Shelf Analytics Across 20,000 U.S. Stores

A computer vision shelf analytics program turned more than one million monthly store images into realograms, planogram compliance data, and measurable evidence for category management across the United States.

Overview

From fragmented store photos to a national view of shelf execution.

A global consumer goods company needed a dependable way to measure shelf position, share of shelf, assortment, and planogram compliance across long aisles and multiple retail channels in the United States. Clobotics combined panoramic image reconstruction, SKU-level recognition, and realogram-to-planogram comparison in one national shelf analytics workflow.

Market
All 50 U.S. states
Coverage
Nearly 20,000 stores
Portfolio
1,500+ SKUs across 11 categories
Data volume
1M+ images per month
600K+Annual digital shelf audits
10×+Increase in audit coverage
65% → 87%+Planogram compliance
1M+Monthly image processing

Customer and business context

The customer is a global, multi-category consumer goods company with a North American portfolio spanning 11 core categories, including fabric care, home care, baby care, grooming, oral care, and hair care. Each category has its own must-stock assortment, price architecture, display logic, and competitive set. Headquarters needed to know more than whether a product appeared in a photo: teams needed to measure brand-block integrity, contracted shelf presence, and whether priority launches reached the intended shelf position.

The program covered nearly 20,000 warehouse, pharmacy, and discount retail locations across all 50 U.S. states. Store evidence came from several third-party retail brokers, dedicated audit teams, and later the company’s own field organization. More than one million images per month had to become consistent data for category management, supply-chain coordination, and annual retailer negotiations.

This could not remain a one-time image-recognition exercise. The customer needed an operating system that could work across channels, organizations, and multiple years: capture had to stay simple for field users, the vision system had to reconstruct 15-to-20-meter shelves, outputs had to fit headquarters workflows, and product models and execution rules had to keep pace with assortment and planogram changes.

Challenge

Long shelves could not be captured in one image

Large U.S. retail stores often use continuous shelves extending 15 to 20 meters. Field teams had to capture each aisle in 10 to 15 overlapping photographs. Differences in camera angle and walking distance created stitching errors, duplicate products, and missed facings, making a reliable store-level shelf record difficult to produce.

Multiple field organizations produced inconsistent evidence

The program involved third-party retail brokers, dedicated audit teams, and later an internal field organization. Each group had different capture habits and operating standards. Manual review could not economically validate more than one million images per month, and inconsistent evidence created disputes over execution results.

Headquarters could not compare the real shelf with the approved design

Planograms defined the intended shelf, but execution in stores could deviate by more than 30%. Without a national realogram, category teams lacked objective evidence to identify missing must-stock products, competitive intrusion, incorrect shelf placement, and unfulfilled display agreements.

Solution

Reconstruct each aisle as one continuous digital shelf

Implementation timeline

2019–2020

Algorithm development and pilot

Panoramic stitching and wide-shelf correction were validated across 500 pilot stores.

2021

National channel rollout

The workflow expanded into major warehouse, pharmacy, and discount channels across the United States.

2022–2023

Internal field-team adoption

The platform supported the customer’s move from external-only collection toward a dedicated internal execution team with faster feedback.

2024–2026

Category-management integration

Realograms became a recurring input for planogram evaluation, supply-chain coordination, and retailer discussions.

Clobotics turned 10 to 15 overlapping mobile photographs into one corrected panoramic shelf, preserving the context category teams need to evaluate the full aisle.

Reconstruct the shelf and rebuild the realogram

Shelf uprights, shelf edges, and product boundaries acted as spatial anchors while perspective and color correction aligned each frame and overlap removal prevented duplicate counts. Computer vision then mapped more than 1,500 customer and competitor SKUs into a structured realogram covering assortment, facings, shelf position, brand blocks, and share of shelf.

Compare the realogram with the planogram

The comparison engine evaluated the realogram against the approved design and highlighted missing must-stock products, incorrect shelf placement, competitive intrusion, and other execution gaps. Category and field teams received one consistent evidence base instead of separate photo collections.

Keep a million-image workflow operational

Automated recognition, quality controls, and exception review kept human attention on severe obstruction, geometric ambiguity, and disputed results. A maintained product library, model and rule updates, batch processing, shared KPI definitions, and role-based permissions allowed brokers, internal field teams, category groups, and headquarters to use the same shelf record without separate reporting stacks.

Results

More than 600,000 shelf audits per year

Annual audit coverage increased from fewer than 50,000 manually sampled shelf visits to more than 600,000 digital audits. The program processed more than one million images per month across nearly 20,000 stores, giving headquarters a consistent view across all 50 states.

Planogram compliance increased by 22 percentage points

Execution against priority planogram rules improved from 65% at the beginning of the program to more than 87%. Teams could identify the exact store, aisle, shelf, and product behind each exception and direct field attention accordingly.

Objective shelf data strengthened retailer negotiations

National share-of-shelf evidence helped the company compare contracted display commitments with actual store execution. The resulting documentation supported the recovery of several million dollars in display-fee adjustments where agreed standards had not been met.

Operating result

A realogram became a repeatable business asset, not a one-time audit.

The program connected field capture, shelf recognition, planogram compliance, and category decisions in one national workflow. Talk to Clobotics about scaling shelf analytics across complex retail channels.