CX Metrics That Lie (And What to Use Instead)
The metrics on your dashboard that are actively misleading you, and what to track instead
Most CX dashboards are lying. Not because the data is wrong, but because the metrics were designed to measure something other than what they are being used to indicate. NPS measures stated intent, not churn risk. CSAT measures interaction satisfaction, not product value. This guide names the lies and provides the replacements.
Most CX dashboards are lying. Not because the data is wrong, but because the metrics were designed to measure something other than what they are being used to indicate. NPS measures stated intent, not churn risk. CSAT measures interaction satisfaction, not product value. AHT measures speed, not quality. Resolution rate measures ticket closure, not problem resolution. Each metric is useful in a narrow context and actively misleading when used outside it. This guide names the lies and provides the replacements.
This is the contrarian companion to Part 5 of the CX Field Guide. Part 5 covers the metrics that matter. This guide covers the metrics that do not, why they do not, and what to track instead. The two guides are designed to be read together: Part 5 for the framework, this for the cleanup.
The core principle: a metric is not lying if you use it for what it measures. It is lying when you use it for what it does not measure. NPS is not a bad metric. NPS as your sole CX health indicator is a bad measurement system. The distinction matters.
The Metrics That Lie
Four metrics dominate most CX dashboards. Each one is useful in a specific context and misleading when used outside that context. The lying happens at the application layer, not the metric layer.
NPS
Measures stated intent to recommend, not actual behavior or churn risk
Useful context: Useful as a directional trend over time, misleading as a sole health indicator
CSAT
Measures satisfaction with a specific interaction, not whether the customer is achieving their outcome
Useful context: Useful for touchpoint quality, misleading as a proxy for relationship health
AHT
Measures speed, not quality. Faster is not better if the problem is not resolved
Useful context: Useful as a diagnostic for training needs, destructive as a performance target
Resolution Rate
Measures ticket closure, not problem resolution. Tickets can be closed without the problem being solved
Useful context: Useful only when controlled for repeat contacts within 7 days
NPS: The Score That Hides More Than It Reveals
NPS (Net Promoter Score) is the most used and most misused CX metric. The original Reichheld claim, published in Harvard Business Review in 2003, was that NPS is “the one number you need to grow.” Two decades of independent research have not supported that claim. The correlation between NPS and revenue growth varies substantially across industries, and several large-scale studies have found CES, repurchase intent, or composite measures to be more predictive.
NPS is not useless. It is useful as a directional tracking metric over time, broken down by customer segment. It is misleading when used as the sole indicator of CX health, because it measures stated intent, not actual behavior.
How NPS Lies
How to Use NPS Honestly
- • Track as a directional trend over time, not a single number
- • Break down by customer segment (enterprise vs. SMB, by industry, by tenure)
- • Send at a fixed cadence independent of product events
- • Always pair with behavioral signals (usage, engagement, support)
- • Never use as the sole CX health indicator
- • Never tie to individual agent or CSM incentives (Goodhart’s Law)
CSAT: Satisfaction Is Not Success
CSAT (Customer Satisfaction Score) measures how satisfied a customer was with a specific interaction. It is the right metric for touchpoint quality measurement: post-support-call, post-onboarding-session, post-feature-release. It is the wrong metric for relationship health, product value, or churn prediction.
The lie: “Our CSAT is 4.5 out of 5, so our customers are happy.” They may be happy with your support team. They may not be getting value from your product. High CSAT on support interactions is not correlated with retention. Customers can love your support team and hate your product. They will give your support agent a 5 and then not renew.
CSAT as Proxy for Health
- • Captures feelings about a single interaction
- • Influenced by agent likability, not product value
- • Inflated by recency bias (last interaction was positive)
- • Not correlated with retention in B2B contexts
- • Gameable: agents can coach customers on scoring
CSAT as Touchpoint Metric
- • Measures specific interaction quality
- • Identifies training needs per touchpoint type
- • Fast to deploy and easy to act on
- • Pairs with CES for effort measurement
- • Segment by touchpoint type before reporting
The replacement: Use CSAT for what it measures (touchpoint quality) and add product usage health and outcome achievement metrics for what it does not measure (relationship health and churn risk). A customer who scores 5 on CSAT and has declining login frequency is at risk. A customer who scores 3 on CSAT and has stable, high usage is probably fine. The CSAT alone tells you nothing about either.
