Before choosing a number, measure the one thing that determines it: how long it takes a customer who converts to get their first useful result.
Go back through your converted customers and find, for each, the gap between signing up and doing the thing your product exists to do — the first report generated, the first integration connected, whatever it is. You will get a distribution, and it is almost always much tighter than people expect. Most of your converters did it in the first session or two.
That number sets the trial length, because the trial only has to be long enough for that plus one realistic delay — a weekend, a colleague's approval, a busy week.
What this usually reveals, and it will probably be your finding too: the people who did nothing were never going to. They did not run out of time; they never started. Extending the trial gives them more days in which to not start. That is why your move from fourteen to thirty did nothing you could see, and why it would have done nothing even if you had measured it cleanly.
Which reframes the question. The lever is not the length, it is the first session. If a large share of signups never reach the first useful result, no trial length fixes it, and that is where the conversion actually is.