Hubpix aimed to revolutionize in-store advertising within the convenience sector by moving away from traditional, inefficient field sales models. With an existing base of 25,000+ independent retailers, they needed a digital platform that could connect brands directly with store owners to drive better media activation and sales. Through extensive field research and ecosystem mapping, we developed a digital model featuring a Cloud Portal for brands and a Mobile App for retailers. The resulting platform eliminated print media waste, provided concrete evidence of media activation, and established a system where retailers earn direct cash rewards for their participation. However, verifying campaign execution meant manually reviewing thousands of in-store photos, a process that was accurate but fundamentally unscalable. We engineered an AI-driven visual intelligence system to automate this entire workflow. By deploying custom computer vision models and OpenAI vector embeddings, we transformed raw store photos into real-time, actionable shelf data. The system now processes complex retail imagery in under five seconds, drastically reducing out-of-stock blind spots. This transformation proved so successful that the platform was ultimately acquired by IB Group, validating the immense strategic value of the technology.
stores daily now undergo successfully automated remote shelf-compliance checks
accuracy in recognizing products and brands in real-world conditions
reduction of out-of-stock situations
Product visibility increased by an average
Seconds per image processing speed
Brands heavily relied on field sales agents for in-store promotional activities, but lacked control over daily schedules and quality checks. To verify product presence and promotional compliance, field sales agents submitted photos that required manual review. This approach crumbled under scale. Diverse store layouts, inconsistent displays, and the sheer volume of photos made accurate verification incredibly slow. High-value promotional data sat unutilized simply because extracting the insights took too long.
This resulted in a complete lack of media activation evidence, leaving brands with poor data visibility and an inability to accurately report campaign status to senior leadership. Additionally, brands experienced massive print media waste by mass-producing assets without knowing which retailers would actually participate. Conversely, the existing model offered independent retailers no incentives to activate media, as entire advertising budgets were absorbed by wholesalers and field sales operations.
Traditional retail checks, where field staff physically walk aisles to note gaps, are prone to human error and simply do not scale for major retail chains. Hubpix needed a system that could automatically “see” and understand retail shelves just like a human eye would, but infinitely faster and without fatigue.
We started out lean. Since precision was required, instead of throwing bloated datasets at the problem requiring millions of images to train a heavy neural network, we trained our initial object detection model using just 250 labeled retail shelf images. By utilizing smart data labeling and iterative training, the model quickly surpassed a 90% detection baseline. We broke the visual challenge into two distinct steps: teaching the system to physically map empty shelf gaps, and teaching it to instantly recognize specific product “fingerprints” using advanced vector databases.
To solve this, we engineered an end-to-end, AI-powered visual auditing platform that transforms standard shelf photos into structured retail intelligence. By combining custom computer vision, vector matching, and a generative AI co-pilot, the system automates gap detection, maps product placements, and delivers highly validated insights at enterprise scale.
We built a custom computer vision model trained specifically to detect two elements: active products and empty shelf spaces. It flags missing stock instantly without requiring the store to upload a predefined planogram.
Recognizing a product shape is useless without knowing the exact SKU. We used OpenAI models to generate embeddings, the digital “fingerprints”, for each product in the catalog. When a new store photo arrives, the system crops the detected products and matches them against our vector database for rapid, highly precise identification.
Knowing what is on the shelf is only half the battle; brands need to know exactly where it sits. We deployed a secondary model that detects entire shelf rows, numbering them from bottom to top. It maps exactly where products sit (e.g., “Brand A chips are on Shelf 3”) and instantly highlights the specific rows with available restocking space.
To make this massive influx of visual data accessible, we integrated a conversational Generative AI layer. This acts as a data co-pilot, translating raw visual tracking metrics into intuitive, actionable insights for non-technical stakeholders and brand managers.
Achieving near-100% product identification accuracy required more than relying on a single AI model or visual similarity score. We engineered a sophisticated Multi-Stage Pipeline (MSP) that validates every detected product through multiple independent layers of evidence. This combination of AI-driven automation, multi-signal validation, and intelligent exception handling enables highly accurate, scalable, and enterprise-grade product identification across large volumes of retail shelf images.
Replacing manual photo verification with a high-speed visual intelligence pipeline delivered immediate, highly quantifiable operational results.
The entire computer vision sequence, detecting products, matching embeddings, and mapping shelves, takes less than 5 seconds per image.
The system successfully matches cropped store imagery against the vector database, recognizing products and brands in real-world store conditions with up to 95% accuracy.
By detecting empty spaces up to 70% faster than manual checks, the platform reduces out-of-stock situations by over 90%, directly rescuing lost revenue.
Automated shelf row mapping empowers brands to optimize their physical placement strategies, increasing product visibility by an average of 20%.
Hubpix now conducts remote shelf-compliance checks across over 50 stores daily, entirely eliminating the need for costly physical field visits and freeing up staff to focus on customer-centric tasks.
What began as a tactical campaign execution challenge evolved into a platform that fundamentally redefined how brands measure retail activation on the ground. Hubpix bridged the gap between campaign planning and real-world execution, enabling sales representatives to capture visual proof directly from stores and providing brands with immediate visibility into compliance and merchandising standards.
The impact extended far beyond basic campaign tracking. By transforming manual store audits into a scalable, evidence-driven process, Hubpix unlocked a new level of accountability and operational intelligence for consumer brands. This approach to field execution proved so compelling that the platform was ultimately acquired by IB Group. This acquisition validated both the market need and the deep technical innovation behind our solution, proving how a focused digital product can evolve from solving a daily operational bottleneck into creating massive strategic value for an entire industry.
We don’t just build software; we bridge the gap between physical operational realities and digital intelligence. Our team combines over 25 years of digital product engineering with practical, cutting-edge AI integration to solve real-world bottlenecks. We understand that in retail, speed and accuracy are your ultimate competitive differentiators. If you are looking to replace error-prone manual processes with scalable, intelligent automation that drives measurable enterprise value, let’s connect.
Results matter. Discover how our customized solutions have delivered measurable business outcomes for clients across various industries. Each case study highlights the strategic impact of our work, turning challenges into opportunities for growth.
Ask Sol anything about digital products, AI, engineering, or growth, and get answers drawn from years of Net Solutions thinking and experience.