When it comes to commitment schemes for data, the state of the art today requires us to go through trade-offs. Do you want fast and cheap proving or fast commitment outside of the circuit? Choose one. Merkle Tree Translation aims to allow picking both options and thus reducing the overhead of general proving over that data, and especially SnapDeals.
SnapDeals currently have a proving overhead comparable to PoRep, with Merkle Tree Translation (MTT) another commitment to the data could be used, reducing the proving time and cost of SnapDeals 40 times.
It would also allow efficient production of proofs using that data, allowing a verifiable version of Compute over Data.
See for the earlier iteration.