Rise of the Machines: How TeleSign Uses Data Science and Machine Learning | TeleSign

We at TeleSign are always looking for ways to improve results for you and your users. Successful user outcomes are critical to the growth of your business. We understand that you want to sign-up good users, protect their accounts and enable experiences that help drive user retention and engagement. We have a team of analysts reviewing data to identify patterns of good or bad behavior and produce actionable insights. Furthermore, we are excited about our forays into data science and machine learning that supplements our human insight and renders enhanced decisions for you.

Judgement Day

Allow me to take you on a journey of how we protect you and your users. This journey begins with our risk assessment product, Score. You provide us any phone number in the world, and we can give you the risk prediction of doing an online transaction with the owner of that phone number. This helps you eliminate potential fraudsters so that the good users get value out of your product or platform and the bad are left frustrated, twiddling their thumbs. In order to provide this risk prediction, we apply machine learning to user supplied data, proprietary and third party data sources and transactions seen across our platform. We run these calculations over millions of daily data records and make prediction models cutting across countries, industries, use cases and customers. Our production data scoring platform is built & optimized for fast queries of data across geo-replicated databases and scoring servers that return responses in milliseconds.

Terminating Spam

Next, we apply machine learning to filter spam on our SMS platform. Spam is a frequent problem that affects both end-users and businesses engaging with these users. We understand the pain of getting your reputation ruined by even an isolated incident, that’s why TeleSign is here to help! Our dedicated team of compliance professionals get ahead of these problems for a living. We are enhancing our spam detection capabilities by applying natural language classification techniques to the content on the SMS. Again, the content of the SMS is only inspected by a handful of approved analysts under certain conditions when they have a clear mandate to do so. Further, the data is only retained for few days as mandated by the SMS regulations.

We’re also applying machine learning to optimize routes for delivering SMS with higher certainty and better cost. As you can imagine there are many different ways to send a text, we work with analysts and computers to figure out the ideal way to send yours.

Lastly, we apply machine learning to predict success or failure of phone verification using our product called App Verify. The approach we use offers a better & faster user experience, higher conversion rates and lower cost than traditional phone verification methods such as SMS and voice call verification. We are now taking App Verify to the next level where we want to predict and minimize failures. How? With data science of course. By training models to better predict failure or success scenarios, App Verify can reduce failures by up to 50% streamlining a process that saves time and money, positively impacting your bottom line.

Here at TeleSign we are just getting started with the fun part of data science and machine learning. We promise there is so much more to come. We have a strong commitment to improving the lives of you and your users, so don’t worry. We’ve always got your back!


Integrate our products seamlessly into your user experience.