Framework for collecting feedback from application users via telemetry and targeted surveys.
- Extensible set of data sources for telemetry.
- Full control for the user on what data to contribute.
- Distribute surveys and offer users to participate in them.
- Survey targeting based on telemetry data.
- Allow the user to configure how often they want to participate in surveys.
This framework consists of the following components:
- Libraries for use in applications.
- QML bindings for the above.
- A server application.
- A management and analytics application.
To use this framework in your application, on a high level you have to do the following steps inside your program:
- Create an instance of KUserFeedback::Provider for the entire lifetime of your application, and configure it as described in its documetation.
- Add one or more data sources to the provider, choosing from the standard built-in ones, or by adding custom ones deriving from KUserFeedback::AbstractDataSource.
- Add a way to notify your users about surveys, and to encourage them to contriubute in the first place, by hooking a suitable user interface up with the corresponding signals in UserFeedback::Provider. KUserFeedback::NotificationPopup provides a reference widget for this purpose.
- Add a way for your users to configure what data to contribute, and how often to be bothered by surveys. KUserFeedback::FeedbackConfigDialog provides a reference implementation of a configuration dialog for this.
Frequently Asked Questions
Why does the telemetry data contain no unique user/installation/machine id?
KUserFeedback is designed to be compliant with KDE Telemetry Policy, which forbids the usage of unique identification. If you are using KUserFeedback outside of the scope of that policy, it's of course possible to add a custom data source generating and transmitting a unique id.
How can duplicated submissions be detected without unique identification?
KUserFeedback addresses this by using a fixed time interval for telemetry data submission. If for example Provider::submissionInterval is set to one week, aggregating data in one week intervals ensures that there's only at most one sample included from each installation. The other way around, setting Provider::submissionInterval at least as high as the average usage interval of your application ensures a substantial amount of installations have submitted a sample in this interval. A frequently used application such e.g. an email program can therefore get good results with a smaller value than e.g. a special purpose application.
How can developments of a user be tracked without unique identification?
If you want to observe how usage patterns change as your users gain more experience with the application, there are multiple ways to do that:
- Include a UsageTimeSource, and correlate the relevant usage data with how long the application has been used.
- It is possible to track arbitrary historic development using custom data sources via AbstractDataSource::reset. This method is called after each successful submission, allowing you to implement a "memory" for your data source, and e.g. submit differential rather than absolute values.