserves a specific note, because it is often described in this market as an open state and that is not accurate. The Texas Data Privacy and Security Act has been in force since July 2024, and it classifies precise geolocation as sensitive data, using the same 1,750-foot definition. Processing sensitive data requires consent, and the Attorney General can seek civil penalties per violation. What Texas does not have is the categorical prohibition on the sale of precise geolocation found in Virginia, Oregon, and Maryland. The practical position is that vendor availability is broader in Texas today, not that the activity is unregulated.
If your event draws from multiple states or you promote a touring property, you need a state-by-state execution map maintained by your vendor and reviewed before each campaign.
The enacted changes discussed here have moved toward tighter restrictions. Plan on the assumption that device-based audience building will become unavailable in a growing number of states through 2027 and beyond. Section 3 describes a category much further along that path: carrier location data was sold, investigated, fined, and withdrawn from the commercial market. That is precedent rather than forecast.
This paper describes the United States market. The mechanics in Sections 1 through 4 travel across the border unchanged, because phones, panels, and advertising auctions work the same way everywhere. The law does not travel, and Canadian organizers face a materially different picture.
Federal law is the Personal Information Protection and Electronic Documents Act. Its application to a business outside Canada turns on the facts of the case rather than on a simple test, but a United States promoter marketing into Canada should assume it is in scope and take advice rather than assume it is not. The federal privacy commissioner has moved toward requiring meaningful rather than implied consent for tracking used for advertising and profiling.
Quebec is stricter again. Under Law 25, profiling, geolocating, and tracking functions must be switched off by default and require the individual to turn them on, and precise geolocation is treated as sensitive information. Alberta and British Columbia maintain their own private sector statutes.
Enforcement differs sharply between levels. The federal commissioner works through investigation, published findings, and referral to the Federal Court rather than direct fines. Quebec’s regulator has direct fining authority reaching into the tens of millions of dollars.
None of this makes Canada a harder market for the approach this paper recommends. It makes it a better one.
Canadian postal codes are far more granular than United States postal codes, covering roughly a block face rather than a district, and the Canadian segmentation standard assigns lifestyle types at that level. A Canadian organizer working from an audience profile has finer geographic resolution than an American one, before touching any device data. That segmentation is built from anonymized and aggregated sources, which generally places it further from the statutes above than device-level targeting sits, though privacy law can still apply upstream depending on how the source information was collected and processed.
The Canadian system also carries something no United States equivalent offers, which is language. It includes a substantial set of francophone segments and a separate Quebec-optimized version, which matters to any exposition drawing across a language boundary.
So Canada makes the distinction in this paper more important rather than less. Audience profiling built on aggregated geography and segmentation generally presents a more durable compliance path than individual device-level targeting. Historical geofencing does remain commercially available from some vendors in Canada, and organizers should not read that availability as establishing compliance. Consent, source data, provincial law, and the vendor’s own collection practices all still need review.
Confirm coverage directly. Several of the analytics platforms named in Section 6 are built primarily around United States data, and Canadian coverage should never be assumed from a vendor’s general marketing. At least one provider does publish Canadian coverage reported at dissemination area level, which is the unit you want, so the capability exists and is worth asking for by name. Ask specifically which Canadian segmentation system a platform carries, whether it reports at postal code and dissemination area level, and whether a Quebec-specific version is included.
Four categories of company sell the five kinds of data. Most organizers meet them in the wrong order, starting with whoever called first.
The companies named below are examples chosen to illustrate the categories. They are not recommendations, TSE has not evaluated their service, and inclusion or omission should not be read as a judgment.
Subscription software that reports who visits a place. This is where aggregated visitation data and, usually, the segmentation layers are purchased together.
Placer.ai is a category leader for trade area and visitation analysis, reporting at census block group resolution. Its marketplace carries more than twenty datasets, including Experian Mosaic, Spatial.ai PersonaLive, AGS, STI and census data, licensed as paid add-ons and used inside the Placer dashboard rather than through a second vendor. Published public-sector contracts show pricing running from the low five figures into the tens of thousands of dollars depending on scope, which can still be a genuine barrier for a county fair.
PassBy offers tiered plans aimed at teams that do not need a full enterprise deployment. Current pricing should be quoted rather than assumed from a published tier structure.
