What Kinsta APM records
Kinsta's APM records timestamped data on PHP processes, MySQL queries, external HTTP calls and Redis calls for one site. It groups requests into transactions by endpoint, such as /wp-cron.php or a /wp-json route. For each transaction it keeps up to three samples: the slowest, the 95th percentile and the median (50th percentile). Each sample has a trace, a timeline of spans showing where the time went, and each span links to a stack trace down to the file and line.
Four tabs organise the data. Transactions shows overall transaction time, split into PHP, MySQL, Redis and External, plus the ten slowest endpoints. WordPress lists the slowest plugins by slug and the slowest hooks, marked as actions or filters. Database lists the ten slowest queries and, only if the Redis add-on is enabled, the slowest Redis calls. External lists slow outgoing requests the server makes to other services, each with its URL and HTTP method.
APM only sees work PHP does. Kinsta defines a request as a visit that needs PHP to render, so anonymous pages served from cache will not explain a slowdown. That is useful for course sites, because the pages that hurt, such as logged-in lessons, quizzes, progress updates and checkout, are the ones that run PHP.
Settings, limits and retention
| Setting | What Kinsta documents |
|---|---|
| Cost | Included at no additional charge on all plans |
| Monitoring durations | 2, 4, 12 or 24 hours; turns off automatically at the end, or can be disabled early |
| Viewing time frames | 30 minutes, 60 minutes (default), 2, 4, 12 or 24 hours, plus an Absolute range |
| Data retention | 14 days |
| Warm-up | Allow 5 to 10 minutes after enabling before reading data; data does not refresh automatically |
| Samples per endpoint | Slowest, 95th percentile and 50th percentile |
| Span highlighting | Orange above 5% of the sample's duration, red above 25% |
| What is not recorded | MySQL and Redis spans under 0.001 ms; other spans under 1 ms |
| Tool conflicts | Kinsta advises against running it alongside Blackfire or New Relic monitoring |
The performance cost and when to turn it on
Kinsta says the APM agent might add load to the server's CPU and RAM and should not be left on for an extended period. It also advises checking that the site still works after enabling it. Treat it as a diagnostic session, not a background monitor.
Pick the shortest window that covers a reproducible slowdown. If students report that quiz submissions hang in the evening, enable 2 hours during that period rather than 24 hours across a whole day. If you can reproduce the problem yourself, enable 2 hours, wait the 5 to 10 minutes Kinsta recommends, then run the slow action ten or more times: submit a test quiz, mark lessons complete, run a test-mode checkout. More samples make the 95th-percentile and median figures meaningful.
Because data is kept for 14 days and you can set an Absolute time range, you can come back later and narrow the view to the 15 minutes around a specific complaint.
Reading a slow quiz or checkout transaction
Start on the Transactions tab. Find the endpoint your action calls in the slowest-transactions list; for a quiz or progress update this is often an admin-ajax.php or /wp-json request, depending on the LMS. Open its samples and compare the slowest sample with the median. If the median is fast and the slowest is many times slower, the problem is intermittent, which points to an outside call, a lock or a scheduled job colliding with the request. If the median itself is slow, every request pays the cost.
Open the slowest sample and sort the spans by Duration (%). A red span takes more than a quarter of that request. Click it for the stack trace and note the plugin folder in the file path. Check the HTTP result as well: 200 means slow but successful, and Kinsta notes a 503 may mean the request timed out.
Then cross-check the other tabs. On WordPress, a plugin with a high Total Duration percentage and high average and maximum durations is a suspect; Kinsta suggests confirming by opening its samples, or by deactivating the plugin and checking APM again once enough data has been collected; doing that on a staging copy keeps live students unaffected. On Database, look for one query that dominates or the same query repeated. On External, look for a payment, email-marketing, licence or analytics API that the request waits on.
