Percentiles in Performance Testing

by Sujay Prabhu, Senior System Analyst

The page load time is an important aspect of the website user experience. The faster the page loads, the happier the users will be. There are various reasons for slow web page loads. Performance monitoring and testing of API endpoints is one way of understanding how websites and services behave when hundreds or thousands of users access them.

What is Percentile ?

Percentile is the common metric in case of Performance testing. If a backend service's 95 percentile is 30ms, it means 95% of requests have response time of less than 30ms.

How to calculate Percentile ?

Here is the response time for one of the endpoint:

RESPONSE_TIMES

Sort the data based on response times

SORTED_RESPONSE_TIMES

Apply the formula

index = (n * (90/100)) where n = Number of requests
index = (10 * 0.9)
index = 9

In the above table, 9th index value is 4.6

So, 90 percentile for the above set of response times is 4.6

Here I am calculating 90 percentile for fewer data where as in real world applications this number can be in thousands or millions.

What about average response time ?

Average response time is also one of the metrics of performance monitoring and as the name says it is the average value of the response times. This can occasionally be misleading if fewer requests have very high response times.

Consider the above set of records and if the response times were as mentioned below instead of 5.1 and 4.6

AVERAGE_RESPONSE_TIMES

Average response time is 4.34s even though majority of the response time hovers between 2 and 3. Thus with average response time metric, fewer requests can bring difference in how the graph looks

More articles

Agentic Memory: A Practical Guide to How AI Agents Remember

A practical guide to agentic memory, covering working, semantic, episodic, and procedural memory, along with storage, retrieval, distillation, and memory evolution.

Read more

What is `llms.txt`, Why It Matters, and How to Use It

What llms.txt is, how it fits into Agentic Engine Optimization (AEO), how it differs from robots.txt and sitemap.xml, and how to add one to your site.

Read more

Your competitors are already using AI.
The question is how fast you want to unlock the value.

Don't know where to start?

AI is everywhere but it's unclear which investments will actually move your metrics and which are expensive experiments.

Your data isn't ready

Most AI projects fail at the data layer. Pipelines, quality, access all need work before LLMs can deliver value.

Internal teams are stretched

Your engineers are shipping product. They don't have capacity to also become AI specialists with production-grade experience.

Legacy systems block everything

Aging, undocumented codebases make AI integration slow, risky, and expensive. They need to move first.

Don't worry. We've got you covered.

Start with the audit.