How many cores do you have on your search head? That is approximately the number of searches that Splunk can run simultaneously. If you have more than that scheduled, then some searches will wait. Also, you must include the count of ad-hoc searches that logged-in users are running in your total number of searches.
Does your search head meet the recommended size?
Do you have enough indexers? Are the indexers the recommended size?
To make searches go faster, your best investment is almost always to add another indexer. In a large environment, some people have a search head that runs nothing but scheduled searches - this is called a job server - and it prevents logged-in users from impacting the scheduled searches and vice-versa.
But these are just general comments. You really need to do more analysis of what's happening in your environment.
Here are some links:
Hardware capacity planning
Summary of Performance Recommendations
