Pagemeter

Your site is slow, or it keeps hitting its PHP worker limit, and you have no idea which plugin is responsible. Most tools are either developer debug bars too heavy to leave on, expensive services that do not run on shared hosting, or lab benchmarks of a plugin on a clean site rather than your site. Pagemeter is a simple, on-demand speed test for your own site. Click Run test and it loads your front page a few times, both cached (what real visitors get) and uncached (a genuine cache miss), then shows you one screen with:

• Runtime by plugin - your plugins ranked by how much PHP time their code spends building a page, so you can see what is actually heavy.
• Caching impact - how much faster caching makes your site, measured rather than guessed. Caching and optimization plugins are recognized and never framed as "slow"; their cost on a cache miss is exactly what buys your visitors a fast, cached page.
• Page performance - the page build time split into Load, Setup, and Render, with peak memory and database query count.
• Outbound HTTP by plugin - every external service your plugins contact during the build, and how slow or error-prone each one is. Dead third-party APIs that quietly time out are a common, invisible cause of a slow site.

Every test is saved, so each time you optimize, remove, or reconfigure a plugin you can watch the numbers drop and prove the improvement. Nothing is recorded except during a test you start yourself, so there is zero measurement overhead on normal visitor traffic. It also includes an optional domain governor: a blocklist you configure to stop a dead or unwanted endpoint from slowing your site. Nothing is blocked by default. Privacy Pagemeter measures locally and stores everything in your own database. It does not contact any external server and contains no telemetry. No data about your site or your plugins leaves your site. What it does not do

• It does not name or rank plugins for anyone but you, on your own screen.
• It does not block any outbound request unless you add that domain to your blocklist yourself.

1.0.2

• Output-escaping hardening: every value shown in the report is now escaped at the point of output (escaping late), and the report markup is passed through wp_kses() on the way out. Passes Plugin Check and WordPress Coding Standards with no notices. No functional or visible changes.

1.0.1

• Finalized the plugin name as Pagemeter and corrected the author, author URI, and plugin URI to match the WordPress.org account. Naming and metadata only; no functional or visible changes.

1.0.0

• First stable release. On-demand plugin speed testing on your own site: runtime by plugin, cached-versus-uncached caching impact, a page-build-time breakdown, and outbound-HTTP-by-plugin attribution, with saved before/after history so you can prove each improvement. Read-only and fail-safe: it records nothing outside a test you start, and can never take your site down.

0.12.9

• Uninstalling now removes everything: all three data tables (previously only one was dropped) plus the test-history and last-test options and the admin-bar cache.
• Hardened input handling on the settings form and the internal test probe, and brought the whole codebase to a clean WordPress Coding Standards run.

0.12.8

• The runtime and per-plugin trend badges no longer show a percentage on very small values (under 10ms), where sampling jitter of a millisecond or two would otherwise read as a large, misleading change (e.g. "90% slower"). Trends now appear only where they are meaningful.

0.12.7

• Test history now names the heaviest plugin by runtime when a page makes no external HTTP calls, instead of showing "n/a" in the Slowest plugin column.
• The caching headline now explains that the cache-miss figure is a full round-trip (network included), while the Page performance section measures WordPress's internal build time, so the two numbers no longer look contradictory.

0.12.6

• Fix: a test on a page that makes no external HTTP calls showed "Run your first test" even though it had measured page-performance and per-plugin runtime. The report now recognizes any recorded data, and leads with the page build time when there is no outbound HTTP to headline. The HTTP stat cards are hidden when a test recorded no external calls, instead of showing zeros.

0.12.5

• Critical fix: the signed test token added in 0.12.1 called wp_salt(), which WordPress does not load until after plugin files are included. On any request carrying the test token (every test load) this was a fatal error, so tests always came back empty and the loaded page returned an HTTP 500. The token is now signed from the site's salt constants, which are available at that point in the bootstrap, so tests actually run.
• Safety: plugin startup is now wrapped so Pagemeter can never take the site down. If it fails to initialize for any reason it disables itself for that request and WordPress continues normally.