Unacast, Foursquare and Azira offer mobility and place-intelligence products ranging from data feeds and APIs to analytical platform interfaces. Their more advanced products generally assume more analytical capability than a typical fair office has on staff.
Advan descends from the SafeGraph visitation methodology and rebuilt it with its own geofences and full population normalization, so counts are scaled to real foot traffic rather than left as raw panel devices. It reports visits, dwell time, and the home block group of visitors for individual locations across the United States and Canada, with several years of history. Its buyers are investment funds, real estate firms, and public sector teams rather than marketers, which shows in the product: it sells data feeds, not a dashboard for a marketing coordinator.
StreetLight handles vehicle and traffic movement, priced to the scope of a study. Relevant if parking and road access are your questions.
Two different products are sold under the same heading, and the distinction decides which one you should be looking at. Some of these are dashboards, built so a marketing coordinator can draw a boundary and read a report the same afternoon. Others are data feeds, built so an analyst can pull records into a model. The second kind is often more granular and more flexible, and almost always harder to use.
A fair office with no analyst wants the first kind. An organization that already produces economic impact reporting for a municipality, or that has a research partner at a nearby university, may get more from the second and pay less for it.
These attach to the platform above rather than standing alone, and they are what turn geography into a description of people.
One point of confusion worth clearing up, because it changes the budget. You do not generally need a second platform to get segmentation. The leading visitation platforms run a marketplace, and the household classification systems are licensed through it as paid add-ons and used inside the same dashboard. One vendor relationship, one login, one invoice with line items on it.
The exception is activation. Reading which segments visited your event is an add-on. Pushing those segments out to an advertising platform as a targetable audience is a feature of the segmentation provider’s own product, and that is a separate purchase. Analysis and activation are priced differently, and most organizers only need the first.
These sell device identifiers and household linkage. They are the category most affected by Section 5, and the one where written state availability matters most.
The fourth category is the one you already pay for. Ticketing platforms, gate scanning systems, Wi-Fi providers, and event apps all hold first-party data, and several event-industry platforms bundle ticketing with marketing tools.
Before subscribing to anything on this page, find out what your ticketing provider already reports. A postal code on every online order is a trade area analysis you own, and a surprising number of organizers have never asked for the export.
First, find out what your ticketing and gate systems already hold. It is free and you own it. Second, buy visitation analytics with a segmentation layer. This is the pairing that changes decisions. Third, and only where the law and the arithmetic both permit it, add device-based advertising. That is the subject of the second white paper.
Analytics platforms report a trade area in two forms, and the difference between them is the most actionable output in this paper.
Laid over one another, the two produce a penetration map. Block groups where potential is high and capture is low are underpenetrated candidates for investigation. They may be growth territory, but competition, travel barriers, price, lineup fit, and simple lack of awareness can all explain the same gap, and the map cannot tell you which. Its value is that you now hold a specific, mappable list of places to look into rather than a hunch about which suburb to add to the radio buy.
It also disciplines the growth conversation. An organizer who wants to grow attendance by ten percent can see which block groups would have to move, and whether that is plausible given how many people live there. Some growth targets die quietly at this step, which is cheaper than discovering the same thing after a season of media spend.
Nearly every analysis turns up a pocket of attendance from somewhere unexpected. A town on the wrong side of a county line, a suburb of a city you do not advertise in, a corridor along a highway you never considered. These are the cheapest growth available, because something is already working there without your help.
The inverse is more useful and less comfortable. Territory inside your natural radius that produces little attendance is either being served by something else, blocked by a travel barrier, or simply never told. Each of those has a different answer and the map will not tell you which. It tells you where to look.
Comparing years is where this becomes a management tool rather than a curiosity. A trade area that contracts while attendance holds steady is an event becoming more local, which is a warning even in a good year. A trade area that expands while attendance is flat means something is working and something else is failing.
One analytics vendor publishes a worked comparison of a major urban music festival across two consecutive years. The earlier year drew from further out, from households with higher incomes and a greater share of families. The following year the trade area contracted. Nothing in the attendance figures would have shown that. It is visible only in the shape of the draw, and it is the kind of finding that should change a lineup decision rather than a media decision.
