A retail company runs dozens of App Service web apps whose CPU and request volume vary widely by day of week and time of day, with heavy seasonal spikes. The operations team currently uses static metric alerts, but they generate excessive false positives during predictable peaks while missing genuine anomalies during low-traffic periods. The architect must recommend a monitoring approach that automatically adapts alert thresholds to each app's historical behavior without requiring manual tuning per app. What should the architect recommend?