Customer Success Stories from AWS Machine Learning Competency Partners
At the AWS London Summit, we announced the new AWS Machine Learning Competency Program for AWS Partner Network (APN) Consulting Partners.
Launch Partners in this program have deep expertise and proven customer success in Machine Learning on Amazon Web Services (AWS), and we are excited to showcase some of their stories.
The AWS Competency Program helps customers identify and choose the top APN Partner for their AWS projects and workloads.
Regit & Peak Business Insight
About the Customer: Formerly Motoring.co.uk, Regit is an automotive tech firm and the UK’s leading online service for motorists. They deliver digital car management services based on a car’s registration plate, and provide drivers with informative reminders such as Ministry of Transport (MOT) tax, insurance, and recalls.
The Story: Regit generates leads for companies in the automotive industry, and provides a way for people to book test drives, buy and sell cars, and request brochures. Regit was not able to predict which users were likely to change cars, or which users had changed cars, until after the event had occurred. That’s why they chose Artificial Intelligence System (AIS) from Peak Business Insight, an APN Advanced Consulting Partner, to pull together user data from Regit’s website and marketing systems.
Peak applied “Categorical Machine Learning models” that handle both category and variable data simultaneously to give predictions about the likelihood of users changing cars, resulting in a sale for Regit. Peak used AWS services such as Amazon SageMaker for real-time ingestion, modeling, and data output.
Amazon SageMaker handles 5,000 API requests a day for Regit, seamlessly scaling and adjusting to relevant data requirements and managing the delivery of lead scoring results. Meanwhile, Amazon Redshift and Amazon Elastic Compute Cloud (Amazon EC2) instances efficiently and continuously optimize model performance and results.
The Result: With Peak, Regit has been able to predict which of its 2.5 million users are going to change cars and when. This means they can serve customers in a more personalized and targeted way, increasing call center revenues by more than a quarter.
In Regit’s own words:
“Working with Peak, we’ve had a 27 percent increase in sales. We’ve also been able to reduce operational costs by up to 35 percent by simply staffing the call center at the times of day most likely to result in a sale to a user or customer. Not only are we now offering improved services to our customers, increasing their satisfaction levels, we’ve been able to grow our revenues in doing so.” – Terry Hogan, CEO at Regit
CQ Roll Call & Quantiphi
About the Customer: CQ Roll Call, a publishing company and part of the Economist Group, produces a number of publications that report primarily on the U.S. Congress. In addition, CQ Roll Call provides essential intelligence and grassroots advocacy resources. They are responsible for creating advanced legislative tracking platforms for both Federal and State levels of government. CQ Roll Call helps customers in government relations to shape policy and influence change.
The Story: CQ Roll Call had a difficult time using objective metrics to evaluate the likelihood of a bill passing and of a particular member of Congress voting for, or against, a bill. Quantiphi, an APN Advanced Consulting Partner, generated two predictive models based on over 20 years of data provided by CQ Roll Call. These models relied on more interpretable classification algorithms (logistic regression, random forest, etc.) instead of ones like neural networks since end users must identify which features are most influential in predicting a bill passing or failing. Through iterative feature selection and engineering, along with the aforementioned algorithms, Quantiphi achieved impactful results.
To enable automation of the bill passage likelihood model on a daily basis, an AWS CloudWatch event was created that started and stopped an Amazon EC2 instance at specific times during the day. A higher instance type of EC2 ensured the growing dataset size was appropriately handled. Once the data is tested, the final probabilities and relevant identification columns are saved in an Amazon Simple Storage Service (Amazon S3) bucket. For testing purposes, an AWS Lambda function could be accessed through an Amazon API Gateway. To ensure quick response times, the file in the S3 bucket was condensed to only include relevant subsets of the data.
The Result: The predictive model built by Quantiphi evaluates each bill and each representative in the current session of Congress every day, as well as each representative’s likelihood to vote for each bill. By the end of a session, this amounts to approximately 10,000 bills and 5.3 million predictions associated with every Congress member’s vote likelihood on each bill. This allows users of CQ’s platform to better predict the likelihood of a bill passing and where Congressional votes are most in question.