If a subscription is out of reach this year, an approximation is available immediately and costs nothing. Export postal codes from your online ticket orders and map them. It covers only the portion of your gate that bought online, it will over-represent advance buyers and under-represent walk-ups, and it will still tell you more than you currently know. Section 11 sets out how to do it step by step, and what free census data adds about the people in those places.
Do this before deciding whether the paid version is worth it. It is also the honest way to judge a vendor demonstration, because you will already know roughly what the answer should look like.
The same analysis can be run on any location in the country. This is where the tool stops being descriptive and becomes strategic.
You can profile a competing event’s audience without attending it, without surveying anyone, and without their cooperation. No permission from the event operator is required and no notification reaches the event operator, for the reasons set out in Section 2.
This is the application that matters most to a buyer, and it is the one nobody sells you. Every other use of location data in this paper looks backward at an event you already produced. Booking does not work that way. The question is almost always about an act you have not had before, and the thing you need to know is what they draw.
The method is the same one described above, pointed at a different place. Draw a boundary around the venue where that act played, over the hours of their show, and read the crowd that turned up. How far did it come. What sort of households. Whether it was local or travelled.
Then hold that against your own trade area and ask the only question that matters before an offer: does the population you would need actually live within reach of your grounds? An act can be genuinely strong and still be wrong for your market, and that is an expensive thing to learn afterward.
One refinement separates this from guesswork. A venue has its own draw whoever is standing on the stage. A regional theatre pulls from its region on an ordinary night. So the figure you want is not the crowd at the act’s show, it is the gap between that night and that venue’s typical night. The difference is evidence of incremental draw associated with that show. Isolating the act still requires controlling for the other things that changed, such as ticket price, date, promotion, weather and competing events.
Three limits, and they are the same ones that constrain reading your own event. A multi-act festival date tells you nothing about any single performer on it, so look for single-headliner nights. Markets differ, and a draw that reaches two hundred miles in one region may not transfer to another. And the act has to have played somewhere readable, recently enough to sit inside the platform’s history.
The value compounds in a way a single study does not suggest. One venue on one night is a data point. The same act across four venues, or the same act two years apart, is evidence. Anyone who reads acts this way over several seasons ends up with something more useful each year than any single study, so keep every analysis rather than treating each one as a one-off.
Every organizer reading this should understand that the capability works against them exactly as well as it works for them. There is no technical defense and no notification.
What there is, is a strategic response: know your own audience better than an outsider querying a database can. An outsider sees where your attendance comes from. You can see that plus what they bought, when they bought it, what they came back for, and what they told you at the gate. That combination is not purchasable, and it is the reason Section 3 ends on first-party data rather than beginning with it.
Some organizers find this practice distasteful when it is done to them and sensible when they do it. That inconsistency is worth resolving before building a program around it. The practice is currently legal in most states and widely used across many industries. It is also, plainly, a use of information that attendees did not knowingly provide to you. Decide where your organization stands and apply that standard in both directions.
This is the strategic argument of the paper, and for an annual event it decides the approach.
Almost everything sold in this category is a device list. It works, it is the basis of most of the second white paper, and it has two properties that make it a poor foundation for a long-term program.
Section 5 covered this. The specific practice being restricted is the sale of precise individual location. A strategy whose foundation is state-by-state availability carries a maintenance burden and an uncertain horizon.
An audience profile describes the kind of household your attendance comes from. Where they live, what they earn, how they behave, what else they do. It is expressed in segments and geography rather than in device identifiers.
A profile does not expire when someone buys a new phone. It is generally not the direct target of statutes aimed at precise individual location, because it does not track individuals. And critically for an annual event, it compounds. Each year of analysis makes it more reliable rather than replacing the last one.
This is the specific answer to the question organizers ask most often, which is how to combine several past years of attendance into one audience. You cannot do it with device lists, because two of those years sit outside the lookback window. You can do it with profiles.
The result reaches scale without device identifiers, without lookback windows, and without the same state-by-state product-availability map required for precise device-based targeting. For an organizer planning a program to run through 2027 and beyond, this is the version worth building.
The reasonable objection to everything above is that a profile is only worth as much as your ability to act on it. Knowing your audience is a particular kind of household in a particular set of counties is of no use if you cannot buy advertising against that description.
