Footfall attribution is the practice of matching advertising exposure to a subsequent physical visit. A device sees an advert, the device is later observed inside the boundary of a shop, restaurant or dealership, and the two events are joined into a reported metric: an attributed visit. It exists because most retail spending still happens in buildings, and the advertising industry needed a way to defend budgets against outcomes that leave no click behind. What it produces is almost never a count, but an estimate projected from a subset of observed devices onto the wider exposed population.

How a visit gets counted

Three measurements must line up: which devices saw the advert, where a device was and when, and the physical boundary representing the business.

The Media Rating Council fixed the vocabulary in its Location-Based Advertising Measurement Guidelines, issued on 29 March 2017 with the Interactive Advertising Bureau and the Mobile Marketing Association after a public comment draft the previous November. Those guidelines define a geo-fence in two forms. Point-radius draws a circle of a stated radius around a set of coordinates. Boundary measurement draws a polygon around the actual footprint of the premises. The document notes bluntly that geo-fencing generally cannot represent multi-storey structures, which is why a mall, an office tower or a supermarket above a car park remains among the hardest cases in the discipline.

Signal quality varies by an order of magnitude depending on source. The guidelines put GPS at roughly 10 to 100 metres, degrading significantly indoors. Wi-Fi is placed at 30 to 100 metres and better suited to indoor micro-location. Cell antenna trilateration resolves no finer than postcode or neighbourhood. Mobile IP addresses are neither accurate nor precise.

Presence alone is not a visit. Measurers are directed to apply speed and duration edit rules filtering out drive-bys and other momentary proximity, and to collect altitude data where feasible to separate floors. Second-level dwell time reporting is encouraged on the basis of at least 200-millisecond polling.

The bid request and the pixel

On the supply side, location arrives in the auction itself. OpenRTB carries a Geo object whose meaning depends on placement: subordinate to Device it indicates current location, subordinate to User it indicates home base. It exposes latitude and longitude, an accuracy value in metres, and a lastfix field counting seconds since the fix. A type field, enumerated in the AdCOM specification, distinguishes GPS or location services, IP address, and user-provided registration data. The specification instructs that centroids of regions such as postcodes not be sent.

That instruction is honoured unevenly, which is why most vendors decline to build on bid-stream location at all. Adsquare states it uses only signals derived from software development kits, a distinction repeated across store visit tracking expansions.

On the buy side, exposure is captured by a measurement pixel populated with macros, the substitution variables a demand-side platform fills at serve time. Azira's Adobe integration writes 22 macros into every pixel, tracing attributed visits back to creative, placement and audience segment rather than only to a campaign identifier.

Google's version of the metric

Google introduced store visits on 18 December 2014, folded into what was then called Estimated Total Conversions in AdWords, and its documentation describes a modelled figure rather than an observation. A signed-in user with location history enabled interacts with an advert, later visits the premises, and Google extrapolates from observed cases to the broader exposed population. A survey layer sits underneath: Google states it questions users drawn from a panel of more than 10 million Google Opinion Rewards volunteers to compare reported visits against predicted ones.

Eligibility is restrictive. Locations must be verified in a Business Profile, campaigns must run location assets, and accounts must clear undisclosed privacy thresholds varying by advertiser. Categories tied to religion, sexual interests, children and certain health conditions are excluded as sensitive. Measurement is offered in 36 countries as of September 2026. Windows differ by interaction: 30 days by default for click-attributed visits, adjustable between 1 and 30, against 3 days for engaged views following at least 10 seconds of video and 1 day for view-through. Google recommends 7 days where store visits feed Smart Bidding, and warns of a reporting delay leaving the most recent five days at zero. Google Ads API version 18, released on 16 October 2024, added a last-click store visits metric at campaign level.

