I don't think the service that 23andMe was selling to consumers was truly scientifically based. They needed a way to attract customers, so they taylored it to be as attractive to customers as possible. 23andMe was never a business where people were the main customers. The idea was for people to use the service and then sell the collective genetic data to businesses that could use it for research and create products. $99.00 per buyer did not cover the true costs of running the business. The failure was that 23andMe could not execute on the commercialization of the generic data.

For whoever can commercialize the data that's in the vaults, the collected data can have great value.

Maybe 23andMe was (certainly they seemed to be based on the idea that all disease would be single-gene, single-mutation, mendelian inheritance).

But many modern treatments for cancer use genetic tests to determine if the treatment would be effective. Because most cancer treatments are truly challenging (serious side effects), we want to know whether the patient would respond. This has already been integrated with the treatment process and from what I understand, it's not junk science, and is fairly successful.