Cool thing: One of the other companies who were helping the effort devised a forensic SNP panel. SNPs are great, because you don't need nearly as much contiguous DNA to find them as with STRs. Then we had to make an efficient many-to-many matching algorithm. *Amazing* to see it work the first time.
Oh, and by "amazing", I mean this:
There were STR (old-style) DNA profiles where, of the 13 sites + gender marker, only the gender marker was there.
And it turned out to be wrong.
Which we knew because from SNP panel an analyst could identify them confidently.
Though also ridiculously garish.
Boss: "Use these exact colors for SNPs. I want them so ugly customer will *have* to specify the real ones."
Customer: "Yup, can see exactly what's going on from across the room. 10/10, no notes."
(We tested on that week's new data, because what else was there.)