Hg & Inawisdom
About the Customer: Hg is a private equity firm and sector expert investor, committed to building ambitious businesses across the technology and services space, primarily in Europe. Hg has over £9 billion funds under management, invested across over 30 portfolio companies.
The Story: Hg specializes in B2B sectors that are often naturally “data rich” and have the potential to deliver tangible bottom-line impact through data insight. However, at the start of Hg’s investment in a company, this data can be disparate, disconnected, or spread across a variety of internal and external systems. Hg’s Operations Innovation (OI) team helps management teams unlock the value of this data.
Inawisdom, an APN Standard Consulting Partner, recognized the requirement for an architecture that could be deployed for Hg and then consistently scaled by deploying dedicated platforms for their portfolio companies. Using their Rapid Analytics and Machine Learning Platform (RAMP), Inawisdom leveraged automated AWS CloudFormation templates for the rapid deployment of tailored, but consistent, data platforms for Hg and their portfolio companies. This includes data ingestion and cataloging (via a data lake), data science (using the Deep Learning AMI) and visualization technologies. The provisioning of the RAMP platform is rapid and fully automated.
In addition to partnering with Inawisdom, this forward-thinking approach led Hg to strategically partner with AWS as the recommended cloud provider for their portfolio of companies. The RAMP platform enabled the delivery of business insights and embedded predictive models for a number of Hg portfolio companies.
The Result: Hg’s in-house Data Science team now uses a cost effective and secure on-demand data platform to deliver and run Machine Learning models, leading to optimized sales conversion rates and better understanding of sales pipelines. Typically, the very first outputs from these engagements create an immediate return on investment.
The flexibility in the operating landscape and the ability to scale (up or down) on-demand lets Hg’s Data Science team focus on creating value from the data, rather than worrying about whether they have sufficient processing power or storage, for example. All the deployments conform to Hg’s agreed product vendor preferences in order to leverage economies of scale in both licensing and skills.
Inawsidom’s RAMP-based AWS data platform uses automated scheduling to minimize run costs. On-demand AWS services are used to minimize not only EC2 costs but also third-party license costs, allowing some services to be running less than 1 percent of the time while still delivering the business outcome.
Atlas Van Lines & Pariveda
About the Customer: Atlas Van Lines is the second largest van line in North America, formed in 1948 by a group of entrepreneurs in the moving and storage industry. The organization was developed with the single goal of moving from coast to coast while adhering to the golden rule of business. In addition to a robust footprint, Atlas boasts stringent agent quality requirements that surpass that of the industry.
The Story: During peak moving seasons, the Atlas agent network works together across markets to meet customer demand. Traditionally, their ability to forecast capacity was manual and labor intensive. They relied on the wisdom and gut instinct of resources with many years of experience. Atlas had the historical data from 2011 forward and desired to find a way to dynamically adjust capacity and price based on future market demands.
Pariveda Solutions, an APN Premier Consulting Partner, has guided many clients on the journey to realization of Machine Learning as a business capability by taking a value-driven approach to planning and executing projects using their proprietary AgileML methodology.
The Result: Pariveda and AWS joined together to help Atlas unlock the possibility of proactive capacity and price management in the long-haul moving industry. Pariveda prepared the data, developed, and evaluated the Machine Learning model, and tuned the performance. They used Amazon SageMaker to train and optimize the model, and then exported it using Amazon SageMaker’s modular nature to run using Amazon EC2.
About the AWS Competency Program
To receive the AWS Competency designation, APN Partners must undergo rigorous technical validation related to industry-specific technology, as well as an assessment of the security, performance, and reliability of their AWS solutions. This validation gives customers complete confidence in choosing top APN Partner solutions from the tens of thousands in the AWS Partner Network.
The key value of the AWS Competency Program is to build customer trust by helping organizations choose the top APN Partners based on workload, solution, or industry designation. The AWS Competency Program identifies, validates, and differentiates APN Partners that have demonstrated customer success and deep specialization in specific solution areas or segments.