Static generation or server rendering.
With static generation the HTML is prepared at build time; with server rendering, on every request. The choice rests on how often content changes.
The decision looks like a technology preference, but it is really a question about content: how often does the page change, and is it different for every visitor? The answers to those two questions determine the architecture; speed is a consequence, not a cause.
What static generation really buys
On a static page the only thing the server does is send a file. No database query, no template compilation, no application start-up. The time to first byte is low for that reason and, more importantly, PREDICTABLE: it does not change when traffic rises.
The second gain is a narrow failure surface. With no running application server there is no application to crash, no connection pool to exhaust and no runtime to patch. The measured time to first byte for this site and the three projects it manages is under 300 milliseconds, and that is the result of the architecture rather than of an optimisation effort.
What server rendering really buys
Personalisation cannot be done at page level with static generation. A user-specific dashboard, a session-bound cart or a list that varies by permission has to be produced on every request. Filling it in from the client is possible, but then the content is absent from the first paint — and invisible to a search engine as well.
The second gain is freshness. For data that changes by the minute — stock levels, a queue number, a broadcast schedule — the build step is a source of delay. Here the question to ask is “how costly is stale content”.
Which one this site uses, and why
We generate this site entirely statically: the content changes rarely, no personalisation is needed, and we meet the budget targets without a plugin layer. The only dynamic endpoint is the contact form — and it runs as a separate function, without affecting the pages. We cover the full list of trade-offs and time to first byte in separate pieces.
If you want your own site measured and fixed at the source, our Core Web Vitals optimization starts from real-user data and delivers LCP, INP and CLS against a budget.
Criterion
Static generation
When the HTML is produced At build time, once
Time to first byte Serving a file — typically the lowest
Personalisation None at page level — done on the client
Content freshness Requires a new build
Server rendering
When the HTML is produced On every request
Time to first byte Production time is added; caching can reduce it
Personalisation Natural: different content per request
Content freshness Immediate
Static generation
Scaling Served from the edge cache, server load stays flat
Infrastructure Static file hosting is enough
Failure surface Narrow — the file to serve already exists
Where it fits Content that changes rarely: marketing, portfolio, documentation
Server rendering
Scaling Server load grows with traffic
Infrastructure A running application server is required
Failure surface Wide — database, application and cache layers
Where it fits Content that changes often or is user-specific: dashboard, cart, feed
| Criterion | Static generation | Server rendering |
|---|---|---|
| When the HTML is produced | At build time, once | On every request |
| Time to first byte | Serving a file — typically the lowest | Production time is added; caching can reduce it |
| Personalisation | None at page level — done on the client | Natural: different content per request |
| Content freshness | Requires a new build | Immediate |
| Scaling | Served from the edge cache, server load stays flat | Server load grows with traffic |
| Infrastructure | Static file hosting is enough | A running application server is required |
| Failure surface | Narrow — the file to serve already exists | Wide — database, application and cache layers |
| Where it fits | Content that changes rarely: marketing, portfolio, documentation | Content that changes often or is user-specific: dashboard, cart, feed |
FREQUENTLY ASKED QUESTIONS
Is a static site always faster?
In time to first byte, typically yes, because there is no production step. But a well-cached server-rendered site can be fast too; the difference lies in the cache hit rate, not the architecture.
If the content changes often, is static out of the question?
It can still be used, but the cost rises: every change means a build. For a publication that adds content a few times a day that is reasonable; for a feed updating every minute it is not.
Can the two be combined?
They can, and often are. Most pages are generated statically, and only the parts that need personalisation are filled in from the client or an endpoint. The contact form on this site works exactly that way.