Metrics & Benchmarks

What gets measured gets managed — but only if you’re measuring the right things. Good metrics connect user behavior to real outcomes. Bad metrics create perverse incentives and waste engineering effort.

This page provides industry benchmarks, measurement frameworks, and validation rules for UX metrics that both humans and AI systems can use to assess interface quality.

---

Metric Specifications

Industry Benchmarks Summary

Metric	Average	Good	Excellent	Source
Task Success Rate	78%	80-90%	>90%	MeasuringU
SUS Score	68	70-80	>80	Usability.gov
SEQ Score	5.5/7	6.0/7	6.5+/7	MeasuringU
NPS (SaaS)	+36	+40-50	+50+	CustomerGauge
NPS (B2B)	+38	+40-60	+60+	Retently
NPS (eCommerce)	+55	+55-70	+70+	CustomerGauge
LCP	—	≤2.5s	≤1.5s	Google
INP	—	≤200ms	≤100ms	Google
CLS	—	≤0.1	≤0.05	Google

UX Metric Categories

Category	Measures	Examples
Behavioral	What users do	Task success, time on task, error rate
Attitudinal	How users feel	SUS, NPS, SEQ, CSAT
Performance	System quality	LCP, INP, CLS, Time to Interactive
Business	Outcomes	Conversion, retention, revenue per user

---

Validation Rules

ux_metrics_validation:
  rules:
    - id: task-success-threshold
      severity: warning
      check: "Task success rate >= 78% for critical tasks"
      threshold: 78
      action: "Investigate usability issues if below threshold"

    - id: sus-score-minimum
      severity: warning
      check: "SUS score >= 68 (industry average)"
      threshold: 68
      action: "Below average indicates usability problems"

    - id: seq-score-minimum
      severity: warning
      check: "SEQ score >= 5.5/7 for all tasks"
      threshold: 5.5
      action: "Low SEQ indicates task difficulty"

    - id: lcp-good
      severity: error
      check: "LCP <= 2.5 seconds at 75th percentile"
      threshold_ms: 2500
      action: "Optimize loading performance"

    - id: inp-good
      severity: error
      check: "INP <= 200ms at 75th percentile"
      threshold_ms: 200
      action: "Optimize interaction responsiveness"

    - id: cls-good
      severity: error
      check: "CLS <= 0.1 at 75th percentile"
      threshold: 0.1
      action: "Fix layout shifts"

    - id: baseline-before-changes
      severity: error
      check: "Baseline measurement exists before design changes"
      rationale: "Cannot measure improvement without baseline"

    - id: sample-size-sufficient
      severity: warning
      check: "Quantitative metrics have n >= 20 responses"
      rationale: "Statistical significance requires adequate sample"

    - id: guardrail-monitoring
      severity: warning
      check: "Guardrail metrics monitored alongside north star"
      rationale: "Prevent gaming at users' expense"

---

Types of Metrics

North-Star Metrics

The one metric that best captures the value you deliver to users. Everything else ladders up to it.

Company	North Star	Why It Works
Spotify	Time spent listening	Reflects value delivered
Airbnb	Nights booked	Captures marketplace success
Slack	Messages sent in channels	Indicates team engagement
Duolingo	Daily active learners	Learning progress
Notion	Weekly active editors	Collaboration value

Criteria for a good north star:

• Meaningful — Reflects real user value
• Measurable — Quantifiable and trackable
• Movable — Your team can influence it
• Leading — Predicts business outcomes

Input Metrics (Leading Indicators)

Leading indicators you can influence directly. They predict changes in your north star.

Metric	Predicts	Example Target
Activation rate	Retention	40% in first week
Feature adoption	Engagement	30% try new feature
Session frequency	Lifetime value	3+ sessions/week
Onboarding completion	Activation	70% complete flow
Time to first value	Conversion	<5 minutes

Guardrail Metrics

Constraints that ensure you’re not gaming your north star at users’ expense.

Guardrail	Protects Against	Alert Threshold
Error rates	Broken experiences	>1% increase
Page load time	Performance degradation	>500ms increase
Support volume	Confusion, frustration	>10% increase
Accessibility compliance	Exclusion	Any regression
Bounce rate	Poor content match	>5% increase
Unsubscribe rate	Over-communication	>0.5% per send

Rule: If your guardrails decline while your north star rises, something’s wrong.

