What is good performance?

 Veröffentlicht Februar, 2026 | 5 Min reading time

Good performance is not a fixed number.
It occurs when response times, context, and user expectations align.

Performance is perception

There are no standardized indicators to tell good performance from bad performance. It are the user’s perception and expectation that determine performance quality. Customers shopping online want to find products, move them into their cart, and finalize the order. This process should be carried out without any unexpected delays. If this is the case, the shopping platform will be seen as performant. Does it take some time until the cart is up to date, or is the checkout unduly delayed, then the system is perceived as having performance problems.

Same measurement variable, different evaluation

These observations are valid for almost every virtual interaction. If a text-editor takes one second to react to a keystroke, it is very likely to be seen as “slow” and having bad performance. If it takes a website one second to load, it will be seen as acceptable and no performance complaints will be issued. Does it take one second to search an entire database, the system will be seen as surprisingly performant.

el.

While the amount of time for these examples has not changed, the appraisal by the users has changed completely. Performance therefore does not consist of short response times but of the interplay of context, expectation and the user’s goal.

Since this definition of performance makes it next to impossible to actually measure it, psychologists have delivered some concrete numbers which we can use to approximate the duration of good performance in different contexts: 

Response time > 15 seconds

Interaktive Nutzung ist nicht möglich.
Solche Verzögerungen sind für Konversationen ungeeignet, können aber für bestimmte Anwendungsfälle akzeptabel sein.
Wichtig ist, dass das Design der Anwendung dies berücksichtigt und Nutzer andere Aufgaben ausführen können.

4–15 seconds

Conversations are heavily inhibited.

Users have to keep information within their short time memory, reducing their problem-solving-capabilities. Entering Data is frustrating. When ending a transaction, such response times are acceptable.

2–4 seconds

Operations needing a high level of concentration are nearly impossible.
Activities like comparing different products and options are straining. These response times are acceptable for clearly defined actions, but not for continuous interaction.

1–2 seconds

Necessary for coherent cognitive tasks.

If users need to process or compare information across multiple responses, the response time should be less than two seconds. This limit is particularly crucial for complex navigation or comparison tasks.

0,1–1 second

Necessary for mental taxing activities.

Some examples are text-writing, working with graphic interfaces, or manipulating visual elements. Short response times are necessary for thoughts to be implemented without interruption so that the workflow can be maintained.

< 0,1 second

Nearly instantaneous reaction.

Keyboard inputs and mouse clicks must be within this range to be perceived as direct. This is particularly relevant for computer games and other highly interactive applications.

While this list might appear persuasive, it is important to remember that real performance is dependent on the users and the context they use the software in.

Quelle: Ian Molyneaux, The Art of Application Performance Testing, 2nd Edition, 2014