Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Canvas-based Power Apps provide a flexible, intuitive way to build custom applications. However, performance issues can occur if apps aren't optimized for performance. This article provides practical strategies to improve user experience and resolve common performance concerns.
Set performance targets
Good app performance means different things to different people.
To set standards for app responsiveness, use the following guide, which shows how users perceive response times in applications.
| Response time | User perception |
|---|---|
| 0–100 ms | Instantaneous; the user doesn't notice any delay |
| 100–300 ms | Slight perceptible delay; feels responsive |
| 300 ms–1 sec | Noticeable delay; user perceives slight lag |
| 1–5 sec | Acceptable delay; user remains engaged |
| 5–10 sec | Noticeable wait; user attention may wander |
| 10 sec or more | Significant delay; user may become frustrated or abandon task |
Source: Nielsen Norman Group–Response Times: The 3 Important Limits
Set performance goals early in the development cycle to ensure your solution starts quickly and remains efficient.
Benchmark the current process: If your solution replaces an old app or process, aim to improve on the current experience. If you're replacing an app, make sure your new app loads as quickly as the old one and allows users to accomplish tasks faster. At a minimum, keep app performance the same.
Don't cause new problems: If you're replacing a business process, focus your benchmark on the timing of the full cycle. Ensure your solution removes bottlenecks so users complete tasks faster, and doesn't introduce new bottlenecks in the process. Help users understand how your new solution fits into their existing processes.
Learn more about defining performance targets.
Perform regular performance audits
Run performance tests and checks throughout solution development to catch potential issues early and avoid extensive rework.
Match the production data size: Ensure your development data sources reflect the size and scope of production data as closely as possible. If you can't match production data, test your solution with small datasets to check functionality. Measure performance with different data sizes—like 100, 1,000, and 10,000 rows—and compare results. If performance increases with data size, estimate how the solution performs at production scale and make sure it stays within your agreed parameters.
Track your solution performance over time: Track your performance test results and compare them across test runs. If you run these tests often, trace what changed since the last test and identify where to make fixes. Track significant performance changes, whether positive or negative, and include them in lessons learned or best practices for future solutions.
Test on target devices: Regularly test your app on actual user devices to make sure it performs consistently across different platforms and network conditions.
Learn more about continuous performance optimization.
Reduce data load
Loading large amounts of data can significantly slow down your app.
Limit data retrieval: Loading more data than your user needs at one time can slow down your app and overwhelm the user with unnecessary information. Instead, show a clear, manageable amount of data by default, and let users easily search or filter to find exactly what they're looking for. This approach keeps your app fast, user-friendly, and focused on what's useful.
Avoid delegation warnings: Ensure that the row limit for delegation warnings in canvas app is set low, to five or ten, to flag any data call that can't be delegated in the formula bar. This practice lets the app developer decide whether to allow potentially non-delegated queries. It also ensures that the issue isn't missed due to the development data source not returning enough data to trigger the default warning limit.
Non-delegated queries are acceptable if there's a constant, small amount of data to be returned, such as in configuration data or other small datasets.
Learn more about optimizing your solution's logic.