---

HEART Framework (Google)

Google’s UX research team (Kerry Rodden, Hilary Hutchinson, Xin Fu) created the HEART framework to measure user satisfaction and product success at scale.

The Five HEART Metrics

Dimension	Definition	Example Signals	Example Metrics
Happiness	User attitudes, satisfaction	Survey responses	NPS, SUS, CSAT
Engagement	Level of involvement	Session depth	Sessions/week, features used
Adoption	New users gained	Sign-ups, upgrades	% new users, feature adoption
Retention	Users returning	Return visits	Churn rate, DAU/MAU ratio
Task Success	Efficiency, effectiveness	Task completion	Success rate, time on task

Goals-Signals-Metrics Process

┌─────────────┐    ┌─────────────┐    ┌─────────────┐
│    GOALS    │ → │   SIGNALS   │ → │   METRICS   │
│ What success│    │ Indicators  │    │ How to      │
│ looks like  │    │ of progress │    │ measure     │
└─────────────┘    └─────────────┘    └─────────────┘

Example:
Goal: Users find what they need quickly
Signal: Low search refinement rate
Metric: % searches with 0-1 refinements

When to Use Each HEART Dimension

Scenario	Primary Dimensions
New feature launch	Adoption, Task Success
Retention problem	Retention, Engagement
Conversion optimization	Task Success, Happiness
Growth initiative	Adoption, Retention
Quality assessment	Happiness, Task Success

---

UX-Specific Metrics

Task Success Rate

The most fundamental UX metric: what percentage of users complete a given task?

Industry benchmark: 78% average (MeasuringU analysis of 1,100+ tasks)

Performance	Task Success Rate
Excellent	>90%
Good	80-90%
Acceptable	70-80%
Poor	<70%

Calculation:

Task Success Rate = (Successful Completions / Attempts) × 100

Types of success:

• Binary — Completed or not
• Levels — Complete, partial, failed
• Assisted — Track moderator interventions

Time on Task

How long does it take? Context determines whether shorter is better.

Task Type	Shorter is Better	Longer May Be Good
Checkout	✓
Error recovery	✓
Content consumption		✓
Learning features		✓
Data entry	✓

Calculation:

Time on Task = Task End Time - Task Start Time
Report: Mean, Median, 90th percentile

Error Rate

How often do users make mistakes? Distinguish between types.

Error Type	Definition	Example
Slip	Wrong action, right intention	Clicking wrong button
Mistake	Right action, wrong understanding	Misinterpreting a label
Failure	Cannot complete	Giving up

Calculation:

Error Rate = (Tasks with Errors / Total Tasks) × 100

Lostness

Measures navigation efficiency — how directly users find what they need.

Calculation:

Lostness = sqrt((N/S - 1)² + (R/N - 1)²)

Where:
N = Number of different pages visited
S = Minimum number of pages needed
R = Total pages visited (including revisits)

Score interpretation:
0.0 = Perfect navigation
0.4 = Some confusion
0.5+ = Lost

---

Survey-Based Metrics

System Usability Scale (SUS)

A 10-item questionnaire producing a score from 0-100. Quick, reliable, and widely benchmarked.

Scoring interpretation:

Score	Grade	Percentile	Adjective
90+	A+	Top 10%	Best imaginable
80-89	A	Top 20%	Excellent
68-79	B	Average	Good
50-67	C	Below average	OK
<50	F	Bottom 25%	Poor

Key benchmark: 68 is the average SUS score across all products.

Single Ease Question (SEQ)

One question asked immediately after a task: “How easy or difficult was this task?”

• Scale: 1 (Very Difficult) to 7 (Very Easy)
• Average: 5.5/7
• Threshold: Below 5.0 indicates difficulty

Benefits:

• Minimal burden on participants
• Task-specific (unlike SUS)
• Correlates with task success

Net Promoter Score (NPS)

“How likely are you to recommend this to a friend?” (0-10 scale)

Calculation:

NPS = % Promoters (9-10) - % Detractors (0-6)
Range: -100 to +100

Industry benchmarks (2024):

Industry	Average NPS	Good	Excellent
SaaS	+36	+40	+50+
B2B Software	+41	+45	+55+
eCommerce	+55-62	+60	+70+
B2B Overall	+38	+45	+60+
B2C Overall	+49	+55	+70+

Caution: NPS measures sentiment, not usability. A product can have high NPS but poor task success.

