Optimize the part of the user journey that is actually slow, then measure whether the change helped. A smaller bundle is useful only when it improves the experience or reduces relevant costs.
Suppose customers on ordinary phones open a product page, see its main image late and experience a delay when choosing a variant. Loading and interaction are separate problems; one optimization may not fix both.
1. Establish the symptom with representative measurements
Choose a journey, device and network condition that matter to the product. Use real-user measurements to find the affected population and controlled browser tests to reproduce the problem. A powerful development laptop can hide delays experienced by customers.
