Customer Experience data as a measurable, monetisable retention signal.
Re-bookings drive the exhibition business: keeping an existing exhibitor is far cheaper than winning a new one. Yet organisers judged satisfaction by gut feel and a single headline score — unable to say which aspects of the customer experience actually drove an exhibitor to return.
Surveys arrived in dozens of incompatible formats across hundreds of events — different question types, scales and structures — making consistent measurement impossible. Nobody could cleanly answer which experience factors predicted retention, let alone quantify what improving them was worth.
Without that translation — from experience to behaviour to revenue — customer experience programmes lacked the financial grounding to justify investment. At-risk exhibitors slipped away undetected, and the cost of churn remained invisible on the balance sheet.
Survey data across 85 events arrived in incompatible formats, question types and scales. Before any model could run, it had to become a single unified read — and the ground-truth outcome (actual retention) had to be computed from sales transactions, not stated intent.
Raw satisfaction scores didn't move with retention. The predictive signal only emerged in year-on-year changes — specifically in paired importance-vs-satisfaction questions. Finding it required interrogating the data carefully rather than fitting a model to the obvious features.
Knowing a score predicts retention is useful. Knowing what a 1% retention improvement is worth in pounds is what gets a budget approved. That financial model had to be built separately — from sales transaction data — and connected back to the prediction engine.
A Customer Experience prediction model
Survey, sales and event data unified in a data lake; retention computed from sales transactions as the ground truth. An automation engine processes ten standard question types and scores free-form answers at scale — 85 events, ~15,000 questions — with no manual coding per survey. A model trained on year-on-year score changes predicts retention direction with ~80% accuracy.
Quantifying the value of retention
Analysis of 56 EMEA events showed that retained customers spend 125% more than new ones (£15.7k vs £7k), purchase more product, and carry larger baskets. From this, the "power of 1%" was modelled: a 1% improvement in retention generates over 1.3% in additional revenue — self-funding Customer Experience investment from the revenue it protects.
Pinpointing the levers to pull
The model doesn't just score — it identifies the specific Customer Experience drivers impacting retention per event, shows current vs target values, and quantifies the score improvement each change would deliver. Commercial teams get an action plan alongside the forecast, and a per-account view of churn risk.
End-to-end data science and engineering — from problem framing and data architecture through survey automation, financial modelling and machine learning to a working analytical product, validated on live events with a global exhibition partner.
Retained customers were found to spend 125% more than new ones — and the financial model showed that a 1% improvement in retention generates over 1.3% in additional revenue. Across the EMEA portfolio, that equates to nearly £3m for every 1% retention gain. On a single event — £6m revenue, ~160 customers — a 1% improvement adds ~£100k in annual revenue and ~£140k in customer lifetime value, typically enough to self-fund the CX programme entirely.