Origin and consolidation

The independent market formed around panels. Placed, founded in 2011, became an early supplier of visitation lift studies. Snap acquired it in 2017; Snap's annual report records total consideration of $185.9 million, of which $139.6 million was purchase consideration including $135.2 million in cash, while contemporaneous trade coverage reported between $125 million and $135 million and dated the deal to June rather than July. Snap never integrated it. Foursquare bought Placed back in 2019 alongside a $150 million investment led by The Raine Group.

Accreditation arrived slowly and partially. In August 2020 the MRC granted Foursquare Visits the first accreditation issued for location data. Foursquare's own announcement noted that Visits underpinned several products including Attribution, which had not been audited or accredited. GroundTruth followed on 3 August 2021.

Measurement then migrated inside buying platforms. StackAdapt and Adsquare launched an integrated footfall attribution product on 10 September 2024 across EMEA, APAC, North America and Latin America. Adsquare turned post-campaign reporting into a live dashboard on 30 June 2025 using a control condition methodology. The Trade Desk brought the same data into Audience Unlimited on 17 June 2026, and Amazon DSP gained Adsquare location segments across six marketsin July 2026.

Why it matters

For anyone buying media against physical outcomes, this is the only widely available signal between a click and a till receipt. Cost per visit replaces cost per click, and a retailer can argue for budget on traffic rather than proxies. Reported results are correspondingly attractive. Adsquare has documented a Walmart Mexico campaign showing a 17% year-on-year uplift across 12 million impressions, and a Visit Vermont campaign achieving cost per visit below $0.50. Azira measured a combined 31% foot traffic rise for restaurant brands advertising around Super Bowl LIX, with Little Caesars at 35%, Dunkin' at 31% and Taco Bell at 30%. Every one of those figures is vendor-supplied and none independently audited.

Limitations and disputes

The oldest complaint concerns the raw material. Analysis published by Placed founder David Shim in 2015, covering roughly four billion daily impressions across some 50 million devices, found that only 27% of bid requests carried location at all, and that 12% of those were precise enough to identify a specific business. A later Placed study reported average accuracy of about four New York City blocks. Both came from a vendor selling panel measurement against bid-stream alternatives, a commercial interest worth naming.

The deeper objection is causal. A matched visit establishes sequence, not consequence. Somebody who searched for a shop, saw an advert and then walked in may well have walked in anyway. Incrementality methods address that with holdouts and geo-experiments, and vendors increasingly bolt control conditions onto footfall products, but a headline attributed-visits figure carries no counterfactual. The endogeneity problem bites directly: targeting concentrates exposure on people already near the shop.

Coverage is a third constraint. Panels skew towards whichever app categories carry a vendor's software development kit, and opt-in rates fall as operating systems tighten permissions. Google's own consumer product moved that way: Maps Timeline, announced in December 2023 and rolled out through 2024, shifted history storage on-device and cut default retention from 18 months to three. Google's advertising documentation still describes an account-level location history opt-in, and has not publicly reconciled the two.

Regulation is now the sharpest pressure. The Federal Trade Commission closed its four-year case against Kochava with a stipulated order filed on 4 May 2026 barring sales of precise location data without affirmative express consent. New Jersey went further. Assembly Bill A5328, signed on 30 June 2026, bans sales of sensitive data including geolocation precise to within 1,750 feet, carries no consent exception, and sets penalties reaching $50,000 per record. Oregon and Maryland acted earlier.

Adjacent terms

Store visits conversions is Google's proprietary implementation, not a synonym for the category. Its modelling, eligibility and windows are specific to Google Ads.

Geo-fencing is the boundary technique shared by targeting and measurement. Proximity targeting decides who sees an advert; footfall attribution decides who is counted afterwards.

Store sales measurement matches exposure to transaction records rather than presence. It establishes whether money changed hands, which footfall attribution never does.

Foot traffic counting totals visitors using door sensors, cameras or Wi-Fi probes. It carries no exposure component and attributes nothing.

