My (lay) understanding is that personality data tend to be clustered such that it's best explained by five factors, with the names of the factors (Openness, Conscientiousness, etc.) having been added later.
The Big Five are a statistically-derived groupings of phrases that are reasonably stable when measured from different angles. I think it would be fair to call it a "modest initial result" if compared to the physical sciences, or "wildly successful" if compared to basically anything prior that came out of the social sciences. There are some quite reasonable critiques listed on the wikipedia page, such as a lack of modeling theory, or only accounting for a portion of personality differences, or arguments for a slightly different number of factors, etc. That said, it seems to be one of the first actually solid findings that ever came out of the social sciences. It's like the whole field has been flailing around, swept down a river in the dark, and for the first time we found a stone that has a bit of purchase, even if it's covered with moss and slippery.
That being said, it's much much harder to develop measures when you rely on other measures of the same type to validate them, which is where personality testing is right now.
Thank you for the correction.
It's a much, much better method for designing and analysing questionnaires, and it is the foundation for all standardised testing (SAT, MCAT etc).