Trade-in automatization software

Our client provides a trade-in service where customers may hand over their device and receive a discount for a new model. The problem was that estimations of old devices might be wrong due to incompetence or illicit behaviour of the trade-in managers. Therefore, the task was to make the process of testing and price estimation automated to exclude human factors.
We have created a distributed system containing central server software, web-application for a trade-in manager workplace and mobile apps for iOS and Android for automated testing of the smartphones.
Volume: 2 person-years
Services: Project management, prototyping, analytics, ML-modelling, architecture development, design, code development, documentation, QA, tech support.
Duration: 2017-2018
Technologies: Angular, Bootstrap, Java, SpringBoot, PostgreSQL, iOS (Objective C), Android (Native Android Java)
Average estimation of similar project: $ 60 000
The project started from the prototyping stage, we have tried various methods of smartphone testing, including automatic screen scrapes detection using ML classification algorithm and simpler inbuild self-checking utilities provided by smartphone manufacturers. As a result of that stage an automated scoring method was developed, which estimates more than 40 parameters of devices. Based on this method the client has the possibility to automatically set a recommended price range for any device.
At the second stage we have created a production version of the software. A central trade-in server, Android and iOS apps, a web-application for the trade-in manager workplace were developed. The system was tested for all supported devices and different distributed configurations.
The final stage was distribution and adaptation. Our client did distribution by itself, but we provided technical support and promptly corrected the system according to feedback from a trial operation.
Solution Architecture:
The initial conception of this project was to provide a tool for the client to rapidly correct scoring models for any new smartphone or price changes without code modification. An intuitive scoring model editor has been developed as a part of the system.