Tokenization is the optimal method because it allows specific sensitive data elements, such as credit card numbers, to be replaced with non-sensitive equivalents, referred to as tokens. These tokens can be used in various operational processes without exposing the actual sensitive data. This is particularly useful for customer data analysis, as the analysis can often be performed with the non-sensitive token rather than needing the actual credit card number. Encryption, while it also obscures the original data, would not be as convenient because data analysis would typically require decryption. Masking affects the utility of the data for analysis because it often involves altering part of the data permanently. Lastly, hashing is incorrect because it is non-reversible and thus unsuitable for scenarios where the original data might need to be accessed again.