You can, and it is the cheapest advertising in this paper. Every major platform accepts postal code targeting as a standard setting, and the independent buying platforms go finer, down to the level of a census block group. The caution from Section 1 applies: a platform can only act on the location signal it has for a given ad request, and on many requests that signal resolves no better than a town. Treat postal code targeting as reliable at neighborhood scale and plan clusters rather than single codes.
So the profile is not stranded. It converts into a list of geographies weighted by how much of your audience each one produces, layered with household characteristics, running on inventory that costs a fraction of specialist location media and carries far less of the precise-location exposure described in Section 5. The analysis is delivered at block group resolution, which is finer than the postal code most platforms will accept, so the constraint is the buying platform rather than the data.
One practical caution. Delivery degrades as the target tightens. Platforms warn that very small geographic targets can reduce match rates and cause advertising to run intermittently or not at all, and the standard remedy is to combine neighboring postal codes rather than isolating one. That suits an event well, since a trade area was never a single postal code to begin with.
Build the profile every year, whether or not you activate it immediately. It costs analysis time rather than media budget, it appreciates rather than decaying, and it is what remains when the other track closes. Run device-based targeting opportunistically, where it is legal, where the audience clears the vendor minimum, and where the venue geography cooperates. Treat it as a supplement with a defined test, not as the foundation.
Everything to this point has treated location data as a marketing instrument. Its most valuable output is not a marketing output at all.
The same analysis that tells you where to advertise tells you which days and which parts of your lineup drew people, and from how far. That is a booking question, and it is worth considerably more than a ticket sale, because it recurs every year and compounds across your entire buying history.
Location data measures days and places, not performers. On a bill with six acts on one stage, the day’s draw belongs to the whole bill, the weather, the ticket price, and whatever else was happening that weekend. Anyone who tells you the data isolates one artist on a crowded day is overselling it.
Attribution becomes possible where an act can be separated from everything around it. A single-headliner night. The same act booked in two different years. A day whose lineup is otherwise unchanged from the prior edition. Those comparisons carry real evidentiary weight. The rest is directional.
An artist’s asking price is generally built from streaming figures, social following, prior guarantees, and comparable market performance. Those are national or regional signals. None of them describe what the act did for your event in your market.
An organizer who can demonstrate that a returning act drew from a narrower radius than the year before, or that a day with an otherwise unchanged bill drew wider than the year its headliner changed, is negotiating from evidence rather than impression. That does not always change the number. It changes the conversation, and over several buying cycles it changes which offers get made at all.
It also protects against the more expensive error, which is not overpaying for one act. It is repeating a booking pattern for several years without knowing whether it is working.
TSE Entertainment has represented event buyers since 1975. We sit on the buyer’s side of the table exclusively, and our fee is built into the deal structure rather than added on top of it. An organizer who knows where their audience comes from, and what kind of households it is, makes better booking decisions and negotiates from a stronger position. If you are building a lineup for 2027, bring what you have learned about your audience to the conversation. We would welcome it. tseentertainment.com
Year one splits into work that costs nothing and purchases that need a budget. Do the first four steps whatever your budget is.
If you have funding, add these in order, and stop wherever the budget does.
If there is no funding this year, steps one to four still leave you with most of what a sponsor asks for and a sound basis for marketing. Step five sharpens who. Steps six and seven add the people your records and surveys still miss, cross-visitation, and other venues. For a talent buyer, step seven is where visitation stops being optional, because it is the only way to read an act before you book it.
None of the work below costs anything but staff time, and together it produces a rough version of the audience profile this paper argues for. Do them before you take a single vendor meeting. You will know where your audience lives, how far it travels, what kind of places it comes from, and where you are underperforming, and you will be able to tell whether a paid study would change any decision you would otherwise make.
Start with your ticket export. Ask your ticketing provider for every online and advance order with its billing postal code, the event date, the number of tickets and, where available, the ticket type. Ask for one row per order rather than a summary report, in a spreadsheet or CSV file. Include complimentary and cancelled orders in the export, flagged as such, so you can decide what to remove rather than having it decided for you.
