Definition:
- We want to find out how long (median) it takes for a Wikidata item and property to be served to the user i.e. the time that goes from the user request till when it gets sent to user, in whichever device the user is at the moment.
Frequency:
- the metric will be retrieved once a week
Purpose:
- Set a baseline of the current state
- Detect long loading times - then investigate where the issue comes from e.g. a bad query, big sized item, too high complexity in the code, too many redirects... It will also be easily caught in case a newly released fetaure is having an impact on this metric.
- In the case that the change of trend can easily lead back to a new release the looking for the root cause will be made easier e.g. the scope will be reduced
- Plus it will give our Product team some more data that may be relevant for their short, mid and long term strategy and planning.
Notes:
- we could already do the dictinction whether the item is requested and served to a mobile device or to a desktop browser. The team working in this will make an estimation of the added effort that such distinction would involve. With that information it will be decided whether we have the MVP with an unique value or not.
- we have firstly only contemplated the case of successful responses thinking that the error cases will be covered in another metric (T274420). It will be discussed with the team in case this assumption is false.
AC:
- retrieve metric
- the metric is to be seen on the dashboard