How to Measure and Optimize the Physical Retail Marketing Funnel With Spatial Intelligence and Computer Vision

Xiaoyi Hua

·October 21, 2026

·14 min read

How to Measure and Optimize the Physical Retail Marketing Funnel With Spatial Intelligence and Computer Vision — Image Source: pexels

Physical stores have always had a marketing funnel. But unlike online channels, its steps have never been tracked or measured. This blind spot means you can't tell if a campaign failed because of placement, creative, product-market fit, or pricing. You see the sale, but not the journey behind it.

Spatial intelligence and computer vision change this. They make every step of the physical retail marketing funnel—Exposure, Attention, Interaction, Basket—observable using the cameras you already have. In-store promotions take a meaningful share of budgets, yet teams still judge them by register totals alone.

This article gives you a practical framework, stage-specific metrics, and guidance for running controlled in-store experiments.

Key Takeaways

Physical stores have a four-step marketing funnel: Exposure, Attention, Interaction, and Basket.
You can measure each stage with the cameras that are already in your store.
Metrics for each stage help you find exactly where a campaign fails.
Do controlled tests to try out changes and see what works.
Use data about how people move around the store to make better marketing choices.

Why the Physical Funnel Stays Invisible

Digital Channels Instrument Every Stage

Online marketing teams track the whole journey. They map each stage and set up tracking for key events, from the first impression to the final click. Awareness campaigns run through content, social media, and search. Teams watch website traffic and organic reach. Consideration content—like comparison guides, case studies, and free trials—feeds engagement metrics and lead nurturing data. Conversion tactics like retargeting and personalized email tie back to conversion rates and ROI.

The tooling makes this possible. Analytics platforms, marketing automation, and CRM systems record user behavior at every step. Teams set KPIs, find drop-off points, and run A/B tests to fix friction. Each stage leaves a data trail, so a weak result points to a specific cause.

Stores Capture Transactions, Not Journeys

Physical stores record the end of the journey, not the journey itself. Point-of-sale systems capture product, price, quantity sold, time of sale, store location, and payment method on every sale, return, or exchange. They also log discounts, coupon redemption, and promotional lift versus baseline. Inventory data adds on-hand stock, stock movement, and replenishment signals.

This tells you what sold. It does not tell you what shoppers saw, noticed, or picked up and put back. A display can fail at placement, creative, fit, or price. The register shows the same result in every case: no sale. You cannot separate those failure modes from transaction data alone. The middle of the funnel stays dark, and your team ends up guessing about causes it could measure.

Defining the Physical Retail Marketing Funnel

Exposure, Attention, Interaction, Basket

A shared vocabulary is the first thing you need for measurement. You need clear definitions for each stage of the physical retail marketing funnel. Without them, your data will describe different things under the same name. The four stages below give you that shared language.

Stage	Definition
Exposure	A shopper has a clear, unobstructed view of a product or display. The item is within their field of view, and a chance to engage exists.
Attention	The moment that chance becomes focus. The shopper notices the product, signage, or display and becomes aware of it. This is different from just being nearby, since a shopper may pass a display without noticing it.
Interaction	The shopper takes a physical action: picking up the product, examining it, comparing it with other options, or putting it back on the shelf.
Basket	The outcome stage. The item is either added to the purchase or removed from consideration.

Each stage narrows the group of shoppers who reach it. Exposure covers everyone with a clear view. Attention covers only those who notice. Interaction covers those who touch, inspect, or compare. Basket covers those who buy. A drop between any two stages points to a different problem, and that separation is what makes diagnosis possible.

A Familiar Sequence, Rarely Applied

This sequence will look familiar to anyone who has run digital campaigns. Impressions, clicks, add-to-carts, and purchases follow the same narrowing logic. Digital teams apply that logic with care. They define each event, track it automatically, and compare results against baselines.

