The one number that kept the biggest client
outcome retained the client worth 50%+ of revenue
Problem
Hospitality operators drown in feedback signals: review scores, survey ratings, complaint cases, free-text comments, all pointing in slightly different directions. Executives wanted one trustworthy answer to “how are we actually doing?”, and the industry default, NPS, was losing their faith. For a feedback analytics company, that gap sat directly under the relationship with its largest client, worth more than half of revenue: the client was renegotiating its contract, and a holistic metric they could actually trust was one of the stipulated initiatives on the table (replacing their manual reporting was another; that story is case study D).
What I built
An industry-first composite metric fusing three signal families into a single 1–5 score: AI sentiment analysis of unstructured guest feedback, key touchpoint ratings (food, drink, speed of service, value), and case-resolution data. It pulls from every stage of the customer journey, and it carries a health score alongside the number itself, so an operator can see not just the score but how much evidence stands behind it. Built in Python within Microsoft Fabric, published to Cosmos DB, and delivered through the embedded dashboards clients already lived in.
The engineering was the smaller half. The metric worked because of the years I had already spent on the client-facing side of the business, sitting with the operators who would use it, hearing which numbers they trusted, which they ignored, and what they actually needed to defend in a board meeting. The design questions (what goes in, how it’s weighted, what makes a score credible enough to benchmark against) were answered from those conversations, not from a spec.
Techniques that made it work
- Discovery before design: the metric encodes what operators said they could act on, not what the data happened to offer
- AI where it earns its place: sentiment analysis over free text as one ingredient in a governed composite, not a black box asked to be the answer
- Shipped into the surface clients already used, with daily refreshes, rather than another report to open
Outcome
The metric was instrumental in retaining the company’s largest client, accounting for over half of revenue, and attracted multiple new clients. It was adopted across the industry as a benchmarking standard: 60 of 171 active clients now run it as their main metric, and because those tend to be the highest-profile estates, they account for roughly 60% of all feedback volume flowing through the platform, month after month. It remains the clearest example of what I sell: the distance from a business conversation to a production system, walked end to end.