SUPR-Q (Website Quality)

An 8-item questionnaire measuring four factors of website experience:

Factor	What It Measures
Usability	Ease of use
Trust	Credibility, security
Appearance	Visual design quality
Loyalty	Intent to return/recommend

Advantage: Provides percentile ranks against database of 100+ websites.

UMUX-Lite

A short, 2-item usability survey:

1. “[Product]‘s capabilities meet my requirements”
2. “[Product] is easy to use”

Scale: 7-point Likert Conversion: Can be converted to SUS-equivalent score

---

Core Web Vitals (Performance)

Google’s metrics for user experience quality, used in search ranking since 2021.

Current Metrics (2024-2025)

Metric	Measures	Good	Needs Improvement	Poor
LCP	Loading	≤2.5s	2.5-4.0s	>4.0s
INP	Interactivity	≤200ms	200-500ms	>500ms
CLS	Visual stability	≤0.1	0.1-0.25	>0.25

Note: INP (Interaction to Next Paint) replaced FID (First Input Delay) in March 2024.

Largest Contentful Paint (LCP)

Measures loading performance — when the largest content element becomes visible.

Common causes of poor LCP:

• Slow server response (TTFB)
• Render-blocking JavaScript/CSS
• Large unoptimized images
• Client-side rendering delays

Interaction to Next Paint (INP)

Measures responsiveness — latency of all interactions throughout the page lifecycle.

Common causes of poor INP:

• Long JavaScript tasks blocking main thread
• Large DOM size
• Excessive event handlers
• Third-party scripts

Cumulative Layout Shift (CLS)

Measures visual stability — unexpected movement of visible elements.

Common causes of poor CLS:

• Images without dimensions
• Ads/embeds without reserved space
• Dynamically injected content
• Web fonts causing FOIT/FOUT

Passing Core Web Vitals

To pass assessment, 75% of page visits must meet “Good” thresholds.

Passing = (Pages with "Good" Score / Total Pages) >= 0.75

---

Setting Benchmarks

1. Baseline First

Measure where you are before making changes. Without a baseline, you can’t prove improvement.

FUNCTION establishBaseline(metric):
  measurements = collectData(metric, days=14-30)

  baseline = {
    mean: average(measurements),
    median: median(measurements),
    p75: percentile(measurements, 75),
    p95: percentile(measurements, 95),
    sample_size: count(measurements),
    date_range: [start, end]
  }

  RETURN baseline

2. Industry Comparison

Compare against public benchmarks:

Source	Benchmarks Available
MeasuringU	SUS, task success, time on task
Google CrUX	Core Web Vitals by industry
WebAIM Million	Accessibility errors
CustomerGauge	NPS by industry

3. Internal Trends

Compare against your own history:

Improvement = ((New Score - Baseline) / Baseline) × 100

Track:

• Week-over-week changes
• Release-over-release changes
• Seasonal variations

4. Target Setting

Set meaningful targets, not arbitrary numbers.

Good targets:

• “Improve task success from 72% to 80%”
• “Reduce LCP from 3.2s to 2.5s”
• “Achieve SUS score of 75 (from 65)”

Bad targets:

• “Get 100% task success” (unrealistic)
• “Be the best in industry” (vague)
• “Improve all metrics” (unfocused)

---

Decision Logic for Metric Selection

FUNCTION selectMetrics(product_stage, goals):
  metrics = []

  // Always include task success for usability
  metrics.push("task_success_rate")

  // Select attitudinal metric
  IF goals.includes("overall_usability"):
    metrics.push("SUS")
  IF goals.includes("task_specific_feedback"):
    metrics.push("SEQ")
  IF goals.includes("loyalty_measurement"):
    metrics.push("NPS")

  // Select behavioral metrics based on stage
  IF product_stage == "new_feature":
    metrics.push("adoption_rate")
    metrics.push("feature_usage_frequency")
  IF product_stage == "retention_focus":
    metrics.push("DAU_MAU_ratio")
    metrics.push("churn_rate")
  IF product_stage == "growth":
    metrics.push("activation_rate")
    metrics.push("time_to_first_value")