0.12.4

• The test loads now identify as a normal browser (the visitor they simulate), so security rules and firewalls that block scripted user agents no longer error the loopback.
• When a test comes back empty, the failure banner now quotes the actual response body, so an origin error, firewall block, or plugin fatal on the uncached build is identifiable at a glance. Empty results caused by an HTTP error are now named as such instead of being mislabelled a cache hit.

0.12.3

• Diagnostic: when a test measures nothing, Pagemeter now inspects the actual loopback response and tells you exactly why: the site was answered from cache (PHP never ran), the loopback could not connect, the signed token was not recognized, or the page ran but nothing was recorded. A compact technical line (HTTP status, server, cf-cache status, size, ran/probe flags) is shown for support.

0.12.2

• Fix: tests could come back empty on sites with a persistent object cache. The probe token is now signed (using your site's auth salt) instead of stored in a transient, so it is verified in-process and no longer depends on the object cache being readable across the loopback request.
• The test now measures your site the way a real visitor experiences it: every load is anonymous (not logged in). Cached loads hit the page cache like a visitor and receive static HTML with no PHP; uncached loads add a cache-busting query argument so an anonymous request misses the cache and rebuilds the page, which is exactly what a first visit or a just-cleared cache costs. (The previous build forced a logged-in request to miss the cache, which was not authentic: logged-in requests always run PHP and skip the cache, so they do not reflect what visitors get.)
• New diagnostic: the toolbar now reports how many uncached loads were actually measured, so you can tell at a glance whether the loads reached PHP.

0.12.0

• Caching impact: each test now measures both cached loads (what real visitors get) and uncached loads (a cache miss), and the headline shows how much faster caching makes your site.
• Caching / optimization plugins (AccelerateWP, WP Rocket, LiteSpeed, etc.) are recognized and badged "cache / optimizer" so they are never framed as slow. Their cost in the uncached breakdown is correctly explained as the price of the speedup they provide.
• The per-plugin and runtime breakdowns are now labelled clearly as the uncached (cache-miss) build cost.
• Test history now records cached and uncached times too.

0.11.2

• Replaced the stale "Data retention (days)" setting with "Test history to keep (runs)", which controls the new Test history.

0.11.1

• Test history and before/after: each test is saved, and the headline, per-plugin, page performance, and runtime now show how much faster or slower they are than your previous test. A new "Test history" panel lists recent runs.
• Removed the redundant "Clear stored data" button and the obsolete "Profile every request" setting (each test now replaces the last automatically).
• Tests no longer trigger a cache-preload cascade: probe loads set DONOTCACHEPAGE so page-cache plugins skip them, keeping a test to a clean handful of loads.

0.11.0

• Major change: Pagemeter is now an on-demand tester. Click "Run test" and it loads your site a few times (bypassing cache) to measure the current cost of each plugin right now, instead of averaging over a rolling time window.
• Nothing is recorded except during a test, so there is zero overhead on normal traffic.
• Removed the Live / Last 24 hours / Last 7 days / Last 30 days windows.

0.10.1

• New "Profile every request" setting: records page performance and per-plugin runtime on every request instead of a sample, so the data fills instantly on low-traffic or test sites. Leave off on busy production sites.

0.10.0

• New: Runtime by plugin (sampled). Estimates the PHP time each plugin's code spends running and ranks plugins heaviest-first, so you can see which plugin is heavy even when it makes no external calls. Measured read-only (it never wraps or alters callbacks) on a fraction of requests; it is an estimate, not exact profiling.

0.9.3

• Page-performance sampling: front-end page loads are always recorded, but high-frequency background requests (AJAX, cron) are sampled, so busy sites stay lean.
• Hardened outbound-HTTP timing against cache plugins that short-circuit requests: stale pending entries are now discarded (and the pending buffer is bounded), preventing occasional inflated durations.

0.9.2

• The per-plugin breakdown ("By plugin") is now the first card in the report, above page performance, so the plugins come first.

