One of the largest global retailers was looking to unlock the value of data and machine learning models to better inform decisions across the brand’s customer experience (CX). After years of explosive growth, the brand needed to focus efforts on creating frictionless experiences for customers across both digital and in-store channels. Technology investments had to be better utilized focusing on building newer CX capabilities while driving improvements in existing digital marketing processes.
To drive CX improvements in the post-purchase journey, the brand had to understand key customer pain points to reduce friction. A new AI-based VoC (Voice of Customer) solution had to be developed to extract customer pain points from over 75 million text mentions across chat, IVR (interactive voice response) and several other sources. Building cloud-driven pipelines that could support this extensive scale was imperative.
Simultaneously, the marketing team was looking to optimize their MXPM (Marketing Experimentation and Measurement pipeline). This would allow them to drive marketing effectiveness measurement at scale with a lower cost footprint, while continuing to stay compliant with the latest customer privacy guidelines. Existing processes required manual interventions, and ran for over 20 hours, often taking days and affecting service level agreements. This impacted downstream marketing operations processes and covered only about 60% of all customer subsegments. Scaling this to all customers subsegments would be prohibitive, operationally and financially.
Merkle deployed two separate teams dedicated to the VoC solution and the MXPM projects, skilled with the right mix of consultants, architects, engineers, data scientists, and analysts. An iterative approach was used, fostering close collaboration with the client brand’s stakeholders to assess and uncover needs, while designing, building, and rigorously testing until the final release was achieved. A Jira sprint board was created for the client’s marketing team to track the progress of implementation. Leveraging the logging and monitoring capabilities of Google Cloud Platform, they prioritized scalability, distribution, and maintainability, while also developing customer monitoring functionalities with Google Looker Studio-based BigQuery Usage Monitoring Tool.
For the CX team, Merkle built an end-to-end VoC platform with auto scale capabilities that runs text data through complex deep learning models chained together to extract themes and attach sentiment scores. The process connects several customer data sources, classifying pain points across +190 different taxonomy nodes. The process enabled daily updates to metrics dashboards created for consumption by various levels of stakeholders, analysts to senior leadership.
Google Cloud Platform (GCP) components involved: Google Composer, Google Logger Service, Google BigQuery, Google Cloud Storage.
MXPM pipeline was optimized for over 16 months improving end-to-end processing time by 65% from 20 hours down to 7 hours covering 100% of the segments (previously on 60% were covered). 15% segments now refreshing daily from weekly prior to the optimization. Estimated total infrastructure cost savings of over $230k on a like to like basis.
GCP components involved: Google Dataproc, Google Composer, Google Logging, Google BigQuery, Google Storage, Google Looker Studio.
This website uses cookies.