40 million datapoint machine learning service brings fleet SMR into “new era”

A new machine learning value-added service from epyx that uses more than 40 million datapoints is described as bringing fleet service, maintenance and repair (SMR) authorisation into a “new era”.

AVA – or Automatic Vehicle Authorisation – integrates with the 1link Service Network platform, used by operators totalling almost five million cars and vans to manage their SMR.

In normal use, epyx expects it to substantially reduce the number of jobs requiring human authorisation, although there is potential to increase or reduce that figure depending on tolerances set by each fleet.

Charlie Brooks, vice president of strategy, growth and data at epyx, said: “’We view AVA as a development that maximises the potential of the latest technology and also increases the value of human intervention where needed.

“It is arguably the biggest single advance for 1link Service Network in its 21-year history and, we believe, brings SMR into a new era. The platform already includes a rules engine that allows some jobs to be automatically authorised but AVA adds several new layers of sophistication and adaptability.”

Using SMR data from the last five years of transactions through 1link Service Network, AVA offers recommendations to approve or review individual jobs across four key areas.

These are: authorisation propensity – based on the type of work, previous negotiations, garage type, mileage and more; prior pricing – rating the proposed price for a job based on similar transactions; item frequency – looking at how often a specific supplier recommends this type of job; and wear rate – showing how the job compares to data covering similar vehicles based on when they have needed the same work.

Recommendations in each of these areas are shown on a scale of 0-100 for both the individual fleet customer and the complete 1link Service Network dataset. A configuration tool allows users to set weightings across a series of key parameters in line with their priorities. 

Charlie said: “The majority of jobsheets submitted by garages are completely legitimate and approved without any changes. AVA is a machine learning platform that helps remove the need for these to be reviewed manually, allowing fleets to potentially auto-authorise jobs in a much more subtle and effective manner than ever before, reflecting exactly their particular approach to SMR management. 

“The advantages are numerous. The speed of response to garages should increase dramatically, as should consistency across similar jobs for each fleet, auto-authorisation of jobs out of hours will be a very real possibility, senior staff can focus on higher value jobs requiring detailed knowledge, and junior staff will be supported through decision making and training until they become fully competent.

“How the technology is used will very much depend on each fleet but we already have some users who auto-authorise a large proportion of their SMR jobs using our existing rules engine and AVA is capable of making more of those decisions, much more effectively.

“Also, on average almost 18,000 SMR jobs are processed through 1link Service Network every day, so our dataset is growing all the time. The more of this unique information we have, the better the outcomes.”