Apdex Scoring

Apdex (Application Performance Index) is an industry-standard metric that quantifies user satisfaction with application response times. Pingara calculates Apdex for every monitor, giving you a single number that reflects real-world user experience.

What Is Apdex?

Apdex converts raw performance data (response times) into a satisfaction score between 0.0 and 1.0.

Higher scores = Better performance

• 1.0 - Perfect (all users satisfied)
• 0.94 - Excellent (94% user satisfaction)
• 0.50 - Poor (half your users frustrated)
• 0.0 - Unacceptable (all users frustrated)

Why Apdex Matters

Traditional metrics like average response time hide important details:

• Your average might be 200ms
• But if 20% of requests take 5 seconds, users are frustrated
• Apdex reveals this instantly

Apdex shows: What percentage of your users had a good experience?

How Apdex Works

The Formula

Apdex = (Satisfied + Tolerating/2) / Total

Every check is categorized as:

1. Satisfied - Response time ≤ T threshold
2. Tolerating - Response time between T and 4T
3. Frustrated - Response time > 4T OR errors/timeouts

Example:

If T = 500ms and you have 100 checks:

• 80 checks ≤ 500ms (satisfied)
• 15 checks 501ms–2000ms (tolerating)
• 5 checks > 2000ms or errors (frustrated)

Apdex = (80 + 15/2) / 100
      = (80 + 7.5) / 100
      = 87.5 / 100
      = 0.875 (Good)

The T Threshold

T is your target response time, the maximum acceptable duration for a satisfactory user experience.

Choosing T:

• Web pages: 500ms–1000ms
• APIs: 100ms–500ms
• Background services: 1000ms–5000ms

Set T based on your users' expectations, not technical minimums.

Tip: If 90% of your requests normally complete in 300ms, set T = 500ms to catch degradation early.

Configuring Apdex in Pingara

Setting the T Threshold

When creating or editing a monitor:

1. Navigate to Advanced Settings
2. Find Apdex Threshold (T)
3. Enter your target response time in milliseconds
4. Click Save

Default: 500ms (suitable for most web applications)

Per-Monitor Thresholds

Each monitor can have its own T value:

• Marketing site: T = 1000ms (users tolerate slower loads)
• API endpoint: T = 200ms (developers expect fast responses)
• Admin dashboard: T = 500ms (internal tool, moderate expectations)

Why different thresholds? Users have different performance expectations for different services.

Interpreting Apdex Scores

Score Ratings

Pingara uses industry-standard Apdex ratings:

Trend Analysis

Apdex is most useful when tracked over time:

Healthy trend:

Mon: 0.95 (Excellent)
Tue: 0.94 (Excellent)
Wed: 0.96 (Excellent)

Stable, high performance.

Degrading trend:

Mon: 0.95 (Excellent)
Tue: 0.87 (Good)
Wed: 0.72 (Fair)

Something is slowing down. Investigate immediately.

Recovering trend:

Mon: 0.60 (Poor)
Tue: 0.80 (Fair)
Wed: 0.92 (Good)

Performance improving after optimization or incident resolution.

Apdex vs Other Metrics

Apdex vs Average Response Time

Average response time can be misleading:

• Hides outliers
• Doesn't reflect user satisfaction
• One slow request skews the average

Apdex shows the full picture:

• Counts satisfied vs frustrated users
• Outliers directly impact the score
• Instantly reveals performance issues

Example:

Average barely changed, but Apdex dropped from Excellent to Fair, revealing that a significant number of requests became slow.

Apdex vs Uptime

Uptime = Did the service respond?

Apdex = Did the service respond fast enough?

Your monitor can have:

• ✅ 100% uptime
• ❌ Apdex = 0.50 (Poor)

This means your site is up but slow, so users are still frustrated.

Use both:

• Uptime catches outages
• Apdex catches performance degradation

When Pingara Marks Monitors as "Degraded"

Degraded is a raw-latency rule, not an Apdex rule. Apdex is computed for reporting only. It does not feed detection, and no Apdex score, on its own, changes a monitor's status. What actually triggers Degraded is a separate, independently configurable pair of settings:

• Slow-response threshold - a response time, in milliseconds. If you don't set one, it defaults to 4× your Apdex threshold (T), which is why the two so often look connected.
• Consecutive slow checks required - how many checks in a row must exceed that threshold. If you don't set one, it defaults to 3.