Physical environments rarely get the same treatment. Teams describe displays as "high traffic" or "good visibility" without defining either term. They judge a promotion by its sales lift and stop there. The stages exist on the floor just as they exist online, but no one labels them, counts them, or tests them. That gap is why the physical retail marketing funnel stays a concept rather than a working tool. Clear definitions close the gap, and the next step is to see where campaigns actually break down.

Where Campaigns Break Down

Point-of-Sale Data Mixes Up Different Kinds of Failure

Your register records only one result for each shopper: they bought something or they didn't. That single piece of data can't tell you why the result happened. A shopper who never saw the display and a shopper who saw it, picked up the product, and rejected the price both show up as a missed sale. The numbers look the same, but the causes need opposite fixes.

This mixing up gets worse over time. You change the creative, sales stay flat, and you decide the creative failed. In truth, the display may sit outside most shoppers' line of sight, so the new creative never got a fair test. Teams repeat this cycle across seasons and budgets. Each round burns money without teaching you anything, because the data can't separate the stages where the campaign broke.

Figuring Out Placement, Creative, Fit, or Pricing

Stage-level evidence turns one unclear result into four clear diagnoses. Low exposure points to placement: the display sits off the main path, faces the wrong way, or hides behind fixtures. Healthy exposure with weak attention points to creative or signage: shoppers have a clear view but don't notice. Strong attention with weak interaction points to product fit or messaging: shoppers look, then walk away without touching. Solid interaction with weak basket conversion points to pricing or assortment: shoppers engage, then reject the offer at the shelf.

Work through the stages in order and the failure mode shows itself. Suppose a seasonal endcap draws strong attention but few pickups. You now know the problem lives between noticing and touching, so you test the product mix or the shelf talker instead of cutting the price. Suppose shoppers pick up the item and put it back on the shelf. Price signage or a competing option likely drives the rejection. Each diagnosis narrows your next experiment, and each experiment gives a cleaner answer than a sales total ever could.

How Computer Vision Captures Each Stage

3D Reconstruction and Line of Sight

The system begins by making a model of the store. AiFi's platform rebuilds the space in 3D and keeps track of where things are and how they move. Every shopper and every fixture has a spot in that model. From there, the platform works out line of sight: whether a shopper has a clear, open view of a product or display. This is the Exposure stage, measured directly instead of guessed from foot traffic.

The capture pipeline follows a clear order:

A 360-degree fisheye camera plus two perspective cameras gather multi-view data, getting past blocking and field-of-view limits.
The system rebuilds 3D skeleton keypoints and uses them to study hand-object interaction and 3D gaze estimation through a Transformer-based encoder.
A Kalman filter and strong calibration help the system rebuild the store accurately.

Working from 3D keypoints instead of raw pixels makes the method hold up well to changes in lighting and background. It also helps with privacy, because the system uses body geometry, not images that identify people. Shoppers don't have to do anything, and you add no new hardware. The same system that runs autonomous retail measures marketing performance with the cameras you already have.

Directed Attention and Physical Engagement

Exposure tells you a shopper could see the display. Attention tells you they did. The platform estimates gaze direction in 3D and pairs it with hand-object signals to tell brief contact apart from steady focus. A shopper who brushes a package in passing looks different from one who holds it and reads the label. Putting hand-object and gaze-object engagement signals together gives deeper behavioral insights than either one alone.

Physical engagement adds the body to the picture. Posture and gesture show intent: leaning toward a shelf, reaching for a product, or turning away. These combined feature sets feed a neural network that sorts engagement, much like attention-tracking systems judge attentiveness from where people look and how they react. The result is a stage-by-stage record of what happened on the floor.

Interaction sits between attention and basket. The platform spots pickups, inspections, comparisons, and replacements as separate events. Each event comes with a timestamp and a location, so you can follow a shopper's path from first glance to final decision. Basket outcomes then link to that path, and the full funnel becomes visible for every display you choose to measure.

Metrics Across the Physical Retail Marketing Funnel

Exposure Rate and Attention Rate

Exposure rate measures how often shoppers have a clear line of sight to a display. You figure it out by dividing the shoppers who pass by and can see it clearly. A low exposure rate tells you the fixture sits off the main path or faces the wrong way.