Fill the gap with surveys and registrations. Your ticket export describes the people who bought in advance. At most fairs that leaves out the cash gate, and at a free festival it may leave out everyone. Two sources close that gap. The first is an intercept survey at the gate: a few volunteers with phones or tablets, and a survey tool that works offline, since cell service often fails in a crowd. Ask for the home postal code first, then party size, first visit or repeat, how they heard about the event, why they came, and which other fairs, festivals or concerts they attended this year. That last question is a consented, first-party measure of the competitive overlap Section 8 describes. Spread interviews across every day, the busiest and quietest hours, and every entrance, because a survey taken at one gate on one afternoon describes that gate on that afternoon. A few hundred completed surveys is a useful target for most events. The second source is every registration you already collect: contest entries, email and text signups, Wi-Fi logins, app installs, and sponsor or vendor lead forms. Each one carries a postal code, or can be made to at no cost by adding the field. Say plainly on the form how the information will be used.
Clean the data before you count it. Most errors in a do-it-yourself trade area come from the spreadsheet, not the method, and they are easy to prevent. Put each source on its own tab, add a column naming the source, and give every tab the same columns: source, postal code, event date, and count, meaning tickets on an order or party size on a survey. Then work through the following.
Combine the sources without double counting. Tickets and surveys measure different things, so do not simply add them together. Ticket orders are counts of attendance. A survey is a sample. Estimate how much of your attendance did not come through advance tickets, which is your gate count minus scanned advance tickets, or at a free event your total attendance estimate. Then spread that figure across postal codes in the proportions your survey respondents reported, weighting each response by party size. Add the result to your ticket totals and you have one column: estimated attendance by postal code. Registration lists are for checking and filling gaps, not for adding attendance, because a contest entrant may already be a ticket buyer or a survey respondent. Where a registration list shows a postal code that neither tickets nor surveys picked up, note it as a lead to investigate.
Turn your cleaned file into a map. In the master tab, total estimated attendance by postal code and sort from largest to smallest. The smallest group of codes that together account for about two thirds of your audience is your core trade area, and everything beyond it is your reach. Add the driving distance from your grounds to each code and group the totals into bands, such as under 25 miles, 25 to 50, 50 to 100 and beyond, and you have a draw radius built from your own records. For a picture, a spreadsheet map chart or the census tools below will shade each postal code by attendance. Because every row carries an event date, you can repeat the exercise by day and have the day-by-day comparison Section 10 describes.
Know what the map leaves out. Ticket records place each order where the card is billed rather than where each attendee lives, so one order may be a family of five or a church bus. Survey respondents are the people willing to stop, not a random sample, and the walk-up estimate is only as good as your gate count and how evenly you spread your interviews. Treat the result as a map of your identifiable audience, not of your whole crowd, and read it alongside your gate count.
Describe the places, using census data. The Census Bureau’s American Community Survey publishes household characteristics for every ZIP Code Tabulation Area, free at data.census.gov. For each of your postal codes you can see median household income, the age mix, household size and the share of households with children, educational attainment, how many own rather than rent, vehicles available, typical commute and language spoken at home. Weight each area by its share of your estimated attendance and you have a description of the kind of places your audience comes from, in the terms a sponsor recognizes. Two cautions. Census tabulation areas approximate postal codes rather than matching them exactly, so a few codes will not line up. And this describes areas, not your audience: a postal code with a median household income of seventy thousand dollars still contains households well above and well below it.
Find where you are underperforming. The same tables give the number of households in each area. Divide estimated attendance by households and you have a penetration rate for every postal code. Sort by it. Large, nearby areas with low penetration are candidates for growth, or at least for a hard question about why they do not come. This is the captured-against-potential comparison from Section 7, approximated from records you already own and public data.
Use a friendlier tool, still for free. The census website is free but unfriendly. Social Explorer puts the same data, and a wider library of public sources, behind a mapping interface: choose your postal codes and variables and it returns a shaded map and a table. Its tools include importing your own data, so your attendance counts by postal code can be loaded and mapped directly; analysis by radius or drive time around your grounds; masking everything outside your trade area; comparing census years to show where your territory grew or aged; and reports describing a chosen area as one place, which is a sponsor page in minutes. It comes in free, paid and enterprise plans. The free plan carries a limited set of data, enough to look around your area but probably not enough for this exercise. Many public and university libraries hold subscriptions their cardholders can use, so check yours first. The paid plan offers a free trial: have your cleaned master file ready before you start it, confirm how long it runs, and cancel before it ends if you do not intend to subscribe. At the time of review the paid plan ran about $135 a month billed annually. Its marketing and advertising materials also mention Consumer Mosaics segmentation, which would add the behavioral layer census data lacks; ask whether it is included in the plan you are quoted, and whether it is reported by area or by household, before counting on it.
