The CX KPIs everyone displays and nobody can read.

AHT, NPS, FCR, CSAT circulate everywhere. Their display gives an illusion of mastery. Their reading, however, almost always remains lazy.

A CX KPI says nothing on its own. It becomes useful when crossed with real customer behaviour, contact reason, and resolution cost.

AHT does not measure quality

AHT measures a cadence. It measures neither accuracy, nor clarity, nor customer effort. An AHT that falls might signal improved flow. It might also signal advisors cutting conversations short, partial answers, or treatment shifted to another channel.

The mechanism is simple. The more you optimise average time, the more you sometimes push the team to close fast. The score improves, but the request is not necessarily resolved. The customer calls back, transfers, or leaves with an incomplete answer. The right signal is therefore not raw speed. It is speed accompanied by stable resolution and reduced effort.

Ask your vendor: "Can we display AHT by query type, channel, and team, with the 7-day reiteration rate?" Require a contractual definition of what is measured: conversation time alone, post-call handling time, or total end-to-end time. Classic trap: AHT falling while re-opens rise. Symptom: queues seem more fluid, but cases keep cycling back.

NPS says nothing without verbatims and journey data

NPS captures a recommendation intention. It does not say why the customer recommends, nor what they still tolerate. A stable score can hide a slow journey degradation. A low score can coexist with good operational resolution if expectations, pricing, or policy irritated elsewhere.

The score alone mixes different causes. A rating may reflect a recent interaction, an accumulation of irritants, or a general brand opinion. Without verbatims, you do not know whether the customer is commenting on the advisor, the product, the delay, or a prior friction point. Without collection timing, you do not know whether the response measures hot emotion or an already consolidated perception.

Ask: "At exactly what journey stage was the score collected, and which verbatim is attached to each segment?" Require breakdown of promoters, passives, and detractors by collection channel and post-interaction delay. Classic trap: NPS sent too early after a positive moment, then the journey degrades without any signal. The score is constructed at the wrong moment.

FCR is only credible if measured from the customer's side

First Contact Resolution as reported by the system and FCR as experienced by the customer often differ. The agent closes a case. The customer calls back three days later on the same subject. If the second contact is coded as a new query, FCR remains artificially high.

The mechanism is subtle. A customer who re-contacts without explicitly referencing the previous interaction is not counted as a failure by most systems. Yet from the customer's perspective, the problem was not resolved. A serious FCR must be built from the customer side: same customer, same intent, within 7 days.

Ask: "How is FCR calculated — from agent closure or from customer recurrence?" Require: "Can we see the rate of same-intent contacts within 7 days, by channel and by query type?" Verify: the FCR definition must be consistent across all measurement points. Classic symptom: a high FCR on voice alongside a high chat recurrence rate — a signal that contact shifts channel rather than resolving.

CSAT measures a moment, not an experience

CSAT captures satisfaction at a precise moment, generally immediately after an interaction. It does not measure the overall experience of the journey, nor the cumulative quality of repeated contacts, nor the long-term evolution of the relationship.

A high CSAT on a given interaction can coexist with a degraded experience across the journey. The customer may have been satisfied with the advisor's response while finding the IVR, the delay, and the transfer process unacceptable. CSAT captures only the part where a human intervened — and misses everything that preceded and followed.

Ask: "What does CSAT measure exactly: the entire contact, the last interaction, or the agent's response?" Require: "Can we see CSAT by journey stage, not just by channel?" Verify: is there a mechanism for cross-referencing CSAT with resolution rate and re-contact rate? Classic trap: a high CSAT on individual contacts while overall NPS and re-contact rate degrade. The symptom is individual satisfaction masking repeated journey failure.