Measuring experience without drowning in surveys
The usual failure of experience measurement is not too little data but too much of one kind. A business discovers survey tools, attaches a rating request to every interaction, watches the scores, and six months later cannot say what changed because the scores moved, or whether anything was done because they did. The customers, meanwhile, have stopped answering.
The alternative starts from decisions rather than instruments. For each measure, the question is what the business would do differently if it moved. A relationship survey score that nobody owns and that no decision depends on is noise with a dashboard. A churn figureShare of customers at the start of a period who left during it. read by the person who sets the onboarding plan is a measure. The measurement plan framework formalises this: objectives, then the KPIA metric tied directly to an objective. for each, then the data needed, then the decisions attached. Applied to experience, it usually produces a short list: one behavioural retention measure, one or two perception measures with a clear trigger, and a handful of operational measures from support. The customer experience hub sets out why those measures matter; this section covers how to collect and read them.
Perception, behaviour, operations
The three kinds of evidence answer different questions and have different weaknesses, which is why they are needed together.
| Kind | What it captures | Examples | Strength | Weakness |
|---|---|---|---|---|
| Perception | What customers say they think and feel | Relationship and transactional surveys, open-text comments, reviews | Explains why; catches problems before behaviour changes | Sampled, biased towards the vocal, slow to move |
| Behaviour | What customers actually do | Churn, retention and repeat rates, active use, cohort curves | Hard to argue with; ties to revenue | Lags the cause; silent about reasons |
| Operations | How the service performs | Response and resolution times, contact volume by cause, first-contact resolution; see customer service metrics | Fast, controllable, owned by a team | Measures the business, not the customer |
The connections are the useful part. Perception explains behaviour: a fall in a post-onboarding score this month predicts the churn figure in three. Operations explain perception: slower response times show up in transactional survey comments before they show up in the relationship score. Behaviour validates the other two: if scores rise and churn does not fall, the score is measuring something the business does not need. The ESOMAR professional standards for market research are the reference for the perception side, in particular for consent, honesty with respondents and the handling of their data, and the voice of the customer guide applies them.
The core guides
- Voice of the customer: NPS, CSAT and CES covers listening programmes rather than isolated surveys, what each of the three common survey measures captures, relationship against transactional surveys, timing, sampling and fatigue, reading results and open text, and closing the loop with customers and teams.
- Churn and retention metrics defines customer and revenue churn, retention, repeat and active-customer rates, introduces cohort views, lists the leading indicators that show churn before it happens, and builds a retention KPI tree.
Two neighbours live elsewhere. Building cohort tables, including the mechanics of grouping customers by acquisition period and following them, is the subject of the cohort analysis tutorial in the growth marketing root. The general method of decomposing an outcome into drivers is on the KPI trees guide in the analytics root, and the KPI tree framework page summarises it; the churn guide here applies it to retention.
Further guides
Further guides cover the operational evidence from support, the comments that do not reach a help desk, and tests of whether a perception measure deserves its place:
- Customer service metrics and how to read them — which service metrics to track and how to read them: first reply, resolution, repeat contacts, contact rateSupport contacts in a period divided by orders, active customers or bookings in the same period, showing how often customers need help., backlog age and satisfaction by reason.
- Tagging support contacts by reason — a tutorial for building a contact reason list: drafting it from real contacts, separating reason from root cause, tagging at close, auditing and reporting monthly.
- Recording what customers tell your staff — a tutorial for the comments that do not reach a help desk: what is worth writing down at the counter or on the phone, a note in the customer's words, a reason from the shared list, complaints passed to their owner when they are made, and a shared read with the people who own the causes.
- Testing whether survey scores predict churn — a tutorial for checking in your own data whether satisfaction or effort scores warn of churn: joining answers to later behaviour, comparing groups and reading small numbers well.
