I have decided to start my family tree again to include sources. I will use a software program to record the research. At what stage / how many individuals is it best to first publish to an online repository?
I am a firm believer in a trial or pilot implementation of any major procedure to identify the potential pitfalls before you go too far. This is particularly important if you have not used the particular repository before.
I recommend that you upload a number of ancestors that is large enough to show up any oddities in the software relating to relationship between generations, numbers of siblings etc BUT still small enough that you can pull it all down and start again if something goes horribly wrong.
Perhaps 6 to 10 people will be enough if you choose then well. In some cases, I have even invented a "difficult ancestor" (such as an adopted child who goes on to marry three times and whose death cannot be confirmed) as a test of how a site handles problems. It is too late to learn this after you have uploaded 500 persons.
There is no requirement to publish to an online tree at all Garry. It depends what your objectives are. Your software program will help you record and manage your data locally, and should allow you to print off reports and trees for relatives.
You can exchange parts of your local data with relatives who might also have software products using GEDCOM files, although - as other posts have suggested - that data standard has a number of problems associated with it.
If you want to publish to an online tree then it may be because you want to contact more-distant relatives, or collaborate with others in adding to your tree. The choice will determine which of the available sites is most suitable. In both cases, though, you need to decide how much of your core collection you want to publish.
I would certainly not use some online tree as my definitive data - especially those whose "small print" T&C's claim they own anything uploaded. Hence, my local copy is my definitive data. I have an online subset of this posted on genesreunited.co.uk (run by the same people as findmypast) but this is only to attract the attention of distant relatives who I can then exchange data with on a voluntary basis. My online data is basically just a tree with names, dates, and places. There are no photos, images, notes, reasoning, biographical history, sources, or citations on there, but I would exchange most of these with someone I'm later in direct contact with. Sensitive information, or information on living relatives or relatives in living memory, is then under my control and may or may not be shared.
I'm not suggesting this is the approach you should take but it's one of several routes you probably need to think about, dependent upon what your objectives are. I personally have little respect for online trees at the moment because the ease with which people can copy-and-paste data from other trees, with little thought and no citations, means the quality can be appalling.