The Problem With How We Look At Performance.
Every Monday, I see dashboards showing "average response time: 2.3 hours" or "average delivery time: 3.2 days." Teams nod, check green boxes, move on.
But there's a fundamental issue: averages hide the story of what's actually happening to your customers and operations.
Most of us weren't taught to think in percentiles. We learned averages in school, use averages in spreadsheets, and build dashboards around averages. But when you're running operations, averages can be dangerously misleading.
Understanding Percentiles (The 5-Minute Explanation).
What percentiles tell you: What experience did X% of your customers actually have?
50th percentile (median): Half your customers had better performance, half had worse
90th percentile: 90% of customers had better performance, 10% had worse
95th percentile: 95% of customers had better performance, 5% had worse
Simple example: If your 90th percentile response time is 8 hours, it means 1 out of every 10 customers waits over 8 hours.
Why this matters: That 10% often includes your most complex cases, biggest customers, or most critical situations.
The Customer Service Disaster You Don't Know You Have.
I was reviewing a support metrics deck for a friend at another company. The dashboard looked great:
Average response time: 90 minutes ✅ Target: <2 hours ✅ Status: Green ✅
But when I dug into percentiles:
50th percentile (median): 45 minutes (excellent)
90th percentile: 8 hours (terrible)
95th percentile: 18 hours (catastrophic)
Translation: While 50% of customers got lightning-fast service, 10% waited over 8 hours. These weren't random customers—they were typically the most complex cases, often from enterprise clients paying 10x more than regular customers.
The damage: Three high-value clients had already started looking at competitors. The "green" dashboard was hiding ₹2.5 crores in annual revenue risk.
The Most Dangerous Illusion - When Nothing Changes But Everything Gets Worse.
Here's the scenario that keeps me up at night. It's from call center performance data:
Month 1 Performance:
60% of calls: 2 minutes
30% of calls: 8 minutes
10% of calls: 32 minutes
Average: 6.8 minutes
Median (50th percentile): 2 minutes
Month 3 "Same" Performance:
70% of calls: 2 minutes
20% of calls: 12 minutes
10% of calls: 38 minutes
Average: 6.8 minutes (exactly the same!)
Median (50th percentile): 2 minutes (exactly the same!)
The dashboard: Everything looks identical. Average unchanged. Median unchanged. Status: Green.
The hidden disaster:
80th percentile Month 1: 8 minutes
80th percentile Month 3: 12 minutes (50% worse)
90th percentile Month 1: 32 minutes
90th percentile Month 3: 38 minutes (19% worse)
What actually happened:
More customers got fast service (70% vs 60%)
But everyone else got significantly worse service
Every single percentile above the median got worse
The business impact: Customer complaints increased 40%. Employee stress skyrocketed. Complex cases (often high-value customers) took longer to resolve. But leadership saw "stable performance" on their dashboards and did nothing.
This is the most insidious form of operational decay—when aggregate metrics hide systematic deterioration.
The Percentile Framework That Actually Works.
Stop tracking these Vanity Metrics:
Average response time
Average delivery time
Average customer satisfaction
Average sales per rep
Start tracking these Reality Metrics:
Customer experience - 90th & 95th percentile response times.
System reliability - 99th percentile latency.
Delivery promises - 95th percentile fulfillment time.
Sales performance - Median + bottom quartile results.
Service quality - 5th percentile customer satisfaction.
Why these numbers matter:
90th percentile = Your "bad day" experience.
95th percentile = Your "disaster" experience.
99th percentile = Your "crisis" experience.
Bottom quartile = Where your team is struggling.
The Monday Morning Action Plan.
This week, do this:
Audit one key metric you currently track as an average
Pull the raw data and calculate 50th, 90th, 95th percentiles
Ask yourself: "What would happen if 10% of my customers experienced the 90th percentile performance?"
Example calculation in Excel/ Google sheet:
=PERCENTILE(data_range, 0.5) // 50th percentile (median)
=PERCENTILE(data_range, 0.9) // 90th percentile
=PERCENTILE(data_range, 0.95) // 95th percentileA Note on Learning This the Hard Way.
I've been tracking percentiles consistently for more than a decade now. Before that, like most operators, I relied heavily on averages and got burned by the hidden problems they don't reveal.
The shift to percentile thinking changed how I approach operations entirely. It's not about being smarter—it's about having better visibility into what's actually happening.
Here's the interesting part: This is one of the most frequent things I have to teach everyone from C-level executives to frontline associates. There's always pushback—they think percentiles are complex math, some esoteric statistical technique.
They're not. This is literally 8th grade math. We just never applied it to business problems, so it feels unfamiliar.
The core insight: Your reputation gets built on your edge cases, not your averages. The customer experiencing your 95th percentile performance tells a very different story about your business than your dashboard average.
What metrics are you tracking that might be hiding problems? Let me know in comments.
~Discovering Turiya@work@life