Attention rate measures how often exposure leads to awareness. You divide the shoppers who noticed by the shoppers who had a clear view. A healthy exposure rate with a weak attention rate points to signage or creative that fails to catch the eye.

Interaction and Basket Conversion

Attention-to-interaction transition measures how compelling the product or messaging is. You divide shoppers who picked up, inspected, or compared the item by those who noticed it. A weak transition means shoppers look and walk away, so the product mix or shelf messaging needs work.

Interaction-to-basket conversion measures how often engagement leads to a purchase. You divide purchases by interactions. A weak conversion points to pricing or assortment, because shoppers engage and then turn down the offer at the shelf.

Compare each metric against a baseline before you judge a campaign. Pair each test store with the most similar control store, then check that both tracked closely in the weeks before the test. Divergent pre-period trends signal a baseline problem you must fix first. For the analysis, simple t-tests suit basic comparisons, difference-in-differences controls for pre-trial trends, hierarchical models account for store-level differences, and time-series methods apply when daily sales feeds are available. Report primary metrics with confidence intervals so your team can see the range around each lift estimate.

From Measurement to Experimentation

Testing Displays, Placement, Pricing, and Signage

Measurement tells you what went wrong. Experimentation gives you a way to fix it. Treat each display choice as something you can test, and change only one thing at a time. Move an endcap into the main aisle and keep the creative the same. Swap the shelf talker and leave the placement alone. Adjust the price sign without touching the product mix. This careful approach lets you tie the result to the change you made.

Run each test in a matched pair of stores. Track both locations for a few weeks before the change, then compare stage metrics during the test window. A display that gains exposure but loses attention tells you the new spot pulls eyes past the fixture instead of toward it. A price sign that lifts interaction-to-basket conversion confirms the earlier rejection came from sticker shock, not from the product itself.

Isolating Effects Stage by Stage

Each stage metric works as an outcome variable for a different kind of test. Use exposure rate to judge placement changes. Use attention rate to judge creative and signage. Use attention-to-interaction transition to judge product mix and messaging. Use interaction-to-basket conversion to judge pricing and assortment. Pick the metric that matches your hypothesis, and read the other stages as controls.

Suppose you test a new shelf talker. Exposure should stay flat, because you did not move the fixture. Attention should rise if the creative works. If attention rises but interaction does not, the message caught the eye without persuading the shopper to reach. That pattern points to a product-fit problem rather than a design problem. Stage-by-stage isolation keeps your physical retail marketing funnel easy to read, and each experiment sharpens the next question you ask.

Where the Framework Applies

Endcaps, Seasonal Displays, and Launches

Endcaps, seasonal displays, and new product launches all work with the same four-stage view. You ask the same set of questions for each one: Did shoppers see it? Did they notice it? Did they engage with it? Did they buy? The answers show you where to take action.

For endcaps and seasonal displays, use stage metrics along with sales lift. Compare featured-product unit sales during the display period to baseline sales from a normal shelf spot. A strong, lasting lift is the clearest sign the display earns its space. Add dwell time to the picture. A high-traffic endcap with short dwell time signals a problem, because shoppers pass by but do not stop. For time-limited offers, track promotional velocity during the window against the weeks before and after. A spike with no drop after the promotion points to new demand rather than shifted sales timing.

New products gain the most from this approach. Endcap placement introduces an unfamiliar brand to shoppers and lets you share key benefits, pricing, and offers. Watch performance after launch to decide whether the extra space pays off.

New products benefit from endcap placement because the added visibility introduces them to shoppers who may not know the brand or product, while also letting retailers share key benefits, pricing, and promotional offers.

Price Signage and In-Store Media

Price signage and in-store media reward the same stage-by-stage reading. Spatial intelligence tracks awareness, engagement, and conversion across placements, and it filters out pass-through traffic so you count verified impressions rather than modeled ones. That difference matters when you judge whether a promotion reached new shoppers or only high-intent ones.

