Why CTV attribution shouldn't stop at the click: measuring store visits, calls, and sales

KEY TAKEAWAYS

• Most CTV attribution was built on web logic: an ad runs, someone clicks, someone converts. For many multi-location and offline-heavy businesses, click-based attribution can miss the majority of CTV’s impact.
• The other 80% happens offline: someone walks into a store, calls the business, books an appointment. If your reporting can't see that, you're only measuring a fraction of the campaign.
• Offline attribution isn't guesswork. It runs on identity resolution, spatial geometry, call tracking, and holdout testing—the same rigor as digital attribution, applied to the real world.
• Knowing what offline attribution can prove, and what it can't, matters just as much as building the infrastructure in the first place.

CTV advertising means running video ads on internet-connected TVs (smart TVs, streaming sticks, set-top boxes) through free ad-supported platforms and subscription services. It looks like TV: full-screen, sound-on, unskippable. It behaves like digital: targetable by audience, measurable at the household level, buyable programmatically.

When platforms first needed a way to measure CTV performance, they reached for the closest existing model: web attribution. An ad runs, a viewer clicks a link or lands on a website, a pixel fires, credit gets assigned. It's a clean, familiar system. It's also built for the wrong medium.  

CTV is now being asked to do more than deliver reach. For multi-location brands, franchise systems and regional advertisers, the real question is not whether the ad ran. It is whether the campaign helped drive store visits, calls, appointments, sales or revenue in the markets that matter.

CTV is a lean-back experience. Viewers watch from across the room, without a mouse or a touchscreen, and there's no hyperlink to tap. A second device might be nearby, but a direct click-through from the TV screen itself basically doesn't happen. So when platforms measure CTV the way they'd measure a search ad or a social post, they're not just missing some conversions. They're missing most of them.

The 20% you can see, and the 80% you can't

Here's the number that should reset how multi-location brands think about CTV measurement: relying only on clicks and immediate web visits can capture only a fraction of the actual commercial impact of a campaign. The larger opportunity is often offline: a customer walks into a store, calls to book an appointment, shows up for a service visit, or completes a purchase in person.

For a purely online business, web attribution may capture more of the picture. But for a multi-location retailer, franchise network, QSR chain, healthcare group, dealer group or regional service business, the offline gap may be where most of the business impact lives. About 80% of commerce still happens locally, in stores, offices, restaurants, and on the phone. If your attribution model only sees online conversions, you're not getting an incomplete picture of performance—you're getting a picture of a different business than the one you're actually running.

The practical risk is real and it's not hypothetical: brands routinely cut budget from CTV campaigns that are quietly driving strong foot traffic and phone revenue, simply because that revenue never showed up in a dashboard built for online conversions. You end up optimizing against the wrong signal, and pulling money away from the thing that's working.

How CTV online attribution works

Connecting a TV impression to something that happens in a store or on a phone requires a few technical steps working together. It's not magic, and it's not a black box—it's a pipeline.

1. Impression logging. When the ad runs, the ad server logs the exposure: when and where the ad ran, household identifier, timestamp, publisher, creative, device data. This is the foundation. Nothing downstream works without a clean log here.
2. Device graph matching. The platform links that TV-level exposure to other devices in the same household, phones, tablets, laptops, using a mix of deterministic signals (like logged-in sessions) and probabilistic signals (like shared network activity). This is what turns "an ad played on a television" into "a household we can actually track."
3. Offline outcome ingestion. This is where the real world enters the picture. Location panels collect privacy-safe mobile location signals to confirm store visits, point-of-sale systems upload transaction data to verify purchases, call tracking software logs inbound calls, and booking or scheduling data captures appointments. These are the outcomes that actually matter for a multi-location brand: someone showed up, someone called, someone booked, someone bought, and they're exactly the outcomes a click-based model can't see.
4. Correlation and attribution. The system checks whether a device from an exposed household shows up in that outcome data, a store visit, a call, a transaction, within a defined window after the ad ran. If it does, that outcome can be tied back to the exposed household within a defined attribution window.

As third-party identifiers keep degrading across mobile operating systems and browsers, this pipeline is leaning harder on first-party data: CRM records, verified household address data, and privacy-safe identity resolution rather than IP addresses that can disappear behind a VPN or a privacy relay overnight.

Retail visit attribution CTV: why store boundaries matter

If you're running a multi-location brand, store visit attribution is the metric that actually maps to your business. But how a platform defines "at the store" changes the answer more than most buyers realize.

The older approach uses a centroid radius: a single point at the center of the property, surrounded by a fixed circle, often somewhere between 100 and 500 feet. The problem is that real buildings aren't circles. A radius that size routinely bleeds into the parking lot next door, the drive-thru lane of an adjacent restaurant, or a neighboring storefront, and it counts people who never walked in as visits. Pull the radius in tighter to avoid that, and you start missing people who genuinely did visit but happened to park or enter from an angle the circle doesn't cover.

Retail visit attribution that holds up under scrutiny uses building footprint polygons instead: the actual outline of the structure, mapped to its real boundaries. Combined with parent-child location data, a polygon can tell the difference between the coffee shop and the shoe store next to it inside the same strip mall, which a circle simply can't do.

Getting from raw GPS location signals to a confirmed visit requires filtering noisy data, identifying likely dwell events, matching those events to the store footprint and resolving ambiguity in crowded retail environments. It is more work than drawing a circle on a map. It is also the difference between a foot traffic number you can defend and one you cannot.

