There is a lot of ego in this industry. Somewhere along the way the standard advice became that if you build complex WordPress websites, you write your own Nginx config on a raw VPS. I have done it. I have spent nights chasing race conditions in cache layers I wrote myself while the client’s site sat on a 502. That badge of honor costs you money.
Calvin Ho and his agency, Knockers Design, worked this out the hard way. They are based in Taipei and the work they take on is the heavy kind: Web3 integrations, CRM connections, custom API builds. While plenty of devs argue over server margins, Calvin moved the whole stack to WordPress.com. At that level of complexity you do not want another server to babysit. You want infrastructure that holds up while you work.
The real math of self-hosting
Self-hosting is supposed to save money, and plenty of people believe it hands you true control on top of that. Price the refactor time instead. Five hours a month on security patches, backups and transient cleanup is not a saving, it is billable dev time you gave away. Knockers Design stopped asking how to host a project and started asking what they could build with it.
Anyone who read my post on why your WordPress server is failing you knows I take stability over a shiny setup every time. Managed hosting takes the plumbing off your desk and leaves you with the architectural work the client is actually paying for.
Architecture for complex WordPress websites
On custom CRM integrations or Web3 projects, the CPU is rarely the bottleneck. How you handle external data is. Self-hosted, a single hanging API call can tie up your PHP-FPM pool unless the timeouts and transients are set up exactly right. A managed platform like WordPress.com arrives already tuned for that kind of concurrency.
For complex WordPress websites I always want a transient-backed API handler in place. Here is a stripped-down version of how remote data should be fetched so it does not drag the whole site down:
<?php
/**
* Senior-level API Fetcher with Transient Caching
* Prefix: bbioon_
*/
function bbioon_fetch_remote_crm_data( $endpoint ) {
$cache_key = 'crm_data_' . md5( $endpoint );
$cached_data = get_transient( $cache_key );
if ( false !== $cached_data ) {
return $cached_data;
}
$response = wp_remote_get( $endpoint, [
'timeout' => 5, // Never wait forever
'user-agent' => 'KnockersDesign-Integration/1.0',
] );
if ( is_wp_error( $response ) ) {
error_log( 'CRM API Error: ' . $response->get_error_message() );
return [];
}
$body = json_decode( wp_remote_retrieve_body( $response ), true );
// Cache for 15 mins to prevent hitting API rate limits
set_transient( $cache_key, $body, 15 * MINUTE_IN_SECONDS );
return $body;
}Monitoring and GitHub deployments
Calvin’s team runs deployments through GitHub integration. Manual FTP is a legacy habit no senior dev should still be carrying. With automated deployments, every line of custom code is version controlled and tested before it reaches production. That is the gap between an agency that fixes bugs and one that stops them from shipping.
They also use Jetpack for monitoring. I used to write Jetpack off as a blogger plugin, but catching a domain expiration or a downtime spike before the client phones you changes the whole conversation. You stop reacting and start leading.
If this complex WordPress websites work is eating your dev hours, hand it to me. I have been wrestling with WordPress since the 4.x days.
The AI filter
For developers building complex WordPress websites, AI works more like a filter than a threat. Clients turn up with ChatGPT-researched briefs, which means the easy work is already automated. What is left are the hard problems, the kind that take 14 years or more of experience to get right. On a platform like WordPress.com, an agency like Knockers Design stays on those problems instead of on server maintenance.
Their full agency story is on the official WordPress blog. It is a good lesson in picking your battles.
