MEASUREMENT
How do you measure customer experience so it predicts revenue rather than reporting the past?
Measure at the level of the individual interaction, as it happens, and join it to what the customer does next. Score resolution, effort and experience on every case, read sentiment live from language and behaviour, and track the revenue that follows. A quarterly survey tells you what already happened. Live interaction data tells you what is about to happen.
Answered by Alex Mead. The dated record behind this answer.
How it is actually done
1. Move the unit of measurement from the quarter to the interaction
A survey sent weeks later measures memory, not experience. Ask at the case, about the case, at the moment the customer has an opinion. At D360 Bank the app let customers rate every closed case on Resolution, Effort and Experience, whenever they chose, with a toggle asking us to follow it up. Those three dimensions were live in a licensed bank's shipped product in April 2021, and they answer three different questions: did you get what you came for, how hard was it, and how did it feel.
2. Read sentiment from behaviour rather than from a form
Most customers never fill anything in, so a measurement system built on forms is reading a self selecting minority. Language, tone, wording, hesitation, what the customer looked at before they made contact and what they did after: all of that is signal, and it moves second by second rather than quarter by quarter. That is the difference between a score that describes and a score that warns.
3. Cover all of the work, not a sample of it
A quality function listening to twenty calls per agent per month is reading roughly one per cent of what happened. Every conclusion it reaches is an inference from a sample chosen by whoever pulled it. GenAI reads all of it. This is the first point in my career at which measurement can actually cover the whole operation, and it changes what the number is worth.
4. Join it to money, or it is decoration
Every interaction score has to be joined to what the customer did next: bought again, upgraded, stayed, referred, or went quiet. That join is the only thing that turns experience measurement into a forecast, and it is the part almost nobody builds. The most recent proof I have is a three month GenAI engagement completed in July 2026, in luxury travel rather than cruise, for a client under NDA: 35 to 40% more enquiries progressed to quotation, 25 to 28% higher quote to booking conversion, and 12 to 15% higher revenue per guest. Those are commercial numbers produced by measuring the interaction properly and acting on it.
5. Publish it internally, and let customers reopen the case
A score nobody can see changes nothing. Share the numbers openly inside the business, down to team level, and give the customer the right to reopen a case if they do not like the outcome. When a closed case can be reopened by the person it belonged to, the closure rate becomes an honest number for the first time.
What most companies get wrong about this
- They run one survey a quarter, on a sample, and call it measurement. By the time the score moves, the customers who moved it have already gone somewhere else.
- They measure satisfaction with the survey instead of the outcome. A customer can rate an interaction five stars and still not have their problem solved, and both facts will be true in the same report.
- They use three smiley faces. It is cheap to deploy and it produces a number with no diagnostic value, so the business collects it for years and never learns anything from it.
- They never join the score to the till, so the customer function argues with the finance function about whether any of it matters, every year, forever.
- They report the average. The average conceals the customers worth the most, and those are the only ones whose experience the board actually needs to hear about.