APAC — India Omnichannel Retail
Challenge
The retail brand was facing critical growth bottlenecks driven by low search conversion rates and frequent stockouts, especially across Tier-2 and Tier-3 cities. Customers struggled to find relevant products due to weak search relevance and lack of personalization across multiple languages. This was further compounded by fragmented user experiences across channels, including mobile apps, web platforms, and WhatsApp, where personalization was either inconsistent or entirely absent.
Additionally, merchandising teams lacked accurate demand visibility at a micro-regional level, resulting in inefficient inventory allocation, overstocking in some regions, and stockouts in high-demand areas. Seasonal spikes, particularly during festive periods like Diwali, intensified these issues due to unpredictable demand patterns.
The absence of localized intelligence and real-time data-driven decision-making led to missed revenue opportunities, higher customer acquisition costs, and reduced customer satisfaction. The business needed a scalable AI-driven system to unify search, personalization, and demand forecasting while supporting multilingual users and diverse regional buying behaviors.
The retail brand was facing critical growth bottlenecks driven by low search conversion rates and frequent stockouts, especially across Tier-2 and Tier-3 cities. Customers struggled to find relevant products due to weak search relevance and lack of personalization across multiple languages. This was further compounded by fragmented user experiences across channels, including mobile apps, web platforms, and WhatsApp, where personalization was either inconsistent or entirely absent.
Additionally, merchandising teams lacked accurate demand visibility at a micro-regional level, resulting in inefficient inventory allocation, overstocking in some regions, and stockouts in high-demand areas. Seasonal spikes, particularly during festive periods like Diwali, intensified these issues due to unpredictable demand patterns.
The absence of localized intelligence and real-time data-driven decision-making led to missed revenue opportunities, higher customer acquisition costs, and reduced customer satisfaction. The business needed a scalable AI-driven system to unify search, personalization, and demand forecasting while supporting multilingual users and diverse regional buying behaviors.
Solution
Viston designed and deployed an integrated AI-powered merchandising intelligence layer that combined multilingual search, personalization, and demand forecasting into a unified ecosystem.
A multilingual semantic search system was built using retrieval-augmented generation (RAG), enabling accurate query understanding and rewriting across English, Hindi, and Tamil. This significantly improved search relevance and ensured users could discover products naturally in their preferred language.
To enhance engagement and conversions, a real-time recommendation engine was implemented, delivering personalized product suggestions, similar items, and dynamic bundles based on user behavior, preferences, and contextual signals across all channels.
On the supply side, Viston introduced SKU-level demand forecasting using machine learning models that operated at a micro-region level. These forecasts were integrated with supply planning workflows, triggering automated replenishment actions and enabling smarter inventory allocation.
The system was built on a scalable cloud-native architecture using Google Cloud technologies, ensuring high performance, low latency, and seamless integration with existing data pipelines. Compliance with the India DPDP Act 2023 was ensured through consent-based profiling, PII tokenization, and robust audit mechanisms.
Viston designed and deployed an integrated AI-powered merchandising intelligence layer that combined multilingual search, personalization, and demand forecasting into a unified ecosystem.
A multilingual semantic search system was built using retrieval-augmented generation (RAG), enabling accurate query understanding and rewriting across English, Hindi, and Tamil. This significantly improved search relevance and ensured users could discover products naturally in their preferred language.
To enhance engagement and conversions, a real-time recommendation engine was implemented, delivering personalized product suggestions, similar items, and dynamic bundles based on user behavior, preferences, and contextual signals across all channels.
On the supply side, Viston introduced SKU-level demand forecasting using machine learning models that operated at a micro-region level. These forecasts were integrated with supply planning workflows, triggering automated replenishment actions and enabling smarter inventory allocation.
The system was built on a scalable cloud-native architecture using Google Cloud technologies, ensuring high performance, low latency, and seamless integration with existing data pipelines. Compliance with the India DPDP Act 2023 was ensured through consent-based profiling, PII tokenization, and robust audit mechanisms.
Results
The implementation delivered measurable improvements across key business metrics, directly impacting revenue, efficiency, and customer experience.
Search performance saw a significant uplift, with click-through rates increasing from 8.2% to 17.3%, driven by improved relevance and multilingual query handling. Conversion rates increased from 2.4% to 3.3%, reflecting better product discovery and personalized experiences.
Average order value (AOV) grew by 14%, rising from INR 1,480 to INR 1,689, as customers engaged more with bundled and recommended products. At the same time, customer acquisition cost (CAC) decreased by 18%, indicating more efficient marketing and higher-quality traffic conversion.
On the operations side, stockout rates across the top 1,000 SKUs dropped from 13.5% to 10.5%, supported by accurate demand forecasting. Forecast accuracy improved significantly, with MAPE reduced from 28% to 14%, enabling better planning and reduced wastage.
Overall, the business achieved a 2.1× improvement in search CTR and unlocked growth in Tier-2 and Tier-3 markets by delivering localized, language-aware personalization. The system also ensured scalability for peak festive demand while maintaining low latency and high reliability.
The implementation delivered measurable improvements across key business metrics, directly impacting revenue, efficiency, and customer experience.
Search performance saw a significant uplift, with click-through rates increasing from 8.2% to 17.3%, driven by improved relevance and multilingual query handling. Conversion rates increased from 2.4% to 3.3%, reflecting better product discovery and personalized experiences.
Average order value (AOV) grew by 14%, rising from INR 1,480 to INR 1,689, as customers engaged more with bundled and recommended products. At the same time, customer acquisition cost (CAC) decreased by 18%, indicating more efficient marketing and higher-quality traffic conversion.
On the operations side, stockout rates across the top 1,000 SKUs dropped from 13.5% to 10.5%, supported by accurate demand forecasting. Forecast accuracy improved significantly, with MAPE reduced from 28% to 14%, enabling better planning and reduced wastage.
Overall, the business achieved a 2.1× improvement in search CTR and unlocked growth in Tier-2 and Tier-3 markets by delivering localized, language-aware personalization. The system also ensured scalability for peak festive demand while maintaining low latency and high reliability.