AWS for Industries

Delivering CPG Demand Planning at Scale and Hyper-Speed With Samya’s Dynamic Demand AI on AWS

Opportunity at the intersection of demand and supply

Complexity and volatility characterized the consumer-packaged goods (CPG) and retail industries even before the pandemic impacted all of us. According to a Boston Consulting Group study, the associated opportunity loss due to complex and volatile business is estimated to be 8–10 percent, and poor anticipation of demand drives one-third of this. The underlying challenges include low forecast accuracy, lack of agility, human bias, and high manual effort. The inability to identify contributing factors and measure their impact makes these challenges harder to overcome for many CPG manufacturers and retailers. The magnitude of these gaps is even greater at granular levels as well as with new products. In the face of disruptions, like a hurricane or a pandemic, the already fragile process starts to break down entirely. Demand planners struggle to meet demand every day while waiting for a real solution to a chronic problem.

AI can help but only if it knows its limitations

In a world of complete and perfect data, artificial intelligence (AI) might be the silver bullet. In the real world, however, AI is necessary but not sufficient. The first and most critical role of AI is identifying high-risk situations where human intervention is essential and highlighting those to the demand planner. AI can also serve up information that the demand planner might find relevant in those situations. The second and equally important function of AI is to automate the process in low-risk cases, thereby freeing up the demand planner’s time to focus on high-value opportunities.

AI-first, future-proof demand planning

Despite its limitations, an AI-first approach works because, unlike humans, AI has no ego and can objectively separate the things it can do from the things it can’t. In a post-AI scenario, there are separate paths for high- and low-risk situations. AI automates the low-risk cases and aids the demand planner in making the best possible decisions in high-risk scenarios. Newer and better data increases the proportion of low-risk scenarios, while disruptions increase the incidence of high-risk scenarios. The process thus self-regulates and remains future proof.


The Samya Dynamic Demand AI value propositionBlog_logo_box Samya_contact

The benefits of implementing Samya’s Dynamic Demand AI (DDAI) include accuracy improvement and productivity gains with proven track records with customers in both CPG and retail industries with the following outcomes:

  • 5–10 percent improvement in accuracy for select categories in 6–8 weeks
  • 5–10 percent improvement across the board before the end of year one
  • Productivity gains freeing up 30–40 percent of demand planners’ time in year one
  • Achieves industry benchmark–level accuracy as defined by Gartner in year two
  • Addresses all significant pain points of demand planners, like long-tail automation, multilevel and multi-horizon forecast reconciliation, etc.

Software-as-a-service solution, powered by proprietary AI, managed by Samya, and runs solely on AWS

To deliver these benefits, it is important to be able to achieve superior predictive accuracy without sacrificing performance, scalability, and explainability. The underlying architecture needs to be scalable, configurable, versatile and allow automation and ease of deployment. To achieve this at a reasonable cost, compute must be on demand while data is always available.

Configurability: Configurations are metadata driven, and the parameterization spans data, AI, machine learning (ML), engineering, and application pipelines.

Data integration: Input and output connectors allow easy integration of disparate data sources and systems. Direct consumption of any customer data that is already on AWS can also be enabled.

Data sources: Combination of disparate data sources—including time series of historical sales, internal drivers, syndicated data, and curated external data—facilitates robust demand forecasting.

Predictive modeling: AI Engine extracts features for seasonality, trends, and drivers and learns relationships using deep learning, ML, and econometric models. Ensemble of time series with automated (dis)aggregation and robust validation is performed to achieve best-in-class accuracy.

Actionable insights: Five to ten percent accuracy improvement within bias thresholds has been achieved in our all customer engagements so far. Forecasts are comprehensive across the short term, long term, and new products. Actionable factor contributions over baseline and low-touch decision support with risk and uncertainty measures drive faster time to decisions.

Security: Samya application is SOC2 Type 2 certified, with all security measures like storage and compute isolation and data encryption during transit and rest. AWS Identity Access Management (AWS IAM) is used to provide a fine-grained access control across all of AWS—role-based access control (RBAC), and AWS Firewall Manager is used to help you to centrally configure and manage firewall rules across your AWS accounts.

Scalability: Forecasting can be done for millions of time series through auto-scaling orchestration, and parallelization along with automating hyperparameter tuning support.

AWS architecture

Connectivity with customer databases is established through AWS Direct Connect, a dedicated network connection to AWS, or AWS Client VPN, a fully managed remote access VPN solution. Samya’s AWS implementation uses StreamSets, Spark MLlib, Django, and GraphQL running on Amazon Elastic Compute Cloud (Amazon EC2), which provides secure and resizable compute capacity. AWS Glue, a serverless data integration service, is used for data cataloguing. The data lake resides on Amazon Simple Storage Service (Amazon S3), an object storage service offering industry-leading scalability, data availability, security, and performance.

For more information, see Samya’s AWS reference architecture.

Fractal Revenue Growth Management System on AWS reference architecture


Demand planning is a day-to-day operation as well as an art. Samya offers a fully managed, AI-based solution ready for them to use. If you would like more information about Samya’s Dynamic Demand AI solution on AWS, leave a comment on this blog post. To request a demo or sign up for proof-of-value (PoV), visit Samya or contact your AWS account team today.

AWS Partner Spotlight (now Fractal Analytics)

Samya offers an enterprise AI SaaS product to help organizations unlock revenue growth at the intersection of demand and supply. Built for any data, at scale, Samya’s purpose-built forecast and prediction solutions deliver improved forecast accuracy and boost revenue growth at hyper-speed.

Danny Yin

Danny Yin

Danny (Yen-Lin) Yin is the Global Technical Lead for AWS Partners in the CPG industry. He joined AWS in 2018 with 18 years of experience in ecommerce application development and operations. Danny helps CPG companies enhance the consumer digital user experience and gain operational efficiency across different lines of business. Danny is also responsible for solutions architecture and technical guidance for CPG technology and consulting partners on AWS. Before he joined AWS, Danny was Director of Digital Engineering at Toys”R”Us, where he successfully migrated the world’s largest toy webstore from an outsourced application to an in-house hybrid cloud application on AWS.

Guha Athreya

Guha Athreya

Guha Athreya is the product and technology leader for He has over 18 years of experience in predictive analytics and product development. Guha specializes in helping enterprises future-proof their business processes. He supports reimagining the business process in a post-AI world and creating products that manifest and embody the future state. He has a keen interest in helping human and machine intelligence to drive higher levels of objectivity and reduce bias in decision making. Guha has previously worked at Grainger, AbsolutData, Volvo, and Ingersoll Rand. He has a bachelor’s degree in engineering from Bangalore University and a master’s in systems and operations from IIT Madras. He has also completed disruptive strategy from Harvard Business School.