Without a correlation between your AI datasets and the desired outcome of your AI model, you will suffer from the Garbage In, Garbage Out problem. No matter how hard you polish, you won’t find a diamond.
Why is AI important to the Telecom & ISP sector?
As with many industries, AI has the potential to radically transform the Telecommunications and ISP sectors. Since much of the business is digital, AI can have a larger impact in this sector than many others.
AI in telecoms could help:
- Fix customer problems faster
- Predict customer & network issues before they happen
- Understand customers behaviour
- Recommend new products and services
- Drive revenue by making product and service recommendations to customers
- Reduce costs through automation
Given the sectorial challenges around profitability, AI represents a great opportunity to radically reduce costs and increase revenue.
ISP AI efforts today
Initial efforts have often gone into customer contact centres and trying to offload as much as possible to intelligent chatbots and so on. Customer care is a significant cost centre for ISPs, so this makes a lot of sense. By using AI, interactions can become more useful and cover more areas, eliminating the need for human interactions.
Another huge cost centre for ISPs is of course Churn. We discussed how Sandvine can help with Churn prediction in this article. For this, AI efforts today are generally focused on using three datasets:
- Network KPIs
- Customer NPS feedback
- Direct customer contact
AI learning and training phase
Operators are investigating how to use AI to try to deduce broadband service quality from these three datasets.
The Garbage in, Garbage out problem
Whilst deliberately provocative, it is an absolute truth that AI is only as intelligent as the data it is trained with.
Network KPIs can be misleading due to their accuracy. Typically, routers and other network elements are not designed to give KPIs capable to deduce the user experience. Their KPIs are based on network operational needs. Throughput may be sampled over a number of seconds for example. This is fine for operations, but almost useless to understand the user experience.
How you measure determines the results
NPS feedback and analysing customer contact is all too late and opinion based, making proactive or near real time use cases very difficult.
How can the Sandvine dataset help AI?
Sandvine brings new visibility of the User Experience, as discussed on this page. A simple 1-5 MOS style measurement is provided for almost every user connection. This can give a comprehensive picture of the overall User Experience an individual mobile user or household is receiving.
In addition to the user experience, Sandvine gives unprecedented visibility of application usage patterns at an individual household or mobile user level. This in turn allows for a deep understanding of your customers preferences – gold dust for marketing teams.
Fact based, real-time and historical user experience data
Sandvine’s long term dataset allows for trend analysis, giving AI the ability to:
- Find user experience degradations, enabling pro-active actions.
- Determine likely root cause of issues.
- See emerging network and application usage trends quickly.
- Build personalisation offers
With a real-time feed, AI can use Sandvine data to:
- React instantly to user issues.
- Create and push out real-time marketing opportunities.
Sandvine insights dataset is accessible via ODBC or Kafka, depending on the use case.
Summary
Sandvine’s dataset can be a critical component in an ISP’s AI efforts. Sandvine’s App QoE data gives the most granular user experience scoring data AND visibility of application usage. This can drive ARPU increases and cost reductions.
Sandvine has always had industry leading data. Now combined with AI powered analysis, the full power can finally be unleashed.
If you’re interested in learning more about our data, contact us.
Topics: Quality of Experience, App QoE, App Quality of Experience, AI, Artificial Intelligence
