Can you try to find the biggest improvements before time runs out with the least effort?
This is a given. By choosing not to run any experiments your impact will be 0% at the end of your testing timeframe. Test to win. :)
Experiments take time to build. You could prebuild tests when you have an experiment running in order to save time and raise your testing velocity. Don't just wait for a test to stop to and begin figuring out what to do next aftwards. Always have the next test ready.
Usually we observe smaller change tests to require less effort, have a smaller impact, and require more time to run.
Usually we observe bigger change tests to require more effort, have a bigger impact, and require less time to run. Bigger change tests might be more effective in the beginning when there are many "high certainty" changes that could be optimized together. The effectiveness of bigger change tests may decay over time.