  // Always include performance guardrails
  metrics.push("LCP", "INP", "CLS")

  // Always include error guardrails
  metrics.push("error_rate")

  RETURN metrics

---

Implementation Tips

Instrumentation

• Consistent events across platforms (web, mobile, API)
• Standardized naming conventions
• User identity resolution for cross-session tracking
• Timestamp precision for time-based metrics

Tracking Pipeline

User Action → Event Collection → Processing → Storage → Dashboard
     │              │                │            │          │
     │         (Real-time)       (Batch)     (Analytics)   (BI)
     │                                                      │
     └──────────────────── Alert if threshold breached ─────┘

Team Rituals

Cadence	Activity	Metrics Focus
Daily	Monitor dashboards	Anomalies, errors
Weekly	Team review	Trends, progress
Sprint	Retrospective	Feature impact
Monthly	Stakeholder report	Business outcomes
Quarterly	Benchmarking study	Industry comparison

Investigation Protocol

When metrics change unexpectedly:

1. Verify data — Is the change real or a tracking bug?
2. Identify scope — Which segments affected?
3. Correlate events — What changed (deploy, external)?
4. Root cause — Why did behavior change?
5. Action — Fix, revert, or accept?

---

Common Traps

Vanity Metrics

Big numbers that don’t connect to outcomes.

Vanity Metric	Better Alternative
Page views	Engaged sessions
Downloads	Activated users
Sign-ups	7-day retained users
Time on site	Task completion

Goodhart’s Law

“When a measure becomes a target, it ceases to be a good measure.”

Example: Targeting “sessions per user” leads to artificial engagement features that annoy users and hurt long-term retention.

Prevention: Use guardrail metrics alongside targets.

Short-Termism

Optimizing for immediate conversion at the cost of retention.

Example: Dark patterns increase sign-ups but increase churn and damage trust.

Prevention: Track 90-day retention alongside conversion.

Analysis Paralysis

Tracking everything, acting on nothing.

Prevention: Limit to 3-5 key metrics per initiative. More data ≠ better decisions.

---

Recent Research (2024-2025)

NPS Decline Trend

CustomerGauge 2024-2025 data shows median NPS remaining at 42, but 10 of 11 industries saw NPS decline. Only Manufacturing improved. Wholesale, Retail & eCommerce, and Digital Marketplaces took double-digit hits.

INP Replacing FID

Google replaced First Input Delay (FID) with Interaction to Next Paint (INP) as a Core Web Vital in March 2024. INP measures all interactions throughout the page lifecycle, not just the first one.

AI in Metrics Analysis

Tools like Mixpanel, Amplitude, and Heap are incorporating AI for anomaly detection, funnel optimization suggestions, and natural language querying of analytics data. However, human interpretation remains essential for understanding causation.

Accessibility Metrics Maturation

The WebAIM Million study continues annual tracking, showing 95.9% of home pages have detectable WCAG 2 failures in 2024. Accessibility compliance is increasingly treated as a guardrail metric alongside performance.

---

References

Frameworks:

• HEART Framework — Google Research — Original HEART paper
• Google’s HEART Framework — IxDF
• How Google Uses HEART — Appcues

Survey Metrics:

• System Usability Scale — Usability.gov
• What is a Good Net Promoter Score — Retently
• SUPR-Q — MeasuringU

Performance:

• Core Web Vitals — web.dev
• Core Web Vitals Thresholds — web.dev
• Core Web Vitals 2025 — EnFuse Solutions

Benchmarking:

• 47 UX Metrics Articles from 2024 — MeasuringU
• Task Success Rate — NNg
• UX Benchmarking — NNg
• NPS Benchmarks by Industry — CustomerGauge

Practical Guides:

• 6 UX Metrics for 2025 — Survicate
• UX Metrics Complete Guide — Evo Design
• Usability Testing Metrics — UXtweak

---

See Also

• Usability Tests — Collecting behavioral metrics
• Cognitive Walkthroughs — Expert evaluation methods
• Accessibility Checklist — Compliance metrics
• Key Principles and Laws — Design principles affecting metrics