superpowers-sage:wp-performance

WordPress Performance Optimization

When to Use

• Page load times are slow or degrading
• Core Web Vitals scores are poor (LCP > 2.5s, INP > 200ms, CLS > 0.1)
• Database queries are excessive or slow (QM shows > 50 queries or queries > 50ms)
• Object cache hit rates are low (< 90%)
• Autoloaded options table is bloated (> 500KB total)
• N+1 query patterns detected in Eloquent models
• Hook execution is slow or excessive
• REST API responses are slow

Scripts

# Dump Query Monitor data (pass optional URL, defaults to <dirname>.lndo.site)
bash skills/wp-performance/scripts/query-monitor-dump.sh
bash skills/wp-performance/scripts/query-monitor-dump.sh https://mysite.lndo.site

# Audit autoloaded options (sorted by size, top 20)
bash skills/wp-performance/scripts/autoload-audit.sh

Scripts: scripts/query-monitor-dump.sh · scripts/autoload-audit.sh

---

Procedure

Step 1 — Install Query Monitor

Always profile before optimizing. Never optimize blind.

lando wp plugin install query-monitor --activate

If the db.php symlink fails (permissions):

lando ssh -c "ln -sf /app/wp/wp-content/plugins/query-monitor/wp-content/db.php /app/wp/wp-content/db.php"

Key constants (Bedrock config/application.php):

Config::define('QM_DB_EXPENSIVE', 0.05);   // seconds threshold for slow queries
Config::define('QM_ENABLE_CAPS_PANEL', true);

See references/query-monitor.md for all panels, QM timers, custom logging, and how to export data as JSON.

---

Step 2 — Identify the Bottleneck

Follow this isolation sequence:

1. Bootstrap — plugins_loaded, Acorn boot, autoloaded options
2. Main query — WP_Query execution time, parameter validation
3. Template rendering — Blade, view composers, ACF block with() queries
4. Frontend — Vite bundle size, image loading, paint timing

QM toolbar shows: page generation time, peak memory, SQL duration, query count.

---

Step 3 — Fix the Top Issues

N+1 Queries (most common)

Enable lazy loading prevention in development:

// app/Providers/AppServiceProvider.php
Model::preventLazyLoading(! app()->isProduction());

Fix with eager loading:

// Bad (N+1)
$posts = Post::all();
foreach ($posts as $post) { echo $post->author->name; }

// Good (2 queries total)
$posts = Post::with('author')->get();

For native WP meta loops:

update_post_meta_cache($postIds);  // batch-load all meta in one query
foreach ($postIds as $id) {
    $price = get_post_meta($id, 'price', true);  // served from cache
}

See references/n-plus-one.md for IN query batching, ACF block patterns, and full detection workflow.

Autoloaded Options Bloat

bash skills/wp-performance/scripts/autoload-audit.sh

Fix oversized options:

lando wp option update <option_name> --autoload=no

Target: total autoloaded size < 500KB.

See references/autoload.md for audit workflow, common plugin culprits, and when to set autoload=no.

Object Cache / Redis

lando redis-cli ping   # → PONG

Monitor hit rates in QM's Object Cache panel. Target: > 90%.

$posts = wp_cache_get("category-posts-{$id}", 'my-plugin');
if (false === $posts) {
    $posts = get_posts(['category' => $id, 'posts_per_page' => 10]);
    wp_cache_set("category-posts-{$id}", $posts, 'my-plugin', HOUR_IN_SECONDS);
}

Every cache write requires a documented invalidation strategy. Bust on save_post, edited_term, or the relevant update hook.

See references/caching.md for transients, Cache::remember(), full-page caching, and group invalidation.

