Accounting for losses and breakage is an imperative part of business operations and it is becoming increasingly difficult to do so when technology and personal devices are involved. In mobile fleet device management, it is cumbersome to track down users when a ‘failure to restore’ happens on a device or in some cases several devices within a workforce. Take for example a scenario where devices are used in a public setting such as patient use in healthcare or airline passengers within an aircraft – in both cases, getting an account for what happened on a device after it is used is not possible.
So how do we account for these issues? Cluster analysis.
Cluster analysis is a technique used in machine learning that attempts to find clusters of observations within a dataset. It is an unsupervised learning algorithm, meaning the amount of clusters within the dataset is unknown before running through the process. The goal of is to find clusters such that the observations within each cluster are similar to each other, while observations in different clusters are different from each other. For example: streaming services use cluster analysis to identify which viewers have similar viewing behaviour to properly allocate advertising spend where it counts most.
Through cluster analysis, M3 Solutions was able to detect a common sequence of datapoints with several devices a client was using to resolve the overarching ‘failure to restore’ problem.
Adapting to digital landscapes
In response to the changing landscape Covid brought, this client introduced use of their tablets to non-employees. That business decision brought with it new applications on the devices: specifically, the use of the YouTube app. A small number of devices would get returned and fail to restore on any given week and would have to be removed, have the Terms of Service accepted, and be reinserted to wipe again.
Our IT experts were able to pull iOS console logs (heavily redacted by apple and in the order of 250, 000 – 1 million lines per check-out) from all the devices to take the user’s subjective account out of the equation. The console logs ran through a pre-processor developed in house, then fed that pre-processed log file into cluster analysis software also developed in house to find the likely culprits.
Armed with the knowledge that: 95% of devices had a specific usage pattern, we focused on trying solutions that fit the usage pattern and were finally able to reproduce the problem and then solve it. With the exact sequence of steps that caused the failures identified, we explained the issue to the customer, and they changed their setup to mitigate the issue. We haven’t seen a failure like it since.
“AI Systems and affective computing are allowing everyday objects to detect, analyze, process and respond to people’s emotional states and moods to provide better context and a more personalized experience,” said Robert Cozza, research director at Gartner. “To remain relevant, technology vendors must integrate AI into every aspect of their devices, or face marginalization.”
Managing customer devices at scale is no simple feat for vendors such as M3 Solutions. On the one hand, a vulnerability that needs to be patched or an app update that breaks functionality becomes an immediate problem that can be service affecting and have a huge impact on a client’s business. On the other, having the luxury of examining every device after use allows an unprecedented opportunity to eliminate the more nuanced problems that only appear occasionally by using unique artificial intelligence techniques in software built in house. Often when a problem has no clear solution and costs of investigating start to rise, businesses learn to live with them; M3 Solutions has a proven track record of eliminating nuisances like elves sneaking in to fix shoes at night.