the question is bit unclear to me. My understanding is you want to do measure task completion for sample size of 20, and lest assume you have 10 tasks fro participants:
There are two main way in usability testing:
For moderated, you need first recruit participants based on your persona. You could do this online, or face-to-face. To do this online you need some tools like Validately, user zoom, … or even you can use Skype or google hang out. The reason I mentioned special tools like user zoom, in this tools you can define tasks, and the tool measure time on task automatically. However to do this in cost effective way you can do this with other free software, and ask some one else from your team to join the session( notetaker), and with stop watch record the Time on tasks.
For moderated, Face-to face, you could ask participants to come to your office, and do the test, and measure the task. Probably you don’t have room with two way mirror, but still you could book a meeting room and again ask your colleague to take note, an recored the TOT. Note, first explain role of note taker for participants, as they would feel overwhelmed to see two person in the room, second only except you and note taker, don’t any one else be in the room, you could record the session, or if you have stockholder to be involve they should watch from live recording, or pre record later.
For unmoderated, there some website like usertetsing.com, that you could upload your prototype. This website has been designed for this, you upload the prototype, and then design the tasks. They will show this to the number of people that you would ask and they would come back with the results.
You could do this with user zoom too.
Last but not least, I don’t recommend Guerrilla testing at all for time on time on task. There are may factors that could effect users effectiveness, on finishing a task. Unless you are doing a test that related to that context, example. You want to design something that people use in airport and you go to airport and see how all those factors in public effect finishing a task ( mix of contextual enquiry + ToT). But if you are designing a website, app,… that people use 90% in the office, home,… you should do this task under control environment.