0.9.1

• Plugins first: the headline now leads with your slowest plugin by name, instead of an aggregate number.
• Re-centered the report on the decision: page performance and the per-plugin breakdown now lead, with the request-context detail moved below the supporting sections.
• Each plugin now shows a plain-language signal ("needs attention", "affects visitors", "slow calls") so you can see at a glance what to optimize or remove.
• Smoother, crisper page-performance trend graph (curved line, uniform scaling, no aliasing).

0.9.0

• New: Page performance. Measures how long WordPress takes to build each page, with a Load / Setup / Render breakdown, peak memory, and database query count, averaged over the window with improvement trends.
• New animated visuals: a self-drawing page-build-time trend sparkline, count-up tiles, and glowing bars (all respecting reduced-motion).

0.8.2

• Added visual bars to the "Where these requests happen" breakdown (total and blocking time on the same scale, so you can see how much of each context actually blocks).
• Moved the loopback and discontinued-service callouts lower on the page, next to the destination data they relate to.

0.8.1

• The report header now shows the real Pagemeter logo mark (gradient) with a subtle radar pulse.

0.8.0

• Refocused on speed, not accumulation. The headline is now the typical time a call takes right now, and it goes down when you optimize (instead of a total that only ever climbs).
• New default "Live (last hour)" view, plus 24 hours / 7 days / 30 days.
• New improvement trend: each plugin (and the headline) shows whether it got faster or slower versus the previous equal period, so removing or reconfiguring a plugin shows up as a real, present win.
• The "By plugin" table now leads with typical call time (sorted slowest first) and a trend column.
• Fixed a timezone drift in the reporting window so short windows are accurate. Toolbar now shows the live average call time.

0.7.0

• The main report now reflects only currently-installed plugins, so totals and tables show accurate current-install performance. WordPress core and must-use plugins are always included.
• New "Previously installed (historical)" section lists removed plugins and their recorded times for before/after comparison, kept separate from the totals.

0.6.0

• New: discontinued-service detection. Destinations that are known-dead services (e.g. AddThis, discontinued 2023) are flagged with a "safe to remove" note in the destination and error tables, plus a callout.
• Admin notices from other plugins now appear in the correct place instead of inside the Pagemeter header.
• Removed plugins are labelled "not installed" (was "unrecognized") so recently-uninstalled plugins read clearly while their history ages out.
• New "Clear stored data" button to wipe recorded history on demand.

0.5.0

• New: request context. Every call is tagged front-end, admin, cron, AJAX, REST, or WP-CLI, so you can tell what actually slows real visitors from harmless background work.
• New "Where these requests happen" breakdown, a "Visitor impact" column per plugin (blocking front-end time), and a headline figure for how much of the total real visitors actually wait for.

0.4.0

• Human-readable plugin names (e.g. "Really Simple SSL" instead of the folder slug), with the slug shown alongside. WordPress's own requests are labelled "WordPress core", and folders that are not a recognized plugin are flagged "unrecognized".
• New "Errors & timeouts" panel: see exactly which plugin failed talking to which destination, why (HTTP status or connection/timeout), how many times, and when. Requests to HTTP 4xx/5xx now count as errors, not just transport failures.
• Admin-bar total now re-syncs when you open the report and refreshes more often.

0.3.0

• New: loopback detection. Outbound requests to your own site are flagged "self / loopback" with a callout, since self-HTTP is slow and can tie up PHP workers.
• Sub-second durations now display in milliseconds (e.g. "26ms") instead of "0.0s".

0.2.0

• New top-level admin menu (Pagemeter) plus an admin-bar shortcut, so the report is easy to find.
• Redesigned report: hero summary, stat cards, and per-plugin / per-destination attribution bars.
• Added count-up and bar-fill animations (respecting prefers-reduced-motion).

0.1.0

• Initial release. Outbound HTTP attribution by plugin, destination report, one-screen summary, and a user-configured domain governor.