AHT: Speed Is Not Quality
AHT (Average Handle Time) measures how long interactions take. It is useful as a diagnostic: an agent with unusually high AHT on a specific query type may need training or better knowledge base access. It is destructive as a performance target: when agents are measured on speed, they rush, cut corners, and reduce quality.
The lie: “Lower AHT means more efficient service.” Lower AHT means shorter calls. Shorter calls can mean faster resolution (good) or rushed, incomplete resolution (bad). Without pairing AHT with FCR (First Contact Resolution), you cannot tell which. An agent who reduces AHT by transferring complex cases to other teams has not improved efficiency. They have shifted the cost.
The Goodhart’s Law Problem
Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure. In contact centres, this manifests as:
- • Agents rush callers off the phone to hit AHT targets
- • Complex cases are transferred or escalated to avoid inflating the average
- • Root-cause diagnosis is skipped in favour of symptom treatment
- • FCR drops; repeat contacts increase; total contact volume rises
- • Customer effort increases; satisfaction drops; churn risk increases
The organisation optimised for lower AHT and got worse service, more repeat contacts, and higher total cost. The metric improved. The outcome did not.
The Replacement: FCR + CES
Replace AHT as a performance target with:
Resolution Rate: The Definition Problem
Resolution Rate measures the percentage of tickets or interactions that are marked as resolved. It sounds like the ultimate quality metric. The problem is the definition of “resolved.” A ticket is typically marked resolved when the agent closes it. That does not mean the customer’s problem was solved. It means the agent decided the interaction was over.
The lie: “Our resolution rate is 92%.” If 30% of those “resolved” tickets are reopened within 7 days (the customer contacts again about the same issue), the real resolution rate is 64%. The metric is lying because the definition of resolved is controlled by the agent, not validated by the customer’s behavior.
The Honest Resolution Rate
The honest resolution rate controls for repeat contacts. The formula:
True Resolution Rate = (Tickets resolved AND not reopened within 7 days) / (Total tickets)
This requires joining your ticketing system data to identify repeat contacts from the same customer on the same issue within a 7-day window. Most ticketing systems can do this with a custom report. The honest number is always lower than the reported number, and it is the only number that reflects actual service quality.
The Goodhart’s Law manifestation: When resolution rate is a performance target, agents close tickets before the problem is solved to hit the target. The rate improves. The repeat contact rate increases. Total contact volume rises. The customer experiences worse service. The dashboard shows improvement. This is the metric lying in real time.
What to Use Instead
The replacement for lying metrics is not a single better metric. It is a signal stack that combines leading and lagging indicators, separates diagnostic metrics from performance targets, and never ties a single metric to individual incentives.
The Minimum Honest Signal Stack
The rule: Leading indicators tell you what is about to happen. Quality indicators tell you how your service is performing. Lagging indicators tell you what already happened. Diagnostics tell you where to investigate. Never use a lagging indicator as a leading indicator, a diagnostic as a target, or a single metric as your entire measurement system.
The Honest Dashboard
An honest CX dashboard has four sections, each with a different purpose. The sections are not equal in size. The leading indicators section should be the largest, because that is where action happens.
Section 1: Leading Indicators (60% of dashboard)
- • Login frequency delta from per-customer baseline (weekly)
- • Feature adoption trend by account (weekly)
- • Support contact rate change by account (weekly)
- • Champion status changes (monthly, via LinkedIn monitoring)
- • Customer health score distribution (weekly, by segment)
This is where you intervene before churn happens.
