Physical retail is experiencing something that looks, from the outside, like a paradox. Consumer demand for the in-store experience remains strong: Ryder’s 2024 survey of 1,306 US online shoppers found that 61% shopped in-store because they enjoyed the experience, up from 40% a year earlier. In Europe, 92% of respondents to ShopFully and Offerista’s 2024 survey of 11,000 marketplace users favoured shopping in physical stores. At the same time, many retailers still rely on a narrow set of measures to understand what happens in those stores. They count entrances. They count transactions. The journey between them remains invisible in those totals.

Ecommerce has built its operations around a richer view. Digital retail has spent twenty years building measurement infrastructure that is, by now, extraordinarily precise. It knows which pages a visitor browsed before buying. It knows which category they reached first, how long they spent there, and which path through the site preceded the transaction. It can identify patterns of purposeful or exploratory browsing, distinguish first-time from returning visitors, and analyse whether the channel that brought them in predicts their likelihood to convert. Much of the customer journey is observable, measurable, and actionable in close to real time.

Physical retail has had a less complete picture. Even sophisticated retailers can find themselves making zone layout decisions, merchandising choices, and staffing allocations based on aggregate footfall and end-of-period sales reports: two measures that describe outcomes without explaining them.

The intelligence arriving in physical retail

A convergence of computer vision, re-identification technology, and zone-level behavioral analytics is giving physical retail a data layer that increasingly approaches what ecommerce has operated with for a generation. According to Grand View Research, the global market for computer vision AI in retail was estimated at USD 1.7 billion in 2024, projected to reach USD 12.56 billion by 2033. The customer experience enhancement segment led this market in 2024, accounting for over 27% of global revenue.

The technology behind this shift is, at its core, a reimagining of infrastructure that retailers already own. Existing 2D security camera feeds can now be processed by AI models to reconstruct individual visitor journeys. Computer vision turns footage into observations; re-identification connects those observations across camera views. The result is a sequence of zones entered, the duration of each stay, and the time taken to move between them, with additional signals such as shopping-bag observations where supported.

This is a different category of measurement from people counting. A visit reconstruction of this kind brings physical retail closer to the behavioral detail an ecommerce clickstream provides for online journeys. Research has explored shopping paths and purchases, including in-store video tracking, for more than a decade. The opportunity now is to make that analysis part of everyday store operations.

Scaled across thousands of visits, this data makes a different set of questions practical to investigate.

• Which zone sequences most reliably precede checkout?

• Where does the journey of a confirmed buyer diverge from the journey of a non-buyer, when transaction data is linked?

• Which zones are associated with progression towards purchase, and which attract dwell time with limited onward progression?

• Which visitors return, and how does their second visit differ from their first?

Who is in the store

Journey reconstruction is one dimension of the intelligence arriving in physical retail. Visitor profiling is the other, and it is the dimension most directly connected to the strategic question of whether a store is reaching the audience it was designed to serve.

The visitor entering a store carries a behavioral signature derived from how they move, how long they stay, and which zones they reach. Estimated audience attributes add another dimension. Matching that profile to the store’s defined target customer transforms footfall from a volume metric into a measure of audience fit.

In Medallia’s research, 82% of consumers said personalized experiences influenced their brand choice in at least half of shopping situations. Digital channels have responded to that expectation by analyzing millions of behavioral data points in real time. When customers receive curated, relevant experiences online, that raises the standard against which physical experiences are judged. Yet physical retail often still relies on static planograms, intuition, and slow reporting cycles. The infrastructure to close that gap now exists.

The metric that makes visitor quality actionable is ICP share: the proportion of qualified visitors in a given zone who match the store’s defined ideal customer profile, computed hourly, by zone, segmented by weekday and weekend.

At a major European fashion retailer we analysed over six weeks, ICP share varied by zone, hour, and day type. Women’s zones during weekday afternoons attracted a meaningfully different visitor mix than the same zones on weekend mornings. Several zones that appeared commercially strong in aggregate footfall terms served a lower-ICP audience than their traffic volumes suggested. None of this was visible in the footfall totals. Our framework for measuring visitor quality develops the approach further.

