Skip to main content

Enabling real time data insights to power a responsive customer experience

Proliferation of digital channels like company websites, mobile apps, social handles and virtual assistants are leading to a sensational shift in interaction paradigm both for organizations and customers. Organizations need to stay ahead of the changing interaction ecosystem by utilizing the channel events generated by each customer touchpoint. Digital channel events are a treasure trove of customer insights and provides an opportunity for organizations to personalize their services. The challenge is to capture the data, process and generate insights in a real time, and then make sure to act on it immediately.

Supporting this goal for our client, a large telecom service provider, Deloitte augmented their data architecture to generate powerful business insights about customer digital channel behaviour by harnessing the power of Google Cloud and leveraging Deloitte’s extensive experience in designing real time digital event stores.

Our client’s vision is to deliver a high-performance personalized service experience to all its customers. To deliver such a service, it is critical to have an effective decision-making engine based on real time customer event data across all channels. The as-is data architecture for capturing digital events was primarily batch oriented (captured once daily or sometimes weekly) and limited to a only subset of available channels. As a result, it was very time consuming and sometimes not possible to generate relevant insights at a customer level in real time, or even close to real time. Furthermore, as not all channels were captured, there was limited visibility into the full picture of true user behaviour. Potential value from the digital events was not being fully maximised, leading to missed sales conversions, protection of churn and inconsistent customer experience across their channels.

What happened next

Deloitte was brought in as a consulting partner to undertake the following –

  1. devise a state-of-the-art digital event architecture that can capture digital events across channels in real time, and
  2. design a data framework that facilitates data availability for analytical and predictive models as well as feed insights to downstream decisioning engines. 

Deloitte approached the solution by first understanding the customer’s as-is architecture landscape, defining key business use cases, and planning the to-be solution architecture. Given the customer’s existing usage of Google Cloud infrastructure and Deloitte’s experience with the platform, the new digital event architecture and model was designed leveraging GCP with BigQuery – Google’s data warehouse solution – at its core. As a Google Cloud Global Industry Solution Partner, Deloitte was well positioned to leverage its expertise in Google Cloud Platform (GCP) products and services to implement a highly performant, reliable, and cost-effective solution built with GCP.

To illustrate the limitations of the previous architecture while also walking through a potential use case, let’s consider the following example; Sarah is a customer of the telecom company and uses the virtual assistant on the company website to learn more about early termination fees. Her intent to cancel her existing plan in the virtual assistant gets picked up and within 30 minutes, Sarah is presented with a personalised communication with a better offer to compare and consider against her current plan. With the as-is architecture being batch oriented, processing of virtual assistant data is typically carried out once daily, thereby a missed opportunity to engage with Sarah within the narrow timeframe before she makes the final decision of cancellation.

The new data architecture was designed to effectively solve these as-is challenges as well as make future proof the platform against new and emerging channels and use case requirements.

The key features of the proposed architecture were as follows –

Inclusion of all digital channels as data sources, such as chatbots, virtual assistants, mobile apps, website events, call centre records, and other enterprise applications. Cross channel behaviour analysis is a key success criteria for personalised services and hence requires data input from all channels.

Layered data architecture – The entire data pipeline is divided into multiple layers. The landing zone provides the storage of all digital events across channels in their original raw form. This layer forms the original source of truth. The curated zone is a filtered and restructured version of the events data which are relevant for business decisions. The modelling zone is based on a single common data model encompassing all channels to enable richer multi-channel customer insights and behaviours. The modelling zone contains derived attributes based on historical digital behaviour data which to augments their customer 360 application.

Google cloud components –

  • Pub/Sub and Dataflow are utilised for real time streaming and ingestion of data in the landing zone.
  • BigQuery is the preferred choice for storage across all the layers. BigQuery serves as a serverless analytics engine capable of automatically handling scale, size and performance. Queries of data with sizes in the TBs can be processed in a few seconds. Further, BigQuery provides built-in machine learning capabilities, generating intelligent insights within the existing cloud storage and query environment. Finally, consumption of data by downstream consumers from BigQuery is seamless with flexible connectivity paths.
  • Security controls – Google cloud provides out of the box security controls which helps in securing the data as well as help to monitor the data pipeline and access/consumption.

After defining the to-be architecture, Deloitte devised a common data model for the analytical/modelling layer. Due to the varied nature of the digital channels the data structure and features available were quite different. At a high level, the channel data was split into two categories, ones with unique customer identifiers, and ones without. While customer level granularity was necessary for our client to leverage the events data for personalisation, the intent was to present all events across all channels in the data model to enable a wide range of insight use cases.

Deloitte proposed two data domains to address the varied nature of the data. The first data domain stores multi-channel event data for all channels where customer identifiers are available. The second data domain stores data where the customer is usually anonymous, such as from over-the-counter cash sales. This data could still be used on a more general level to identify consumer patterns, even though the data could not be linked to individual subscribers.

The wins

  • Built a foundation towards our client’s vision of capturing real time digital events and enhancing the experience of their customers
  • Expedited build work by using out of the box features of Google Cloud services, ensuring swift delivery and saving time and money for our client
  • The common data model facilitated data capture from all existing digital streaming channels and maintained flexibility to accommodate new channels as they became available in the future
  • The digital event store became a central data source for downstream applications, maintaining accuracy and generating key insights around the digital behaviour of customers.

Author : Sid Bhattacharya - Manager, Consulting Strategy AI & Transformation

Contributor : Vincent Chen – Consultant, Consulting Strategy AI & Transformation