John Toby Idowu

Most SREs don't understand key metrics like P50, P95, and P99. Let me explain it with pizza. You order pizza from Domino's. They promise 30-minute delivery. They deliver 5,000 orders a day. P50 means 2,500 pizzas arrived in 30 minutes or less. Half the customers got the promise. Half waited longer. This is your typical experience. P95 means 4,750 pizzas arrived in 45 minutes or less. But 250 customers had a worse experience than that. Pizza is getting cold. Nobody called them. P99 means 4,950 pizzas arrived in 60 minutes or less. But 50 customers got cold pizza, a ruined evening, and a reason they moved to Pizza Hut. That is the problem with averages. Your dashboard is green. Your average looks healthy. But 50 real people had a bad experience today. And tomorrow. And the day after. This is exactly what happens in production. Your API average response time shows 80ms. Looks great. But your P99 is sitting at 3 seconds for a small group of users. They are not complaining loudly. They are just quietly leaving. At 1 million requests a day, P99 means 10,000 requests are getting cold pizza. Every single day. Silently. This is why senior engineers never trust averages alone. They look at P95 and P99 first.