DNA evidence has revolutionized our understanding of mammal relationships over the last decade or more. It was a challenge to draw this tree showing the big picture, which is still developing. I could fit 67 mammals out of more than 5000, to represent 25 orders. 
I'm sure some of these groupings are not quite right, or will undergo further revision (or recent revisions I've missed!). The deep groupings among the placental mammals are well-supported but still a bit up in the air.
I approached this with a good background in mammal evolution, but a few details still stood out to me as surprising, and I bet most people don't know about them. For example, zoologists used to think that tree shrews were the closest living relatives of primates. No more!
Tree shrews now look like a relative of the rodent-rabbit clade, and the living sister group of primates are cologos, which some scientists used to call "flying lemurs".
My favorite is one that may not stand up: a bat-odd-toed-ungulate connection. Biologists have argued about bats since Linnaeus (who put them into primates). Even with mtDNA in the 1990s, it looked like there was a chance of a bat-primate relationship. But no. Horses? Could be.
Meng-Yun Chen and coworkers 2017 (doi.org/10.1093/gbe/ev…) found the strongest support for the bat-perissodactyl sister grouping, but this seemed weaker as examined by Jacob Esseltyn et al. (doi.org/10.1093/gbe/ev…).
A new large review by Nathan Upham, Esseltyn, and Walter Jetz puts Chiroptera on its own (HT @ed_hagen). (doi.org/10.1371/journa…) So maybe we'll have to "tapir" that sister-grouping down a bit. The tree from Upham and coworker's paper is massive!
@ed_hagen A well-supported finding from fossil and DNA evidence is the placement of cetaceans (whales and dolphins) into the tree of even-toed ungulates, now together known as Cetartiodactyla. Deer are closer to narwhals than either are to camels and pigs.
Knowing less about marsupials, I was surprised at the little monito del monte, which is closer to Australian marsupials than to other South American marsupials. Maybe a remnant of Gondwanaland diversity of the late Cretaceous.
With most phylogenetic trees, the details may change but the big picture is fairly constant. But mammals have so many divergences clustered in a narrow time frame, that adding more DNA data is really shifting the deep connections of orders and larger groups.
I'll point people back to the paper by Nathan Upham and coworkers that just came out last month -- it's the best and newest species-level phylogeny for mammals available, but most of the comparisons are based on a set of 31 genes. Genome-wide datasets are going to shift things!
