Fabrizio Trovato
Data Scientist

Published: August 14, 2024

The unsung hero of AI: How machine learning is driving Telco optimisation

In the dynamic and crowded world of AI-driven tech, finding a product that’s genuinely cutting-edge can feel almost impossible, and what is often missed in the mix of Artificial Intelligence (AI), is the power of machine learning. In this blog, we’ll cover how the team at 15gifts are actively utilising machine learning to drive increased user engagement for our guided selling engine, through a feature called ML Engage.

ML Engage is designed to identify and accommodate the individual needs of online consumers, creating a personalised sales experience. Particularly in the competitive Telco sector, it is crucial to go beyond generic offerings and deliver exactly what the customers need, precisely when they need it. ML Engage does just that – tailoring each online interaction to individual preferences and suppressing needless interruptions for users who won’t benefit from it.

What is ML Engage?

While Machine Learning (ML) is part of the broader AI landscape, what sets it apart is its more refined approach – ML only performs the specific job it was trained to do, following its unique purpose. ML uses statistical algorithms and data points to optimise the decision-making process and ultimately choose the most efficient outcome.

ML Engage is a machine learning-driven solution designed to enhance user engagement on our clients’ websites. The goal is to be able to consistently identify users who can benefit from using the engine, and enable targeted optimisation opportunities.

By understanding what each user needs and likes, ML Engage can create an experience that’s truly unique to each individual. The result is a user-focused approach that tailors the digital experience, making each interaction more relevant, engaging, and ultimately, more successful.

The evolution of 15gifts engagement strategy

The team at 15gifts are always looking at ways to improve the performance of our guided selling engine, ensuring maximum user satisfaction and ultimately driving more sales for our clients. Over time, we noticed that an increasing number of users were opening the engine but choosing not to engage with the initial question. This signalled the need to further refine the engine’s engagement strategy.

In January 2023, we began the implementation of ML Engage, aiming to introduce a more tailored approach while mitigating any “accidental engagement”, which can lead to customer annoyance and intrusiveness. With ML Engage in place, the engagement bubble appears exclusively for those who show all the right characteristics of a user in need of assistance.

To identify the right users, our ML Engage model underwent extensive training to learn exactly who to target and when. It considers factors such as user behaviour, traffic source (direct, organic, paid etc), page visits (to the customer’s website), device, and browser type. The engagement bubble only appears for users with a high likelihood of interaction, improving the experience for both engine and non-engine users.

Since implementing ML Engage on select client websites, we have seen a +11.2% increase in conversion rates per engagement bubble displayed and an increase in questions answered per engine activation. These encouraging results show that we are successfully identifying and engaging the right users, leading to increased conversions and user satisfaction.

Lessons learnt and continual development

Just like any new endeavour, it’s been a learning curve. As we recognise evolving trends, we identified the need to shift our engine. When we first began the development of ML Engage, it took a while to drive the performance we had anticipated. One factor we identified quite quickly,  was addressing the challenge of intrusiveness on smaller mobile devices. Since then, we’ve made great progress with our ML Engage solution, addressing the underlying issues and establishing an effective model.

Looking ahead, our focus is on continuous improvement. We’re exploring a hybrid approach to further refine our methodology, reducing engagement visibility for less likely users, while amplifying it for those predicted to engage.

Book a demo with us to learn more about how ML Engage could help increase conversion rates and your customers’ confidence when shopping online.

Humanisation, anthropomorphism and intuition

No matter how effective a personalised online experience is, there is one aspect of traditional in-store selling that many digital platforms still struggle to compete with: real people.

Personalised online interactions are effective, yet they often lack the warmth of genuine human connections that enhance brand loyalty and customer satisfaction.

Button Text
A group of people that are looking at a phone.

Popular posts

Read the case study
Read the article
Download the paper