Once both are met, the transition to Degraded is confirmed the same way a Down transition is: through quorum across your configured regions, not a single check.

Why the distinction matters: because the threshold only defaults to 4T, changing your Apdex threshold changes where Degraded triggers unless you've set the slow-response threshold explicitly. And setting the slow-response threshold on its own, without touching Apdex, moves Degraded without moving your Apdex score at all. Treat them as two settings that happen to share a default, not one setting under two names.

Why "Degraded" matters:

• Alerts you before a full outage
• Lets you investigate performance issues proactively
• Visible on status pages (users see "Partial Outage")

Example (using the defaults):

• T = 500ms → slow-response threshold defaults to 2000ms
• If 3 consecutive checks (the default) exceed 2000ms, and quorum confirms it → Status = Degraded

Practical Tips

Set Realistic Thresholds

Too strict:

• T = 100ms for a database-backed web page
• Constant false alarms
• Apdex always Poor

Too lenient:

• T = 5000ms for a simple API
• Users frustrated but Apdex says Excellent
• You miss real issues

Right approach:

1. Monitor your baseline performance for a week
2. Check p95 response time (95th percentile)
3. Set T slightly above p95
4. Adjust based on user feedback

Monitor Apdex Trends, Not Just Scores

A single low score might be a transient blip. Watch for:

• Sustained drops - Performance degrading over hours/days
• Spikes during peak hours - Capacity issues
• Regional differences - Network or server problems in specific regions

Combine with Latency Breakdown

If Apdex drops, drill into the performance breakdown:

• High DNS time → DNS server issue
• High TCP connect → Network congestion
• High TTFB → Backend processing slow
• High total duration → Large response size or transfer bottleneck

Use the latency chart to identify where the slowdown occurs.

Use Apdex for SLA Reporting

Many teams report Apdex in monthly SLAs:

Example SLA:

"Our API will maintain an Apdex score ≥0.90 (T=200ms) for 99.5% of calendar days."

Apdex gives a user-centric performance guarantee beyond simple uptime.

Apdex on the Dashboard

Overview Card

The dashboard shows overall Apdex, a weighted average across all active monitors:

• ✅ 0.94+ → Green, Excellent
• ⚠️ 0.70–0.93 → Yellow, Fair to Good
• 🚨 <0.70 → Red, Poor to Unacceptable

Per-Monitor Apdex

Click any monitor to see:

• Current Apdex score
• 24-hour, 7-day, 30-day trends
• Breakdown by region (Pro feature)
• Histogram of response times

Apdex in Incident Summaries

When an incident is resolved, the incident report includes:

• Average Apdex during outage
• Apdex recovery timeline
• Comparison to baseline Apdex

This helps diagnose whether the issue was a full outage or performance degradation.

Troubleshooting Low Apdex

Apdex < 0.5 (Unacceptable)

Investigate:

1. Check for active incidents (site down = Apdex near 0)
2. Review error rates (timeouts, 500 errors)
3. Look at recent deployments (did code change break something?)

Common causes:

• Database overload
• Memory leak
• DDoS attack
• Infrastructure failure

Apdex 0.5–0.7 (Poor)

Investigate:

1. Check backend resource usage (CPU, memory, disk I/O)
2. Review slow query logs
3. Check third-party API latencies
4. Verify CDN performance

Common causes:

• Unoptimized database queries
• Cold cache (after restart)
• Increased traffic without scaling
• Slow external API calls

Apdex 0.7–0.85 (Fair)

Investigate:

1. Look for gradual performance regression
2. Check time-to-first-byte (TTFB)
3. Review recent code deployments
4. Check for N+1 query patterns

Common causes:

• Code inefficiency introduced in recent deploy
• Gradual data growth slowing queries
• Network path degradation

Next Steps

• HTTP/HTTPS Monitoring - Configure monitors and performance tracking
• Monitor Intervals and Regions - Optimize check frequency and location
• Understanding Incidents - How Apdex triggers degraded status
• Setting Up Alerts - Get notified when Apdex drops