Database Query Optimization

• Add indexes for columns in WHERE, ORDER BY, JOIN
• Always use $wpdb->prepare() for raw queries
• Avoid SELECT * — specify columns
• Use Eloquent query scopes to encapsulate reusable logic
• $wpdb->show_errors() in development only — never in production

Vite Bundle Size

lando theme-build -- --analyze   # open bundle analyser

• Dynamic imports: const m = await import('./heavy-module.js')
• Split large vendor libraries via manualChunks in vite.config.js
• Target: no single chunk > 200KB gzipped

Core Web Vitals

Metric	Target	Common Sage causes
LCP	< 2.5s	Large hero image, render-blocking CSS, slow TTFB
INP	< 200ms	Heavy JS bundle, long main-thread tasks, Livewire hydration
CLS	< 0.1	Images without dimensions, dynamically injected content

Always declare width and height on images. Preload the LCP image:

<link rel="preload" as="image" href="{{ $heroImageUrl }}" fetchpriority="high">

See references/core-web-vitals.md for full LCP/INP/CLS remediation, critical CSS with Vite, and third-party script strategies.

---

Step 4 — Deep Profiling (if bottleneck unclear)

Use QM timers for targeted code sections:

do_action('qm/start', 'my-operation');
$result = $this->expensiveComputation();
do_action('qm/stop', 'my-operation');

For call-graph analysis, use Xdebug profiler in Lando:

# .lando.yml
services:
  appserver:
    xdebug: profile

Trigger: append ?XDEBUG_PROFILE=1 to the URL.

See references/profiling.md for Xdebug setup, microtime() patterns, and WP hook profiling (pre_get_posts, save_post).

---

Cron: Action Scheduler vs wp-cron

• wp-cron: triggered by page visits, unreliable under low traffic
• Action Scheduler: queue-based, retries, logging, concurrent processing
• Development: lando wp cron event run --due-now
• Production: disable wp-cron (DISABLE_WP_CRON=true) and use real system cron

---

Critical Rules

1. Check Query Monitor before optimizing. Profile first — never optimize blind.
2. Benchmark before and after every change. Record QM query count, page time, and memory before the fix; verify improvement after.
3. $wpdb->show_errors() only in development. Never leave error display enabled in staging or production — it leaks sensitive schema information.
4. Every cache write requires an invalidation strategy. Document the hook and timing for every wp_cache_set, set_transient, or Cache::remember().
5. Do not use $wpdb->show_errors() in production. Use QM logging instead.

---

Verification

• QM shows no duplicate queries and no queries > 50ms
• Autoloaded options total size < 500KB
• No N+1 patterns (Model::preventLazyLoading() throws no exceptions)
• Redis object cache hit rate > 90%
• Core Web Vitals pass: LCP < 2.5s, INP < 200ms, CLS < 0.1
• Page load time improved from baseline measurement
• Vite bundle sizes reasonable (no single chunk > 200KB gzipped)
• QM PHP Errors panel clean — no warnings or deprecations from theme
• QM HTTP API Calls panel shows no timeouts or errors

Failure modes

Symptom	Likely cause	Fix
TTFB > 1s	No object cache or slow queries	Enable Redis; audit autoloaded options
Query count > 50 per page	N+1 or missing eager loading	->with(), update_post_meta_cache()
LCP > 4s	Large unoptimized hero image	Compress, WebP, preload hint
CLS > 0.25	Images without explicit dimensions	Add width/height
Redis connection refused	Redis not running in Lando	Check .lando.yml, lando rebuild
Object cache not working	Drop-in not installed	Cross-ref acorn-redis skill
Stale cached data	No invalidation on update	Add wp_cache_delete to save hooks
QM toolbar not visible	Not logged in as admin	Set QM auth cookie via Settings
QM db.php symlink missing	Permissions issue	Manually symlink (see Step 1)
QM panels empty	QM_DISABLED is true	Check config/application.php constants
Slow hooks detected	Expensive callbacks on frequent hooks	Queue async; use QM timers to measure

Escalation

• Performance issues persist after all fixes → profile with Xdebug or Blackfire
• Database bottlenecks beyond query optimization → DBA for schema/read replica
• CWV blocked by third-party scripts → document constraint, escalate to project lead
• Infrastructure-level issues (server response time, CDN) → hosting/DevOps