Recent developments

Attribution is being pushed towards purchases and across borders. StackAdapt linked its platform to Affinity Solutions card data on 30 July 2026, covering over 150 million cards and $4 trillion in spend. Azira and Adobe announced cross-border footfall attribution inside Adobe Advertising DSP on 25 August 2026, live first in Asia-Pacific, disclosing neither attribution windows, minimum dwell time, control methodology nor match rates.

Out-of-home has become a growth vector. Azira and Carvertise connected rideshare vehicle advertising to verified destination visits on 2 June 2026, reporting more than 2,600 visits for Visit Fort Worth alongside a documented discrepancy between the $800,000 economic impact figure in the release and a $1 million figure circulated elsewhere. Azerion's acquisition of mJourney on 27 August 2026 sold footfall and store visit rate as packaged deliverables without naming the supplier. Nielsen went the other way, counting 243,400 shoppers at the new IKEA in New Zealand through panel recall: a description of who visited, attributed to nobody.

Timeline

  • 2011: Placed is founded, building an opt-in panel for visitation measurement
  • 18 December 2014: Google adds store visits to Estimated Total Conversions in AdWords
  • October 2015: The Mobile Marketing Association publishes its Demystifying Location Data Accuracy whitepaper
  • February 2016: IAB publishes its Mobile Location Data Guide for Publishers
  • November 2016: MRC issues draft Location-Based Advertising Measurement Guidelines for public comment
  • 29 March 2017: The final guidelines are published with IAB and MMA
  • 2017: Snap acquires Placed, recording total consideration of $185.9 million
  • 2019: Foursquare acquires Placed from Snap alongside a $150 million investment
  • August 2020: MRC grants Foursquare Visits the first location data accreditation
  • 3 August 2021: GroundTruth receives MRC accreditation for location, place and visitation data
  • December 2023: Google announces the move of Maps Timeline on-device, cutting default retention from 18 months to three
  • 10 September 2024: StackAdapt and Adsquare launch integrated footfall attribution
  • 16 October 2024: Google Ads API v18 adds a last-click store visits metric
  • 30 January 2025: Adsquare extends store visit tracking to Austria and Switzerland
  • 30 June 2025: Adsquare converts attribution reporting into a live dashboard
  • 4 May 2026: The FTC files a stipulated order closing its Kochava location data case
  • 2 June 2026: Azira and Carvertise link rideshare advertising to verified visits
  • 17 June 2026: The Trade Desk brings Adsquare data into Audience Unlimited
  • 30 June 2026: New Jersey enacts A5328, banning sales of precise geolocation data
  • 30 July 2026: StackAdapt links to Affinity Solutions card transaction data
  • 25 August 2026: Azira and Adobe launch cross-border footfall attribution in APAC

Summary

Who. Location intelligence vendors including Adsquare, Azira, Foursquare and GroundTruth supply the measurement. Demand-side platforms including Adobe Advertising, Amazon DSP, StackAdapt and The Trade Desk carry it inside the buying interface. Google operates its own implementation as store visits conversions. Standards sit with the Media Rating Council, IAB and the Mobile Marketing Association.

What. A method for matching advertising exposure to a subsequent physical visit, producing an attributed visit count and derived metrics such as cost per visit and store visit rate. The figure is projected from an observed subset rather than counted, and the projection method is rarely disclosed in full.

When. Panel-based visitation measurement dates to the early 2010s. Google added store visits in December 2014, formal measurement guidelines followed in March 2017, the first accreditation was granted in August 2020, and the capability moved inside demand-side platforms from September 2024 onwards.

Where. In the mobile location layer beneath programmatic buying: software development kits inside applications, geo-fenced polygons around premises, OpenRTB Geo objects in the bid stream, and measurement pixels populated by macros at serve time.

Why. Most retail spending still occurs offline, and no click connects an advert to a doorway. Footfall attribution supplies the missing signal, at the cost of resting on precise location data that regulators in several jurisdictions now treat as sensitive by default and, in New Jersey since June 2026, refuse to let anyone sell.