Illustrative example, not a measurement: a checkout sample takes 4,000 ms. One External span, a call to an email-marketing API to tag the new buyer, takes 2,800 ms and shows red at 70%. More PHP threads or memory would not shorten that request at all. Moving the tagging to a background job, or adding a short timeout, would.
What the trace points to, and what to do before upgrading
| Pattern in APM | Likely meaning | Try before paying for more resources |
|---|---|---|
| One External span dominates | The request waits on a third-party API | Move the call to a background task, add a timeout, or check the vendor's status; a bigger plan will not help |
| One slow query in the Database tab | An unindexed or broad query, often a plugin report or meta lookup | Update or reconfigure the plugin, ask its developer about the query, or limit the report's date range |
| Many fast, repeated queries | Work that an object cache could serve | Consider Kinsta's Redis add-on ($100 per site per month, 2GB) only after APM shows the repetition |
| One plugin high in the WordPress tab | A plugin hook runs on every request | Check its settings, test without it on staging, look for a lighter alternative |
| Transactions fast but users still wait | Requests queue for PHP threads rather than run slowly | Check the PHP thread limit chart under Analytics, Performance before buying memory |
| 503 results on samples | Requests may be timing out | Find the span that runs long; timeouts usually have a single cause |
When paying for more is the right answer
If APM shows requests are individually fast but the PHP thread chart keeps hitting its maximum, you are short of capacity, not efficiency. Kinsta notes that frequently reaching the maximum number of PHP threads may indicate performance issues and that dynamic sites benefit more from threads than static ones. Kinsta sells extra memory pool at $10 per 500MB per month, and memory per thread is the pool divided by the number of threads; it advises keeping at least 128MB per thread for WooCommerce. Standard plans with a default pool of 4GB or more cannot be increased.
If the Memory limit reached indicator shows Allowed memory size errors, visitors on those requests see a 500 error. That is a separate signal from slowness and is worth fixing first.
Who should use it and who can skip it
Use it if you host an LMS, membership or WooCommerce site on Kinsta and are about to upgrade because pages feel slow for logged-in users. A two-hour session costs nothing extra and can tell you whether the upgrade would help.
Skip it if the complaint is about anonymous, cached pages; APM records PHP work and will show little. Skip it, too, if you already run New Relic or Blackfire on the site and do not want to switch them off, since Kinsta advises against running APM alongside them.
Checklist for an APM session
- Write down the slow action, the page and the time it happens.
- Disable other APM tools on the site first.
- Enable APM for the shortest window that covers the slowdown, then confirm the site still works.
- Wait 5 to 10 minutes, then reproduce the action at least ten times in test mode.
- Note the endpoint, median and slowest durations, and the red span's file path.
- Check WordPress, Database and External tabs for the same culprit.
- Fix or test the fix on staging, then run a second session to compare.
- Only then decide on Redis, more memory or a bigger plan.
Questions before you decide
Does Kinsta APM cost extra?
No. Kinsta says APM is included at no additional charge on all plans. The cost is possible extra CPU and RAM load while it runs, which is why it switches off after 2, 4, 12 or 24 hours.
How long does Kinsta keep APM data?
Kinsta retains APM data for 14 days. You can view it in windows from 30 minutes to 24 hours, or set an Absolute range to focus on the minutes around a specific incident.
Will Kinsta APM show why cached pages are slow?
Not really. APM records work done by PHP, such as transactions, queries and external calls. Pages served from cache do little PHP work, so focus it on logged-in lessons, quizzes and checkout.
Sources & verification
Product details and prices can change. Check the linked provider before buying.
- Kinsta: APM tool documentation Accessed 30 Oct 2026
- Kinsta: WordPress PHP performance and threads Accessed 30 Oct 2026
- Kinsta: Redis cache add-on Accessed 30 Oct 2026
- Kinsta: Site caching Accessed 30 Oct 2026
Sources link directly to providers. Product buttons may use separately labeled affiliate links. Read our disclosure.