Imagine the impact that 26% of daily purchases would have in your business if they would go south. And by that we mean that the customers' satisfaction would be highly impacted. How would that impact your Net Promoter Score and the costs related to calls and extra truck rolls to install equipment?

There are lots of debates covering the preventive maintenance aspect of services, so customers can get their uninterrupted feed of data. However, the challenges of keeping issues away from customers start much earlier in the journey: they start when customers purchase their most wanted products.

Although at VodafoneZiggo we aim to help customers to understand what they bought and how to install it themselves, sometimes it gets complicated due to outdated takeover points installed at their homes. The problem that comes along with this is that the customers won't be able to follow the DIY guide, generating excessive costs in terms of extra calls and truck rolls (when a mechanic has to come over to finish the installation). But how to solve this in a preventive way?

To approach this issue, we collected and labelled thousands of pictures of customers in-home installation, aiming to automate the identification of their takeover points prior to the product delivery. With this in mind, and a sophisticated Deep Learning model trained in house, we created a strong pipeline for the continuous integration and deployment of our solution, reaching thousands of customers on a daily basis.

Our model has been trained to identify a few components in the in-home installation, but we have also closed the feedback loop by collecting more pictures as the service is used, hence increasing our dataset and the potential of Computer Vision and Deep Learning within the telco industry.