Section 2: Quality Indicators (20% of dashboard)
- • True FCR (controlled for 7-day repeat contacts)
- • CES by touchpoint type (post-interaction)
- • CSAT by touchpoint type (segmented, not aggregated)
This is where you measure service quality, not speed.
Section 3: Lagging Indicators (15% of dashboard)
- • NPS trend by segment (quarterly, directional only)
- • Churn rate by cohort (monthly)
- • Renewal rate by segment (quarterly)
This is where you confirm whether your interventions worked.
Section 4: Diagnostics (5% of dashboard)
- • AHT by query type (identify training needs, not performance)
- • Ticket volume trend (identify product issues)
- • Escalation rate (identify complexity trends)
This is where you investigate, not where you target.
Implementation: Replacing Lying Metrics
Replacing lying metrics is a political exercise, not a technical one. The data exists. The tooling exists. The resistance comes from people whose performance is measured by the lying metrics and who fear the honest replacement will make them look worse. It will, initially, because the honest number is always lower than the lying number. That is the point.
The implementation sequence:
Run both metrics in parallel for 90 days
Keep the lying metric on the dashboard and add the honest replacement alongside it. Do not remove the lying metric yet. Let the team see the gap between the two numbers.
Educate leadership on the gap
Present the parallel data to leadership. Show that the honest resolution rate is 64% while the reported rate is 92%. Explain why the gap exists and what it means for actual service quality.
Remove the lying metric from performance targets
Stop tying individual incentives to the lying metric. This is the hardest step politically. The metric can stay on the dashboard as a diagnostic, but it must not drive individual performance reviews.
Replace with the honest metric in the next cycle
Once the team has adjusted to the honest metric not being a target, make it the primary metric. Set realistic targets based on the honest baseline, not the inflated lying baseline.
Review quarterly and recalibrate
The honest metric may reveal new patterns. Review quarterly, adjust the signal stack, and continue refining. Measurement is not a one-time project; it is an ongoing discipline.
The honest answer: Replacing lying metrics will make your numbers look worse before they make your outcomes better. That is the cost of honesty. The alternative is a dashboard that looks good while your customers churn. Most organisations choose the good-looking dashboard. That is why most CX measurement does not work.
Related Reading in The CX Field Guide
This is the contrarian companion to Part 5 of the CX Field Guide. Read Part 5 for the full measurement framework, then this guide for the metrics to remove from your dashboard. The customer health scoring framework provides the leading indicator system that replaces the lying lagging indicators.
Analytics & Measurement Tools
Analytics platforms, VoC tools, and sentiment analysis systems for building an honest measurement stack:
Amplitude
Product analytics platform that helps businesses understand user behaviour and optimise customer experiences.
Buyer fit
Enterprise, Mid-Market, Startup teams that need Behavioural Targeting and Cohort Analysis coverage.
Key Features
- Behavioural Targeting
- Cohort Analysis
- Customer Segmentation
- Funnel Analysis
Hotjar
Behaviour analytics and user feedback platform that helps understand visitor behaviour on websites.
Buyer fit
Small Business, Mid-Market, Startup teams that need Customer Segmentation and Funnel Analysis coverage.
Key Features
- Customer Segmentation
- Funnel Analysis
- User Behaviour Analysis
- API Integration
Mixpanel
Advanced analytics platform for product teams to analyse user behaviour and understand engagement.
Buyer fit
Mid-Market, Small Business, Startup teams that need Cohort Analysis and Customer Segmentation coverage.
Key Features
- Cohort Analysis
- Customer Segmentation
- Funnel Analysis
- Predictive Analytics
Alchemer
Survey platform with advanced logic features, customization options, and integrations (including Salesforce and Zendesk).
Buyer fit
Mid-Market, Enterprise teams that need Sentiment Analysis and Email Distribution coverage.
Key Features
- Sentiment Analysis
- Email Distribution
- In-App / In-Product Surveys
- Multi-Channel Distribution
Qualtrics XM
Enterprise experience management platform with advanced survey capabilities, real-time analytics, and AI-driven insights.