The storefront as a measured system

The intelligence arriving in physical retail extends to the storefront itself, and to a question that most retailers currently answer only by intuition: how effectively does passing traffic become qualified visits from the audience the brand is designed for?

This question produces a single number, the Qualified Reach Rate, defined as the share of passersby who become a qualified, ICP-matched visitor inside the store. It connects four measurements: passing traffic, entry, qualification, and target-profile match. The entry step on its own is drop-in rate.

The diagnostic format this produces gives retail executives a clear view of where the funnel narrows. Across a six-week measurement period at a major European fashion retailer, the storefront funnel ran as follows:

Qualified Reach Rate: 2%. Source: Indivd’s six-week analysis at a European fashion retailer, discussed in Beyond footfall.

Each gap between those numbers identifies a stage to investigate and an intervention to test. A display window that maximizes drop-in rate must also be evaluated on the qualified visits it produces and what those visitors do next. A higher target-profile share, even at lower entry volume, can represent an improvement when the resulting commercial outcomes support it. The window becomes testable: change it, then see who entered and what happened next.

Where qualified visitors convert

Once traffic quality is established at the entry level, the commercial question shifts to zone-level performance: which zones do qualified visitors explore before reaching checkout, and at what rate?

The metric that answers this is cross-conversion to tills: for each zone, the share of qualified visitors in that zone who reach the tills in the same visit. It connects activity inside the store with a later stage of the journey. It adds diagnostic detail to store-level conversion by revealing which zone visits are associated with checkout progression.

A zone with high dwell time and low cross-conversion to tills records long visits but relatively few subsequent checkout-area visits. A zone with strong aggregate footfall may therefore deserve closer attention than its traffic ranking suggests. The cause could be product availability, waiting for assistance, or browsing that leads to a later purchase. Zone-level measurement identifies where to investigate, and it can be produced continuously from suitable camera infrastructure. Fitting-room analysis offers a concrete example of how those connections inform staffing and merchandising decisions.

Paired with ICP share, cross-conversion to tills produces a quadrant classification of store zones. Core zones show high ICP share and strong checkout progression. Potential zones have the intended audience but weaker progression, a signal that product placement, staffing, or zone design deserves investigation. Zones with lower ICP share and strong progression can reveal valuable audiences the profile overlooks; zones low on both measures warrant a review of their intended purpose. The classification updates hourly, making it a tool for daily floor decisions, read alongside sample size and normal variation.

What visitor quality is worth in currency

The final layer of new intelligence connects visitor profile to estimated commercial value: zone-level value expressed in currency, informed by the visitor mix in each zone.

The logic is straightforward. Where observed purchase behavior establishes differences in purchase probability and expected basket size, those differences provide a basis for assigning value to visitor segments. Assigning value multipliers to each ICP tier translates visitor volumes and ICP share into a weekly currency figure per zone that estimates value alongside raw traffic. Without that behavioral basis, the ranking puts a currency symbol on an assumption.

The result is a ranking that can diverge from the footfall ranking. Zones that generate large visitor numbers may rank below smaller zones with stronger purchase outcomes. That difference gives retailers another basis for deciding where to investigate, test, and invest.

Inventory management, loss prevention, and checkout automation are important applications of computer vision. Visitor quality and zone-level commercial intelligence extend the opportunity: understanding how the store experience relates to the value of the visits it already attracts.

The store that knows itself

The intelligence described in this article can run on existing 2D security camera feeds. Camera coverage, configuration, and validation determine what each location can measure reliably. The barriers that have historically separated physical retail from the analytical depth ecommerce takes for granted are, in this domain, organizational as well as technical: a question of whether retailers are asking the right questions of the infrastructure they already have.

A customer who browses a website and adds to cart generates a detailed behavioral record that can inform subsequent improvements to the experience. The same customer who walks into a store, spends 35 minutes moving through six zones, engages deeply with two of them, and leaves with a shopping bag generates, in a store relying on footfall and POS data, a single entry in a footfall counter and a transaction if they purchased.

The ambition of current computer vision infrastructure in retail is to close the gap between those records. A store that understands visitor journeys, the audience it attracts, which zones hold attention, and how those patterns connect with commercial outcomes has a stronger basis for every floor decision. For retailers putting this intelligence to work today, the time to act is this quarter.

Common questions