Spark embraces deep customer insights and intelligent automation

Who: Spark New Zealand

What: Creation of a customer experience framework, supported by intelligent automation to streamline the process and remove friction in customer journeys

How: Using TM Forum’s Application and Information Frameworks, part of the Open Digital Framework

Results: An increase of around 40% in the uptake of offers which in turn resulted in savings and improvement in customer NPS

One of the typical motivations for digital transformation in the telecoms world is becoming more customer-centric. Telcos need to know who their customers are, what they want and how to give it to them quickly. That’s a tall order when the only things you know about your customer are their line number and where to send the bill. This was precisely the case for incumbent operator Spark (formerly Telecom New Zealand) 10 years ago, when it was offering the usual menu of legacy telco offerings and an inflexible green-screen back-end system. “The system understood customers as line numbers,” says Kallol Dutta, Tribe Lead, Spark New Zealand. “That was how we saw the customer.” Spark initiated a transformation program to move away from that impersonal system to a flexible, personalized system that would see customers as people with specific needs. But the objective wasn’t simply to have richer customer data profiles; it was to create a customer experience framework that could track the customer journey in detail, with intelligent automation underneath it to streamline the process and make it as frictionless as possible – even to the point of anticipating pain points before they happen.

Transformational groundwork

Spark laid the groundwork with a three-stage transformation program, starting with the IT, although Dutta says this was as much about transforming company culture as it was about technology.

“Culturally, we changed so the teams are really able to take risks and become more outcome focused,” Dutta explains.

The result was the development of agile, cross-functional teams capable of taking action at pace, using real-time insights, to make the business highly responsive and adaptive, Dutta adds. “We developed teams so that everyone was aligned to the wider objective of moving from legacy and becoming digitally aligned,” he says. “We also developed a lot of internal system integration capabilities, which then brought in other partners as required.” The second stage was to become more a data-driven company with a data architecture that was grounded in real-world business cases. “We worked from the business case down – we did it end to end, from PoC to operational scale,” Dutta says. “The idea was to get scale on just a few things, but to do them well.” Using open source solutions, Spark built a data architecture based on TM Forum’s Information Framework (SID), part of the Open Digital Framework, to deliver big data capabilities that enable teams to use real-time/near-time data and insights to see how their products and journeys are performing for customers.

“We are essentially using the Information Framework as a way to develop the common data layer, which is represented in a common enterprise catalog,” Dutta says. “This enterprise catalog is a really important thing, because it helps you to discover data elements so that you don't duplicate work. It also provides data governance which is very important.” Dutta adds that Spark is also using TM Forum’s Application Framework (TAM) in conjunction with the Information Framework to map the data to applications. This approach, he notes, lends itself well to Spark’s cloud migration strategy.

“We will have a multi-vendor, multi partner strategy,” he explains. “We have Azure and Snowflake as our key partners, but that doesn’t confine us because our architecture is generic – we are building an open architecture, which TM Forum recommends, so that we can integrate any architecture.”

Encouraging deeper customer relationships

One of the specific things Spark has done with its transformed agile data architecture is develop a customer experience framework to identify and track how customers are going through their journeys. So in a metaphorical sense, Spark is digitally walking alongside their customers on these journeys. The core concept is that from the customer’s viewpoint, their experience with Spark is not a one-off interaction but a more complex relationship. Spark’s so-called JUCCI (Join, Use, Care, Change, Involve) framework illustrates the complexity of that relationship. Customers join services, use them, pay for them, change them (by adding new services, cancelling them, or upgrading to bigger service bundles) and give feedback. Obviously this is not a one-way relationship – Spark is not just signing up customers, but offering customer care and technical support, managing whatever modifications customers want, notifying them of offers or problems that may affect them and offering rewards. The JUCCI framework enables Spark to categorize customers into journeys and then implement big data analytics to stitch different journeys together to understand the holistic journey the customer is traversing. The result is a “journey catalog” that enables Spark to understand its customer experience performance in a much more holistic way across multiple dimensions by analyzing the various data points in the data lake.

“We have used data to understand customers’ profiles and segmentation,” Dutta says. “Based on various customer parameters we are able to cluster the customers in groups. Then we can understand their cluster, their journeys with our journey framework and then understand what they really like or don’t like in their journeys.”

Dutta admits this is not easy to do in a complex telco environment where experiences can vary widely. “For example, a prepaid top-up journey is a few seconds, whereas a fiber installation journey can be six months,” he says. “But by mapping out those two types of journeys and having the ability to track the two journeys together, we can understand not only the journey of the customer, but also the fallouts, so we can then evaluate whatever friction is involved and remove it.”

Intelligent automation is key

Key to making all of this work is intelligent automation, which is also made possible by Spark’s digital transformation initiative. Spark has implemented robotic process automation (RPA) to handle repeatable tasks and functions that – when handled manually – tend to be slow and error-prone.

Spark is implementing automation in three ways – in addition to the actual process automation implicit in RPA, there’s also cognitive automation (using machine learning) and conversational artificial intelligence (AI) for things like chatbots. Dutta is particularly enthusiastic about cognitive AI, which is already delivering some impressive results. “With that one, we are using machine learning to not only understand customers, but also target our campaigns, especially the retention campaigns,” he says. “We can predict which customers are likely to leave, and then target them with offers they will likely take so that they don't leave.”

According to Dutta it’s working like a charm, with initial results showing around 40% higher uptake of offers compared to before machine learning was added to the mix. “That essentially resulted in not only savings, but also improvement in customer NPS, because we could identify the customer’s problem and solve it.”

Spark currently has 100 bots handling different functions, from field services and processing fiber orders to IT operations and proactively troubleshooting fiber connectivity problems. Dutta says the bots have helped Spark deliver the automation pillar of its three -pronged strategy in 2018 (the other two being simplification and digitization) which partially contributed to the telco’s 4.3% drop in OpEx for its annual results released June 30. Moving forward, as AI becomes a more prominent component of Spark’s operations, Dutta says it is closely following the ongoing work in TM Forum’s AI for IT & Network Operations (AIOps) Catalyst. “We are on the boundary of creating a different AI model for Ops, because the DevOps model – though it is relatively mature – is not directly usable in AI,” Dutta says. “We need to create a new AIOps model, and the level of collaboration TM Forum has with various parties who are developing AIOps is really impressive.”