A leading Indian automotive manufacturer, with a network of 3,500+ franchised retail outlets, faced mounting challenges in enforcing store brand compliance. Like many automotive retailers, the client relied on manual audits by field teams from the compliance department to ensure adherence to SOPs for visual merchandising, logo placements, promotional displays, and asset usage. However, this process was fraught with inefficiencies:
Human Bias & Pilferage: Field teams could overlook or intentionally ignore violations, leading to inconsistent image analysis.
Lack of real-time insights: Weak processes leading to inefficiencies and increased reliance on extensive auditing processes.
High Operational Costs: Manual data entry consumed significant resources, and real-time insights minimise the need for extensive auditing
The client aimed to fine-tune deep learning algorithms, specifically Convolutional Neural Networks (CNNs), for image classification and object recognition. This initiative was intended to enhance compliance enforcement, optimize operational costs, and ensure the protection of brand integrity.
Business Challenge
The automotive retail sector thrives on consistent brand experiences to drive customer loyalty. However, decentralized franchisee operations often lead to deviations from brand guidelines. Key pain points included:
Inability to monitor compliance accurately.
Lack of accountability due to human-dependent audits.
Rising costs from manual processes and delayed issue resolution.
Solution
Akaike Technologies integrated a computer vision model-based feature into their task management platform, automating brand compliance inspections. This deployment enabled the company to transition towards a precise and streamlined brand compliance monitoring process, enhancing operational efficiency and accuracy.
Comparison: Manual vs. AI-Enabled Brand Compliance Monitoring
Key Steps:
Step 1: Automating Compliance Data Capture
The field teams click images using their task management platform which inturn get uploaded to the clients centralised server, and our team receives an json payload to process the image.
Digitally mapped brand assets (e.g., logos, posters, product displays, employee uniforms, store front facia etc.) to create a compliance reference database.
Step 2: Training Computer Vision Model for Image Classification and Object Recognition
Trained a custom computer vision model using sample annotated images to flag non-compliance (e.g., missing signage, incorrect product placement, shabby staff uniform).
Augmented the dataset to create diverse data representations and tackle class imbalances in a training dataset.
Applied multiple models for different scenarios. For example, for front facia image analysis, applied segmentation and classification model vs. test drive area inspection object detection model with unique scoring mechanism
Step 3: Deploying Actionable Compliance Insights
Fine-tuned Computer Vision based product feature ensured precision and compliance, eliminating the need for manual supervision
Centralized Dashboard: Compliance scores, trends, and vendor rankings were visualized for regional teams.
Model Refinement: Non-compliant cases were fed back into the AI to improve accuracy.
Integrated anomaly detection to identify pilferage risks (e.g., missing test drive vehicle i.e. 1 out of 3 test drive vehicles is missing in the test drive zone).
Impact Delivered
Enhanced Compliance Accuracy: Achieved an impressive 90% overall model accuracy.
Consistent Brand Representation: Fine-tuned models enable accurate identification and classification of brand-related images, fostering a cohesive brand image.
Protection of Brand Integrity: By detecting unauthorized or inappropriate use of brand assets, companies can safeguard their brand integrity and prevent potential reputational damage.
Elimination of Pilferage Incidents: AI-powered cross-verification has successfully flagged discrepancies in field teams reports, effectively deterring fraudulent activities.
The Akaike Edge
Akaike's expertise lies in end-to-end management of AI lifecycle processes including problem identification and data collection to model deployment and ongoing monitoring, utilizing industry-leading frameworks, tools, and libraries
.png)
%20(1).png)
