Assessing pre-owned smartphones manually is not easy nor efficient. First, there’s a problem of scale—massive amounts of trade-ins are conducted every year. Here at HYLA, we processed 3 million pre-owned phones in 2019, which amounts to over 10,000 phones per day. Hiring the number of workers needed to assess these phones purely by hand would be cost-prohibitive, especially since some of these phones aren’t able to return much value other than the cost of their recycled raw materials.
In addition, there’s a subjectivity problem. No matter how much training is provided, a phone that looks badly damaged to one worker might look mildly damaged to another. Workers can make errors in judgement, get distracted, or slow down unpredictably. Meanwhile, we need to maintain extremely high throughput while also maintaining high accuracy.
To solve these issues, we automated our operations. Using our automation and machine learning initiatives, we’ve been able to achieve throughput and accuracy while increasing both quality and efficiency which reduced our operating costs. Here’s a look at our approach.
Why is it Important for Us to Automate?
Pre-owned smartphones are a commodity—and they depreciate. HYLA is one of the world’s largest sellers of these devices, and so we’ve been able to get a good sense of how a pre-owned smartphone declines in value over time. According to our analytics, a pre-owned smartphone loses one percent of its value every week.
As such, it’s important that once phones are received that they are charged, data wiped, graded and then sent to inventory with deliberate speed. This means understanding which phases in this process represent bottlenecks, and which bottlenecks can be reduced. The phone charging stage could represent a bottleneck, for example, but we check charge levels when devices arrive and only charge the ones required. Further we are able to charge a large number of devices at the same time. Another area that we have improved upon is the diagnostics process, which has greatly aided by automation and machine learning.
Diagnostics is vital to HYLA’s processes because it informs the way that we grade our phones. Higher grades of pre-owned phones get resold at higher prices. Understanding what grade a phone falls into is central to our business model. Consistent grading builds confidence with buyers.
Our facility is set up so that it can perform a comprehensive assessment of every smartphone we receive, with factors including:
- Carrier locks
- Lost and stolen status (e.g., Find My iPhone)
- Speakers and headphone jack
- Bluetooth and other connectivity
Introducing Vision Tunnel
Cosmetic grading is one of the most complex and important series of tests we perform.
The equipment we use to cosmetically grade pre-owned smartphones appears to be relatively simple—it’s a conveyor belt with a camera in it. Under the hood, however, we’ve created a powerful application that can make complex judgements and decisions.
We have developed a machine learning application based on data from the hundreds of thousands of phones that we’ve already processed. Using this training data, our application can recognize when a phone has a scratched or cracked screen or body, understand the extent of the damage, and then grade the smartphone based on its assessment.
A cracked screen can lower the value of a device by more than 80%, so it’s important that we achieve a high level of accuracy—but speed and consistency also remain important. With the addition of Vision Tunnel, the total time spent assessing a phone—as well as all other tests mentioned above—takes just about two minutes, several minutes shorter than it can be done manually by an operator. It also allows an operator to work on more devices at the same time and focus on more complex tasks improving operator productivity. What’s more, Vision Tunnel can intelligently route smartphones based on the grade it assigns. This means, for example, that we won’t spend time testing the phone lock, camera, etc. if we know that the screen is too badly damaged for it to be resold. This saves even more time on testing.
Results of Automation at HYLA Mobile
As a result of implementing automated diagnostics at HYLA, we’ve been able to increase the number of devices processed per person-hour and reduce shelf time—in other words, we get phones through our warehouse to minimize value losses. What’s more, our machine learning application performs uniform assessment of every device and provides a greater detail of accuracy than a human can provide. Linked with data-driven analytics from HYLA Device IQ, we are able to accurately calculate the value of each device.
For more information about HYLA and how we can help you with trade-in and asset recovery programs, feel free to contact us today!