Separate reading from interaction
Write a small functional brief before selecting technology. A price comparison might provide immediately readable text and a table, with an optional filter for exploration. In a configurator, several choices and a calculated result might be the core task. These needs should not receive identical treatment simply because both are called components.
Astro's documentation distinguishes rendered content from framework components hydrated through client directives. Treat this as a possible boundary for interactivity. It does not establish how fast a particular page will be. Other scripts, third-party services, images and data still affect the actual experience.
Define a contract for the waiting state
Describe what the visitor sees before the client code is ready and what they can do then. A button that looks ready but does not respond creates uncertainty. Can the function wait until needed? If so, the waiting presentation still needs to make sense.
Consider an illustrative product filter: a usable product list appears before activation; afterwards the visitor can narrow the selection. Hiding the entire range until a script succeeds creates a different accessibility and failure profile. Document and test that choice rather than assuming deferred activation is automatically better.
| Part | Decision before implementation |
|---|---|
| Core content | What remains readable before activation? |
| Start condition | When does the visitor actually need the function? |
| Waiting state | What is visible and usable during the wait? |
| Failure route | How can the visitor continue if activation fails? |
Ask the developer to test a bounded option
Request a pilot with one component and a clear user task. The developer chooses an appropriate implementation and checks support in the project's Astro version. Not every interface detail requires a client framework: ordinary HTML, CSS or a small script may be sufficient.
Use a test card covering a slow connection, keyboard and mobile. Read the core content before activation, then attempt the task. Note unexpected focus movement, shifting content or lost choices. The test should examine real behaviour rather than only the presence of a particular source attribute.
Compare function and cost together
Measure transferred client code and observe when the feature becomes usable under documented conditions. Compare alternatives on the same page with the same material. A small script measurement is insufficient if the visitor cannot complete the task. A working component also provides no reason to ship its code to unrelated pages.
Save the decision alongside the feature: task, core content, activation and known limitation. Test again when its purpose changes. This process supports choosing interactivity, not promising a particular Core Web Vitals result. Performance assessment also needs measurements that distinguish controlled laboratory observations from real usage.