Articles on Optimization Loop.

Measurement, A/B testing, retraining — a closed loop that makes the system better every month instead of decaying over time.

Optimization LoopWebsite speedOptimization LoopMeasurement

Fast on your device, slow on your customer's phone: measure with real-user numbers before you fix anything

A business owner opens their website on a new phone over office Wi-Fi and sees the page in a second and a half. A customer on a mid-range phone over 4G can wait twice as long. Google judges speed by the 75th percentile of real users, not by what the site owner sees. This article explains the three Core Web Vitals in plain language, why a Lighthouse score is not one of them, three calculations you can do yourself, and six steps to measure, fix and measure again.

18/10/202615 min read

Optimization LoopMeasurementOptimisationStatistics

Before running an A/B test, work out one number — for most Vietnamese sites it exceeds a year of traffic

You do not have to run an A/B test to find out whether it can conclude anything. One calculation, about a minute's work, gives the minimum number of visitors required. For a site with 500 visitors a day, detecting a 10% improvement needs 322 days — and the habit of checking results daily pushes the false-win rate from 4% to 20%.

13/09/202611 min read

Optimization LoopCloudInfrastructure costOptimisation

When cloud prices stop falling: 'ship now, optimise later' just expired

On 4 January 2026, AWS quietly raised H200 GPU prices by 15% — ending two decades in which compute prices moved in one direction only. That does more than make the bill larger. It breaks the unspoken assumption that has bailed out every wasteful architecture for fifteen years: ship it, prices will fall.

15/08/202613 min read

Optimization LoopMonitoringAlert PipelineAnomaly DetectionSRE

Designing an alert pipeline that avoids alert fatigue: from static thresholds to anomaly detection

An alert system with thresholds set too sensitively will send dozens of notifications a day — the team gradually ignores them, and real alerts get buried in the noise. Thresholds set too high miss incidents until it's too late. A three-layer alert pipeline — static threshold, dynamic baseline, and anomaly detection — combined with routing and escalation policies to deliver the right alert to the right person at the right time.

20/06/20268 min read

Optimization LoopAI & MLModel MonitoringData DriftMLOps

How production AI models degrade — and how to detect it before it's too late

After deployment, a model keeps receiving inputs and returning predictions — but the world around it changes. Data drift and concept drift are the two mechanisms by which a once-accurate model gradually loses reliability without producing a single log error. How to detect degradation early, design a three-layer monitoring pipeline, and decide when to retrain versus fix the root cause.

17/06/20268 min read

← See all articles