AI-Powered Customer Insights: Turning Service Data into Retention Strategy
Customers are not reliable narrators of their own behavior. What they say, what they feel, and what they ultimately do often diverge—especially after purchase, when experiences unfold largely outside traditional digital analytics. For e-commerce businesses, this creates a critical blind spot: the moments that most directly shape trust, repeat purchase, and long-term retention.
At Farfetch, customer service interactions held the missing signal. Embedded in millions of emails, chats, and contacts was a real-time record of customer sentiment, anxiety, and behavioral intent. The challenge was not access to data—but transforming fragmented service signals into actionable insight that leaders could use to make confident, customer-centric decisions at scale.
The Customer Experience Index
Nominated for the AIconics AI Innovator of the Year Award in 2022
Business Challenge:
As the company was about to IPO in 2018, it faced several challenges in the customer experience. One challenge was that critical parts of the post-purchase journey were completely opaque. There was no data or feedback to support decision-making for or to measure impact on the post-purchase customer experience. The purchase journey can be measured with onsite metrics, but e-commerce post-purchase largely occurs off-screen. While the onsite journey is absolutely critical to customer acquisition and conversion, post-purchase is where the customer promise is fulfilled, directly impacting customer retention. Riding blind at this critical phase was proving to be costly in more ways than one.
The Opportunity:
Fortunately, data existed that could shed light on post-purchase, it just hadn’t been tapped. My team of expert data scientists and analysts were able to change that. Starting manually to build the concept for a post-purchase journey metric, we dug into Customer Service data to find patterns that aligned to volumes of contacts along critical customer journey moments, then read hundreds of emails to evaluate the general sentiment and anxiety levels of each of those moments. Even in it’s early manual phases, we were able to identify critical friction points and root causes to surface highly actionable insights for strategic decision making in the post-purchase journey. The reports were instantly adopted by company executives to invest in customer self-service and customer service Robotic Process Automation initiatives. This early research served as the foundation of the Customer Experience Index.
Data was sporadic and hard to access. Sources were spread across the organization. Decisions were made without customer input. CX suffered as a result.
Important customer feedback knowledge was held within Customer Service teams as a compilation of individual experiences, making it hard to extract patterns or validate.
Customer verbatims were researched manually to find patterns and root causes of negative customer sentiment. Human-power was limited and reports rolled out slowly.
Machine Learning and Sentiment at Scale:
In partnership with the Marketing department and an external software partner, we were able to scale this research and further develop use cases that would continue to shed light not only on the post purchase journey, but the total customer lifecycle. Leveraging AI, the team trained the software to read customer sentiment in the thousands of weekly emails as well as comments from customer surveys - offering real time insights into the holistic customer experience.
Expanded global view.
We were able to pull data from across global regions, reading feedback in 12 customer languages, allowing us to identify regional customer patterns and compare insights to reveal cultural nuances.
Surfaced actionable insights.
We taught the machine to read customer sentiment and identify patterns that led to root cause identification. These root causes enabled business action for continuous improvement and transformative business initiatives.
Increased pace of change.
It expanded human workforce capacity, capabilities in pattern recognition. We could produce more reports more quickly, to respond to customer expectations and behavioral changes.
12 Customer feedback languages monitored for global insights.
1000s of Customer interactions monitored for sentiment each week.
100s of Business cases supported for customer-centric strategies & business initiatives.
Development of the “Customer Experience Index (CXI)”
Leveraging emerging technologies, we continued to evolve metrics that would support understanding of the customer’s experience and behaviors. The CXI was developed as a way to better understand the impacts of our business’ performance on the customer behavioral response (conversion, cancellations, repeat purchasing and retention). By linking service KPIs to customer sentiment and contact-volume data, the CXI made it possible to identify which moments in the journey could make—or break—long-term customer relationships, enabling more confident decisions that directly drove repurchase and retention.
This research and metric development led to the nomination of my direct report, Senior Head of Voice of the Customer, for the AIconics AI Innovator of the Year award in 2022 for he and the team’s brilliant and ground-breaking work.
Customer Experience Index (CXI) Results
15% reduction in cost to serve as a result of improved operational efficiency and mitigated negative customer friction.
Meaningful improvement in the company’s Trustpilot score through targeted CX initiatives guided by the CXI metric.