Customer Experience Survey: Metrics, Journey Questions and a Touchpoint Template

CES. Dixon and colleagues report a study of more than 75,000 people dealing with contact-centre staff or self-service channels. They write that CES "outperformed both" CSAT and NPS "in customer service interactions" as a predictor of loyalty, and they note NPS "has been shown to be a powerful gauge at the company level" (HBR, 2010). Read that as a claim about service contacts, made by the people who built the metric. On their scale a high score is bad. Some tools ask an agree-disagree version instead ("The company made it easy..."), where high is good; we could not open a primary source for that version, so check which one your tool uses before you compare numbers.

NPS. The definition, its margin of error and its critics are covered in our NPS methodology guide. The short version: it is a relationship measure, and a single score hides a wide interval.

CSAT. There is no canonical wording, so define yours and keep it fixed. If you want a formal satisfaction index, the American Customer Satisfaction Index, begun in 1994 at the University of Michigan, combines three 1-to-10 questions: overall satisfaction, how the experience compared with expectations, and how close it came to an ideal (summary on Wikipedia).

This mapping is our suggestion, not a standard:

Whatever you pick, ask one open question after the score. The number says how many people struggled; the open answer says where.

Nielsen Norman Group defines a journey map as "a visualization of the process that a person goes through in order to accomplish a goal", built from an actor, a scenario with expectations, journey phases, actions, mindsets and emotions, and opportunities. It adds that actions in the map should be "rooted in data" (NN/g, Sarah Gibbons, 2018).

Run interviews first, to learn the phases in customers' own words. A survey then sizes what the interviews found. Questions that feed each part of the map:

Report each phase on the people who went through it, not on everyone. A low emotion score at "set up" from 60 people deserves a different weight from one at "ordered" from 600.

Send soon after the event. Ask about one recent event while it is fresh, and name it: "your delivery on 3 October", not "your recent experience".

Keep it short. One score, one resolution check, one open question, and only then anything else. Add a question only if someone will act on its answer.

Pass facts, do not ask them. Channel, agent, order value band and store already sit in your systems. Attach them to the response instead of asking the customer.

Do not survey the same person at every touch. Set a rule, such as no more than one survey per customer in a period you choose. Platforms implement this differently; our customer satisfaction platform guide lists what to check.

Watch the bases. A branch with 25 answers a month can swing from month to month on noise alone. Show the base beside every score and combine months for small units.

A template for a support contact. Adapt the wording for a delivery or a visit, but keep the order: score first, then resolution, then why.

Translate the scale labels with the same care as the questions; our Likert scale guide covers label wording in Hindi.

A voice-of-customer (VoC) programme puts solicited feedback (the surveys above) next to unsolicited feedback: complaints, call notes, app-store reviews and social posts. The two answer different questions. Surveys give you rates on comparable bases. Unsolicited feedback shows problems you did not think to ask about, but it comes from whoever chose to speak.

Three conditions make a VoC programme more likely to change anything:

• An owner for every touchpoint, named in the journey map, who sees that touchpoint's scores and comments each week.
• One coding frame across sources, so "refund delay" in a survey comment and in a call note count as the same thing. Our open-ended analysis guide shows how to build and check one.
• A follow-up loop for individual customers who asked to be contacted, tracked until someone has called them back.

Touchpoint surveys go to your own customers through your own systems, so they are not a panel job. The study our panel fits is the journey survey: category users describing how they bought, with your brand and its competitors in the same sample.

The phone-buying journey above, with 500 recent online buyers, is one Pro bundle: 500 × ₹120 = ₹60,000 (pricing). Our panel methodology says the panel spans Tier 1, 2 and 3 cities and runs through an app, so it reaches people with smartphones. That limit matters less for an online-buying journey than for one that happens offline.