A customer experience measurement plan and dashboard
A measurement plan for experience fits on one page: the retention outcome at the top, the perception and operational measures that explain it beneath, and for each a definition, an owner, a cadence and the decision it informs. The definitions deserve more care than they usually get. "Active customer" and "churned" have to be defined once, in writing, with the edge cases settled (a paused subscription, a customer who bought once a year, a refund), because every later argument about whether retention improved is really an argument about definitions.
The dashboard follows the plan rather than the other way round. A useful experience dashboard has a small number of panels in a fixed order: the retention outcome and its trend, the cohort view beneath it, the perception measures with their trigger points labelled, the operational measures from support, and a panel of recent open-text themes. The panel most dashboards lack is the last one, and it is the one that turns the others into action. Experience changes should be tested like anything else: an onboarding change rolled out to a random half of new customers is an A/B TestControlled comparison of two variants. on retention, and the dashboard should be able to show the two groups side by side. Customer lifetime value (LTVTotal value a customer brings over time.) belongs on the dashboard as the outcome retention feeds, with the maths left to the growth economics guides.
A broadband provider builds a one-page plan
Larkline, a regional broadband provider with about 38,000 customers, has four survey tools, a weekly report of eleven scores, and monthly churn of roughly 1.8% that nobody can explain. The team writes a one-page plan: monthly churn by tenure cohort as the outcome; a single post-installation survey and a quarterly relationship survey as perception; time to first engineer visit, repeat-fault rate and support response time as operations. Three of the four survey tools are switched off. Within two quarters the post-installation survey shows that customers who wait more than five days for installation rate the experience far lower, and the cohort view shows those customers churning at nearly double the rate in their first six months. Installation scheduling becomes the retention project for the year, and churn among new customers falls to about 1.2% monthly.
Closing the loop
Measurement earns its cost only when something changes because of it. Closing the loop has two halves. The inner loop goes back to the customer: someone reads the low score or the complaint and responds, within a defined time, with a human reply and where possible a fix. This is also the best recovery mechanism a business has, because a customer who was asked and then heard is often more committed than one who never had cause to complain. The outer loop goes to the teams: the themes from open text, the operational causes behind the scores and the cohorts that are churning are reviewed on a fixed cadence by the people who own onboarding, product, support and marketing, with decisions recorded and followed up.
The failure mode is a monthly meeting that reads the numbers and ends. The fix is structural: each measure on the plan has a named owner, each review produces a short list of actions with dates, and the next review opens by checking them. The voice of the customer guide covers the inner and outer loops for survey data in detail; the service section covers the root-cause loop for complaints.
Common customer experience measurement problems
| Problem | Usual cause | Guide |
|---|---|---|
| Scores collected everywhere, nothing changes | No decision attached; no owner; no outer loop | This hub, above, and Voice of the customer |
| Response rates falling | Too many surveys; wrong timing; no visible result for the customer | Voice of the customer |
| Churn figure disputed every month | Active and churned never defined; different teams count differently | Churn and retention metrics |
| Retention looks fine until it suddenly is not | No cohort view; blended rate hides the newest customers | Churn and retention metrics, Cohort analysis |
| Survey score rises while churn does not fall | The score measures something unrelated; sample biased | Voice of the customer, Testing whether scores predict churn |
| Nobody can trace a retention change to a cause | No KPI tree; measures not connected | Churn and retention metrics, KPI trees |
| Dashboard has many panels and no story | Built from available data rather than from the plan | This hub, above |
| Service report is a row of averages | Metrics read alone, with no guardrail and no split by contact reason | Customer service metrics |
| Support contacts counted but their causes unknown | No contact reason list; reason and root cause mixed | Tagging support contacts by reason |
| Customers say things at the counter or on the phone that nobody records | Comments reach staff and stop there unless someone writes them down | Recording what customers tell your staff |
Read the churn guide first if the business cannot yet state its retention rate with confidence; read the voice of the customer guide first if it can, but cannot say why customers leave. When churn or a survey score has already moved and a review wants to know why, the customer experience troubleshooting section tests whether the number changed before the customers did, then reads the causes one at a time.
