The Future Of Forecasting: How AI Can Help Industries Predict Demand For Products That Don't Exist Yet

Forbes

5 days ago

Dileep Kumar Rai is a Global Supply Chain Optimization Expert, Oracle Fusion Cloud architect, and demand forecasting leader.
Launching a new product in today's market isn't just risky; it's like setting sail into uncharted waters during a storm, with no compass and relying solely on your instincts to guide you. Throughout my years of working alongside leaders in the automotive, tech, fashion, FMCG and pharmaceutical industries, I've uncovered a widespread yet underappreciated challenge: the 'no-data' dilemma.
When introducing something genuinely new to the market, the usual tools—such as ERP forecasts, historical sales curves and even seasoned intuition—often fall short. The result? Overproduction, lost revenue or empty shelves.
My Industry Finding: The 'No-Data' Dilemma
Through extensive research and direct collaboration with industry innovators, I have identified four recurring pain points that keep executives awake at night:
This isn't just an abstract problem; it's a daily operational reality that costs companies millions in missed opportunities and excess inventory.
The Cross-Industry AI Forecasting Framework
To address this, I developed a proprietary and adaptable AI/ML-powered forecasting framework that combines the art of human judgment with the science of machine learning, tailoring it to each industry's unique dynamics.
The Secret Sauce: Understanding The Framework
Consider launching a new product as similar to crafting a new recipe:
• Data Ingestion: My 'pantry' is stocked with both internal ERP staples and the freshest external ingredients: social buzz, competitor actions, and economic factors signals.
• Feature Engineering: Here's the spice rack mixing predictors like launch buzz, product attributes and campaign reach to create a distinctive flavor profile for every launch.
• Tailored ML Models: These reflect the cooking techniques of XG-Boost for FMCG, LSTM (long short-term memory) for fashion, and Bayesian models for pharma, each adapted to the industry's unique texture and volatility.
• Scenario Simulator: This is my test kitchen, where I experiment with different launch 'recipes,' adjusting prices, channels and competitor responses to determine which flavors succeed.
• Executive Dashboard: The tasting table provides real-time insights, confidence intervals and scenario comparisons, empowering leaders to select the best options before introducing them to the market.
Industry-Specific Flavor Profiles
No two industries have the same palate. This framework adapts like a master chef:
Key Drivers: Seasonality, social buzz
Modeling Approach: LSTM combined with NLP to capture fast-changing trends and consumer sentiment.
Key Drivers: Economic cycles, supply chain lag
Modeling Approach: DeepAR paired with time series boosting to handle production volatility and macroeconomic shifts.
Key Drivers: Retail promotions, regional preferences
Modeling Approach: XGBoost with promotion uplift models to track short-term promotional lifts and geographic variability.
Key Drivers: Clinical trial data, prescriber behavior
Modeling Approach: Bayesian Models combined with ARIMAX to incorporate clinical data uncertainty and prescribing trends.
The Unique Ingredient: Comp Chaining For New Product Launches
A cornerstone of my framework is a technique I call Comp Chaining. This method enables us to 'borrow' the sales history of analogous products and blend it with new launch scenarios, even when there's no direct precedent.
Here's how Comp Chaining works:
• Choose A New Product: Select a target new product and a source-comparable (Comp) product.
• Align Sales Histories: Shift the company's sales timeline to match the new launch, week by week.
• Adjust For Differences: Use ratios and statistical adjustments to account for differences in initial orders, program length or market conditions.
• Blend With AI Forecasts: Where the Comp's history runs out, machine learning fills in the gaps, seasoning the forecast with real-time external data.
• Simulate And Refine: Run multiple scenarios, adjusting the blend until the forecast matches the unique profile of your new product.
This approach doesn't just copy the past; it creates a new recipe, tailored for today's market and tomorrow's challenges.
How To Bring This Framework To Your Organization
Final Thoughts: Turning Uncertainty Into A Strategic Ingredient
Forecasting demand for new products was once a tedious and uncertain game of guesswork. This framework demonstrates that it doesn't have to be that way. By blending real-world data, AI and Comp Chaining, you can turn uncertainty from a risk into your secret ingredient for growth. In a world where innovation cycles are accelerating and consumer attention is fleeting, the question isn't whether you can afford to add this flavor to your forecasting. The real question is: Can you afford not to?
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?