HVAC Manufacturer Uses EZForecast to Transform Forecasting and Build Planning

Quick Overview

Client: HVAC/Industrial Heating Manufacturer 

Technologies used: Microsoft Azure, Wavicle EZForecast, and Power BI

Goal: Improve demand forecasting accuracy and enable efficient, load-level & scalable production planning. 

Challenges: Legacy, Excel-based forecasting and planning processes delivered low accuracy, lacked visibility into seasonality and external signals, and required significant manual effort across planning teams. 

Solution: Wavicle implemented EZForecast on Azure, combining machine learning, external data signals, and business expertise to deliver accurate 12-month forecasts, intelligent build planning / MPS that were showcased in intuitive dashboards. 

Result: Automated ML-driven forecasting using EZForecast improved forecasting accuracy from 66% to ~74% with 90% reduction in planning effort.

The Challenge

This HVAC industrial heating manufacturer was operating in a highly demand-driven environment where forecasting accuracy and planning speed directly affected revenue, inventory, and service levels. However, three core challenges were holding the business back:

• Limited forecasting accuracy and insight 
  Forecasting relied on basic statistical methods and a legacy tool, delivering only 66% accuracy and consistently under-forecasting demand by nearly $17M. The approach lacked the ability to incorporate seasonality, promotions, or external signals such as weather data and construction indices, leaving planners reactive rather than predictive.

• Manual, excel-driven build planning 
  Production and inventory planning were managed through fragmented Excel workflows. Monthly build plans took several days to create and reconcile, with no systematic visibility into safety stock, capacity constraints, production loads, or customer-level demand.

• An underutilized data and cloud foundation 
  Although Azure infrastructure was already in place, it was not designed to support advanced analytics or machine learning. Data was inconsistent and unstructured, legacy on-prem systems limited scalability, and teams lacked a unified platform to operationalize forecasting and planning at scale.

To move faster and plan with confidence, the manufacturer needed a modern, cloud-based approach to forecasting and planning.

Solution

Wavicle’s solution for the manufacturer followed a structured, multi-phase approach that unified cloud readiness, data engineering, advanced forecasting, business customization, planning, and inventory alignment into one seamless operating model. Here’s the seven-step journey from data to smart decisions:

• Built a scalable cloud foundation
  Laid the groundwork for AI-driven planning by designing and deploying a modern Azure-based data and ML foundation, including cloud infrastructure and automated pipelines to support scale and growth.
• Modernized legacy planning systems
  Moved forecasting and planning from on-premises systems to Azure, creating a more agile, future-ready environment.
• Prepared data for AI and analytics
  Standardized, cleaned, and enriched historical data to ensure it was reliable, consistent, and ready for forecasting and planning use cases.
• Delivered intelligent demand forecasting
  Implemented EZForecast to generate accurate 12-month demand forecasts using multiple machine learning models at the item and customer level.
• External demand signals integrated
  Forecasts were enriched with real-world signals like weather and construction activity (Dodge Construction Index), helping teams anticipate demand shifts and plan with greater confidence.
• Enabled smarter production planning
  Extended forecasting into build planning, aligning demand, inventory, safety stock, and production capacity to create a stable, optimized year-round plan.
• Unified decisions with full data visibility
  Brought everything together through intuitive dashboards in Power BI, giving teams a clear, shared view of forecasts, inventory, service levels, and production performance.

Implementation timeline

• Proof of concept (6 weeks): Validated measurable forecasting accuracy improvements using EZForecast.
• Full implementation (3 months): Delivered an end-to-end forecasting, build planning, and dashboard solution on Azure.
• Total time to value (4.5 months): Implemented in months, compared to the 2+ year timeline typically required for a ground-up custom planning system.

Results

In under five months, Wavicle transformed the manufacturer’s forecasting and production planning using a cloud-native, ML-driven approach built on Azure and EZForecast. 

• Higher forecasting accuracy with financial impact
  ML-driven forecasting using EZForecast improved accuracy from 66% to ~74% by factoring in seasonality, promotions, weather data, and the Dodge Construction Index, reducing under-forecasting from $17M to $5–6M.

• 90%+ reduction in planning effort through automation
  Excel-based workflows were replaced with an automated planning engine on Azure, cutting planning cycles from days to hours for both master and divisional planners.

• Better service levels and production stability
  Real-time visibility into inventory, capacity, and production enabled backlog reduction during peak months and supported smoother, more predictable monthly builds.

• Fast delivery, built to scale
  The solution was delivered in 4.5 months using Azure ML, CI/CD pipelines, and automated data pipelines, far faster than traditional multi-year planning system implementations.   

Forecasting & planning transformation 

What’s next 

With a scalable Azure foundation in place, the manufacturer is positioned to extend EZForecast across additional plants, acquisitions, and make-to-order planning while introducing agent-based planning capabilities already prototyped and tested.