In my opinion only the third is really worth it. Ethnicity estimates are a curiosity, are not accurate at the level most of us care about in genealogy - the big exception is African descendants of slaves, who have both higher accuracy due to the genetic diversity of Africa, and fewer other options.
But that really puts it closer to family discovery, which is a real use case. Caring who you're related to and how is damn near a cultural universal, and some of us care a lot.
Health/traits stuff I see as little better than astrology. Yes, you can find out stuff like if you have blue eyes - but so can a mirror tell me. Odds are you already know yourself far better than the DNA traits testing companies can infer from your DNA.
it's even worse if one of your siblings or close relatives does this. people have been implicated in crimes due to relatives getting tested. on the one hand that's great they're closing cold cases, but the disaster where this is used incorrectly against you is horrifying.
The ends do not justify the means. The police should not have an easily searchable database of everyone's genetic profile, no matter how they gathered it.
I'm not interested in services of finding relatives or health coach gimmicks, I just want an one-time purchase of a raw data blob with solid personal privacy guarantees. Am I asking for too much?
Why isn't there a privacy focused company that just doesn't do this.
i mentioned in an earlier comment i'd be happy to start this company if there were that kinda demand. it'd cost you around $1000 USD to get this done and get a hard drive mailed to you with raw data (fastq files).
EDIT: it probably would be cool if this kind of thing existed so consumers could shop their analysis around. it might actually open up some consumer bioinformatics startups
I think there is some sort of middle ground here. Although 99% of consumers don't care about privacy, it's an org risk as we saw with 23&Me. A consumer focused, privacy heavy but cloud based platform will emerge in the wake of 23&Me.
Gemini says a reasonable cost estimate from saliva to SNPs is $163.84 - $211.84 using Ultima $80 for sequencing (funny that sequencing is no longer the bottleneck on cost).
FASTQ is fine but I'd prefer BINSEQ: https://github.com/ArcInstitute/binseq. I'm writing my own pipeline from BINSEQ.
They do have a family matching database, but you can opt out of it I'm pretty sure. They do try to upsell you on frightfully expensive Y-DNA long read sequencing for niche genealogy use cases, but you can just ignore that...