CTV call attribution: measuring calls, appointments, and buying intent

For categories like healthcare, home services, legal, financial services, auto and high-consideration retail, the conversion moment is often a phone call. Someone sees the ad, looks the business up, and calls to get a quote or book an appointment.

Connecting that call back to the ad requires dynamic number insertion: when a visitor from an exposed household lands on the site, the phone number they see is swapped for a unique, trackable number tied to that session. When they call, the platform can trace the call back through the visit to the original ad exposure and route it to the right local store or agent automatically, no menu-punching required.

Raw call volume alone isn't a useful metric, though. Wrong numbers, billing questions, and customer service calls all show up in the same bucket as new business. More advanced setups can use conversational AI to help classify calls by intent: was an appointment booked, was a quote requested, did the conversation indicate a real sales opportunity? That structured signal can then feed straight back into bidding platforms, so media spend follows calls that actually convert, not just calls that happen.

CTV incrementality: proving what the campaign actually caused

Counting visits isn't the same as proving your ad caused them. Plenty of people would have walked into your store anyway: loyal customers, people who happened to be nearby, habitual visitors. That's baseline traffic, and if you don't subtract it out, you're taking credit for business you didn't create.

The fix is a holdout test: split your audience into an exposed group that sees the campaign and a matched control group that doesn't, then compare visit rates between the two. The difference is your incremental lift, the traffic your ad actually caused. Run the same comparison against revenue, and you get incremental ROAS: the return generated specifically by the campaign, not the return you'd have gotten anyway.

This only works if it's set up before the campaign launches, not bolted on afterward. A holdout group configured retroactively doesn't tell you anything real. It also only works if the attribution window matches how people actually buy. The right attribution window should match how people actually buy. A QSR visit may happen within 24 hours. A car purchase, healthcare appointment or home services quote may take weeks. Categories with longer consideration cycles need windows to match, typically 7 to 21 days for multi-location retail, and up to 30 days for automotive.

What this looks like in the real world

In practice, offline attribution can show very different signals by category. A QSR campaign may be evaluated on visit lift and cost per visit. A healthcare campaign may focus on appointment calls. An auto campaign may look at showroom traffic, lead quality or in-person conversions. A retail campaign may compare exposed and control markets to understand incremental store visits or sales.

The consistent theme: CTV’s physical-world impact is real, but it only becomes useful when the measurement matches the business outcome.

What CTV offline attribution can and can’t prove

Being direct about the limits here matters, because overclaiming is exactly the kind of thing that erodes trust in a measurement system.

Offline attribution can:

• Show statistically significant visit lift compared to a real control group.
• Verify building entry through polygon boundaries.
• Connect an ad exposure to a downstream call or transaction at the household level. 
• Compare performance by market, location, audience, creative or publisher when enough scale exists.

What it can't do is prove, for any single individual, that they walked into your store because of one specific ad they saw. It can't track someone's path once they're inside the building. And it can't attribute anonymous cash transactions that never touch a loyalty program or a CRM record. Good measurement partners will tell you this upfront. Ones that promise individual-level certainty on every conversion are telling you something that isn't technically true.

Building it right from the start

If you're running CTV for a multi-location brand, four things matter more than any single vendor feature:

1. Build your identity foundation on first-party data, CRM records and verified addresses, rather than volatile third-party identifiers that keep losing ground to privacy changes.
2. Use real building footprint polygons for every location, not radius circles, and pair that with call tracking that actually validates conversations instead of just counting them.
3. Set up holdout groups before launch, not after, and measure success by incremental lift, not raw visit counts.
4. Feed those validated offline outcomes back into your bidding engine, so spend actually follows what's driving revenue in the real world, not just what shows up as a click.

This is the measurement model CTV needs if it wants to compete for performance budgets. Advertisers should not have to choose between TV’s reach and digital’s accountability. Jamloop was built to connect what happens on screen to what happens in the real world, including store visits, calls, appointments and sales.

FAQ

What do I need to have in place before turning on CTV store visit attribution? 

At minimum: a clean list of your physical locations with accurate addresses, a first-party data source (CRM, loyalty program, or POS system) you can match against exposed households, and enough campaign scale to generate a statistically meaningful exposed and control group. Attribution set up after a campaign already launched can't retroactively build a valid holdout group, so this needs to be in place before day one, not added mid-flight.

‍

Does retail visit attribution CTV work for stores inside a mall or strip center, or only standalone locations? 

It works for both, but shared buildings need more precision. A store inside a multi-tenant complex needs its footprint mapped as its own polygon, with parent-child data distinguishing your location from the mall or center around it. Without that distinction, a visit to the shoe store next door can get misattributed to you, or vice versa.

How long does it take to see reliable offline attribution results after a campaign launches? 

You need enough time for the exposed and control groups to accumulate a meaningful sample of visits or calls, on top of the attribution window itself (which can run anywhere from a few days to a month depending on category). Most brands should plan on four to six weeks of live campaign data before treating early lift numbers as reliable rather than directional.

Is offline attribution worth it for a brand with only a handful of locations, or is it only useful at national scale? 

Scale helps statistical confidence, but it's not a hard requirement. The bigger factor is category: high-consideration, high-ticket categories (auto, home services) tend to show clearer offline signals even at a smaller scale than low-ticket, high-frequency categories.

Can CTV attribution measure offline sales?

Yes, when the right data is available. Offline sales can be measured through POS data, loyalty data, CRM matchbacks or other privacy-safe transaction signals. The quality of the measurement depends on the quality of the data, the match rate and whether the campaign was set up with a proper control group before launch.