Buyer fit
Enterprise, Mid-Market teams that need AI-Powered Text Analysis and Customer Segmentation coverage.
Key Features
- AI-Powered Text Analysis
- Customer Segmentation
- Robust Statistical Analysis
- Email Distribution
SurveyMonkey
Survey platform with a drag-and-drop builder and templates, used by businesses of all sizes.
Buyer fit
Small Business, Mid-Market, Startup teams that need Email Distribution and Multi-Channel Distribution coverage.
Key Features
- Email Distribution
- Multi-Channel Distribution
- Multilingual Support
- API Integration
Frequently Asked Questions
Which CX metrics are misleading?
The CX metrics that most commonly mislead organisations are: NPS (when used as the sole CX health indicator, it hides churn signals because satisfied customers churn for behavioral reasons unrelated to survey scores), CSAT (when used as a proxy for relationship health, it captures touchpoint satisfaction not product value), AHT (when used as a performance target, it incentivises agents to rush and reduces first contact resolution), and resolution rate (when the definition of ‘resolved’ is not controlled for repeat contacts within 7 days). Each metric is useful in a specific context and misleading when used outside that context.
Why does NPS lie?
NPS lies because it measures stated intent to recommend, not actual behavior. A customer who scores 9 on an NPS survey can still be shopping competitors, reducing usage, and preparing to churn. NPS is susceptible to survey timing bias (scores inflate after positive events), response bias (satisfied and dissatisfied customers respond at higher rates than neutral customers), and the promoter definition (scoring 9-10 excludes the majority of satisfied customers who score 7-8). NPS is useful as a directional tracking metric over time but misleading as a sole indicator of account health.
What should you use instead of NPS?
Use a combination of metrics instead of relying on NPS alone: product usage health (login frequency delta from baseline, feature adoption) as the leading indicator of churn; CES (Customer Effort Score) as the predictor of loyalty in transactional relationships; customer health score (multi-vector composite) as the per-account risk indicator; and NPS as one of several signals, not the signal. The replacement is not a single metric; it is a signal stack that combines leading and lagging indicators.
Why is CSAT not a good measure of customer success?
CSAT is not a good measure of customer success because it measures satisfaction with a specific interaction, not whether the customer is achieving the outcome they paid for. A customer can be satisfied with your support team’s friendliness while failing to get value from your product. They will score you an 8 on CSAT and then not renew. CSAT captures feelings at a moment in time; success is a state. Use CSAT for touchpoint quality measurement (post-interaction) and pair it with product usage health and outcome achievement metrics for relationship health.
What is the Goodhart’s Law problem in CX metrics?
Goodhart’s Law states that when a measure becomes a target, it ceases to be a good measure. In CX, this manifests when metrics tied to individual performance incentives are gamed: agents ask customers to give them a 10 on NPS (score improves, experience does not), support teams close tickets before resolution to hit resolution rate targets (rate improves, repeat contacts increase), and contact centres reduce AHT by rushing callers (handle time drops, FCR drops and customer effort increases). Any CX metric tied to individual incentives will be gamed by rational people optimising for what they are measured on.
How do you build an honest CX dashboard?
Build an honest CX dashboard by combining leading and lagging indicators, separating diagnostic metrics from performance targets, and never tying a single metric to individual incentives. Include: one relationship metric (NPS or relationship CSAT) tracked as a trend, not a target; one transactional metric (CES or post-interaction CSAT) for touchpoint quality; two to three behavioral signals (login frequency delta, feature adoption, support contact rate) as leading indicators; and one operational metric (FCR, not AHT) for service quality. Review the dashboard weekly for delta, not for absolute scores.
The Cleanup to Part 5
This guide is the contrarian companion to the CX Field Guide’s measurement framework. Part 5 covers what to measure. This guide covers what to stop measuring and why. Together, they form the complete measurement discipline: build the honest signal stack, remove the lying metrics, and never tie a single metric to individual incentives.
View All Insights →