What you hold at the end of this is a baseline. Every later purchase should be measured against it, and if a paid study would not change a decision you were going to make anyway, you have saved the money.
Repeat the analysis on the new season and compare. The second year is where this becomes a management tool, because a single year is a snapshot and two years is a direction.
This is also the point to add device-based advertising, if the law in your state permits it and your audience clears the vendor minimum. By now you know what your audience looks like, which means you can judge whether the device-based version is adding anything the profile does not already give you.
Three years of profile is where you can begin distinguishing recurring event draw from booking-related shifts, which is when the lineup analysis in Section 10 starts to carry real weight. That is the point at which this stops being a marketing expense and starts being an input to the buying decision.
Find out what you already own, buy the analysis that tells you who your audience is, and only then spend money reaching them.
Terms defined as they are used in this paper. Vendors sometimes use them differently, which is itself worth watching for.
Advertising identifier. An anonymous code assigned to a phone that lets advertising systems recognize it across apps. Users can reset it, and many do.
Aggregated data. Reported in groups rather than as individuals. The basis of visitation analytics and the reason it carries less legal exposure.
Captured market. A trade area weighted by actual visits. Who came.
Census block group. A census unit of roughly a few hundred to a few thousand people. The finest resolution most visitation analytics report, and finer than a postal code.
Cross-visitation. What other places a group of visitors also go. Useful for sponsorship and for choosing which events to study.
Device identifier. See advertising identifier.
First-party data. Information you collect directly through your own relationship with attendees, from ticket sales, gate scans, signups, surveys, or your own app. Surveys, where included, are the only method described here that can directly ask why somebody came.
Geofencing. Drawing a boundary and serving advertising to phones currently inside it. Also used loosely for at least three other things, which is the subject of Section 1.
Geofence warrant. A court order requiring a provider to identify the devices its own location dataset indicates were present within a defined area during a defined window. Similar in geometry to historical geofencing, different in source data and legal mechanism.
Geoframing. Identifying which phones were inside a boundary during a past window, then buying ordinary advertising slots against those phones afterward. No direct message is sent and no event app of yours is required.
Proximity messaging. Sending a notification through your own event app to attendees physically on site. Frequently also called geofencing, and unrelated to anything in this paper. See the third white paper in this series.
Geotargeting. Serving advertising to a defined geographic area, from a television market down to a single postal code or finer. Cheap, widely available, and generally not subject to the precise-location restrictions discussed in Section 5.
Household linkage. Connecting a phone to the home it returns to at night, and from there to that home’s internet connection or postal address.
Location intelligence. Aggregated reporting on who visits a place. Measurement rather than advertising delivery.
Lookback window. How far into the past a vendor will search for devices at a location. Commonly capped near twelve months.
Match rate. The share of a list a system can actually recognize and reach. Falling match rates are why device lists lose value.
Panel. The set of devices a visitation platform observes, from which it estimates total attendance.
Polygon. A custom shape drawn on a map, as opposed to a circle around a point. More accurate for irregular venues.
Potential market. A trade area weighted by the population living in it rather than by visits. Who could have come.
Precise geolocation. Location specific enough to identify an individual within a small radius. The specific thing several states now restrict.
Segment. A category of household grouped by shared behavior and characteristics. The unit of an audience profile.
Trade area. The geographic territory an event actually draws attendance from, measured rather than assumed.
Universal opt-out signal. A browser or device setting through which a person declines data sale. Several states now require it be honored.
Walled garden. An advertising platform that sells only its own inventory and will not accept an audience built by an outside vendor. Facebook and Google are the main examples.
ZIP Code Tabulation Area. The Census Bureau’s approximation of a postal code’s area, used to publish statistics. Close to the postal code but not identical, which is why a few codes will not line up.
Questions to put to any location data vendor in writing before signing. A vendor who answers these readily is a different proposition from one who deflects them.
Platform policies, vendor product terms, and state privacy statutes referenced in this paper change frequently. Every item below should be confirmed as current before it is relied on. Sources were reviewed September 21, 2026.