There is no set number of sales needed, but instead this depends on how correlated your data is and how many variables you’re trying to solve for.
Often, more sales will help provide “better” results, but this is only the case when the data is properly correlated. Always seek “better” data than “more” data. Think “quality first” and then find as many sales as you can that fit that criteria.
Fewer sales will mean fewer variables that can be solved, particularly secondary drivers (e.g. bathrooms and garage bays) and tertiary factors (e.g. fireplaces and patios). Fewer sales also typically means more “work” to get credible results.
Using sales from history is a good way to keep strong correlation for the property features, while increasing the number of sales that you have to work with.
See the Knowledgebase article Proper Data Collection Tips for more information.
See the YouTube video Getting Results in Rural or Custom Markets for practical examples of working with limited data.