Attention data often reveals a messaging problem that sales data hides. In one coffee-brand display case, shoppers focused on the top of the display and mostly ignored the side panels, the panels most visible as shoppers walked past. The display was fine; the messaging sat where shoppers were not looking. Computer vision also reads price tags and turns shelf photos into structured data, catching pricing errors and missing tags. The system returns a compliance score and a prioritized fix list within 90 seconds. Each use feeds the same physical retail marketing funnel, so every activation answers the same four questions.

A New Starting Point for In-Store Marketing

From "Did It Work?" to "Where Did It Work?"

One sales lift number answers only a small question. It shows the campaign sold more product, but not which shoppers it reached or which part did the work. Detailed measurement lets you ask a better question. The Birds Eye frozen aisle project shows how this works in real life:

The brand tested the project before rolling it out, and the goal was to measure results in as much detail as possible—not just overall sales growth.
Measurement was split into single in-store elements, so the brand could see which elements worked well and which did not, which helped them roll out more efficiently and effectively.
Detailed measurement showed where the impact happened: the biggest lift came from shoppers with children, drawn in by bright colors, not the health-conscious shoppers they first aimed for.

That third finding is the real reward. The campaign worked, but it worked in a place the team did not expect. You can only act on that insight when your data separates the stages and the audiences.

Grounding Decisions in Floor Behavior

Stage-level data also shows which conditions help a promotion succeed. Two Eurocash case studies make this clear:

Case Study	Measured Uplift vs. Control	Where It Worked
Ajinomoto Oyakata	27 percentage-point units uplift vs. control	Stores running the same offer with media vs. without media; gains spread beyond promoted SKUs
Plant-based drinks brand (2+1 promo)	43% hero-SKU volume growth vs. 23% in control stores	Promoted hero SKU plus non-promoted SKUs also beat control

Both results point beyond the promoted item. Media presence and halo effects on other SKUs shaped the outcome, and a sales total alone would hide that. Base your decisions on floor behavior you actually observe, and the physical retail marketing funnel becomes a working diagnostic tool instead of a scorecard.

Your four-stage funnel now works as a system. Spatial intelligence turns a shopper journey that was never measured into stages you can diagnose and test. Stage-specific metrics take performance measurement beyond sales alone, and they let you compare each result against a baseline.

The shift matters. You stop guessing about placement, creative, fit, or pricing and start optimizing from floor behavior you actually observe instead of register outcomes alone. Begin by setting baselines for your key displays. Then run a first controlled experiment: match test and control stores, run across full weekly cycles, and check that stores apply the change the same way. Build that feedback loop across activations, and the physical retail marketing funnel becomes a tool you can repeat.

FAQ

Do I need to install new cameras or sensors?

No. The platform works with the camera setup your store already has. The same system that runs autonomous retail also measures how well marketing performs. Shoppers don't act any differently, and your team doesn't add any hardware to the ceiling or the shelf.

How does the system know a shopper actually looked at a display?

It rebuilds the store in 3D and tracks where each shopper stands inside that model. Then it estimates which way they are looking and pairs that with hand-object signals. A quick glance and a long look count as different events, so attention numbers show real focus.

Can I measure a single display without tracking the whole store?

Yes. Choose the fixtures you care about and set baselines for each one. The platform reports exposure, attention, interaction, and basket results for those displays. You can add more coverage later as your testing program grows.

What is the smallest useful experiment I can run?

Start with one display and one change. Pair a test store with a similar control store, track both for a few weeks beforehand, then make the change in the test store only. Compare stage metrics during the test window and read the result with confidence intervals.

How is this different from foot-traffic counters?

Traffic counters tell you how many people walked past a spot. They can't tell you whether anyone saw the display, noticed it, or picked up the product. Stage-level measurement separates those events, so a weak result points to placement, creative, fit, or pricing instead of one traffic number.