Machine learning meets logistics
Traeger Grills partnered with Consilium to optimize logistics forecasting for their global network of 3PL partners. Using machine learning models and other AWS services, they were able to automate and improve forecast accuracy for thousands of SKUs.
The challenge
The demand for Traeger grills has been on fire in recent years. As a result, the company realized its antiquated spreadsheet-based approach to logistics forecasting was not going to scale with its business. The team turned to Consilium to help develop a modern solution that harnessed the power of data scattered across disparate systems. That data was used to power a state-of-the-art collection of AWS services that streamlined and optimized global product forecasts.
How Consilium helped
Like most companies, Traeger’s corporate data is fragmented across a myriad of disparate platforms and data sources.
Using Amazon EventBridge, Amazon Kinesis Data Firehose and Amazon S3 (Simple Storage Service), Consilium was able to extract transactional data used to model customer purchase behavior.
Data Extraction & Ingestion
Once forecast algorithms and models were tuned to acceptable tolerances (i.e., +/- 5% from actual results), the entire process was automated with forecast outputs stored within Traeger’s corporate data warehouse powered by Amazon Redshift.
Any anomalies/errors are surfaced in real-time using a combination of Amazon CloudWatch, Amazon SNS and AWS Chatbot.
Data is surfaced consistently to internal resources and external partners, eliminating the need for emails and spreadsheets.
Machine Learning Models
Traeger and Consilium iterated on a variety of machine learning algorithms using Amazon Forecast.
Through testing and analysis, it was determined that separate algorithms should be used to support distinct long-term and short-term forecasts.
Long-term forecasts leveraged algorithms that accurately model seasonal trends, while short-term forecasts were focused on driving interval and daily predictions.
Scale
Value delivered
Traeger’s business processes weren’t keeping pace with the growth of it’s business. Logistics forecasting was a collection of manual processes that required days of tedious data extraction, transformation and analysis. The results were less than ideal, which left customers waiting for parts that were needed to bring friends and family together to enjoy the flavors only a Traeger can provide.
The cross-functional partnership between Traeger and Consilium was able to develop and implement a state-of-the-art machine learning-based forecasting solution in a matter of weeks. Within a month, logistics forecasting was entirely automated.
Forecast accuracy almost doubled, which allowed external 3PL partners to more effectively and efficiently plan warehouse resources.
Results were apparent almost instantly. Instead of waiting weeks for replacement parts, customers were cooking on their Traeger within 1-2 days, which reduced the volume of customer service contacts by almost 25%.
“We knew there was a better way to forecast warehouse order volume - we just didn’t know what it was.
The Consilium team was nothing short of amazing.
In a matter of days, we had initial results. Within a week, we knew we had a viable and scalable solution. Within a month, our manual processes had been entirely automated.
We went from hoping we were accurate to knowing we were.”
-Corey Savory-Venzke, VP of Customer Experience, Traeger Grills
Forecast Accuracy
+82%
Cost Savings
~$1.2M
“I’ve been fortunate to have spent 20 years driving customer experience for a variety of global consumer brands. I’ve never seen anything close to what we’ve built with Consilium and AWS.”
— Corey Savory-Venzke, VP of Customer Experience, Traeger Grills
