For anyone interested in detecting mail delivery to their mailbox,
I use a Samsung/SmartThings (ST) Multipurpose Sensor (MPS). While these are generally promoted for their Open/Close (door/window) sensors, they also have a very good accelerometer in them. As a mail detector, in my hands, the MPS accelerometer does pretty good, but not 100%. Very roughly 90% of the time.
Actually sensing Open/Close, per se, with a metal mailbox is a very awkward proposition both because it’s a metal enclosure, and just the pure geometry of trying to mount the sensor around the arched opening and/or where it’s not blocking/knocked off by mailman or seen and taken by kids walking down the street. Maybe you could put it up underneath (the bottom edge of our plain metal mailbox door swings back when you open it) but then the unit would have to be mounted right up against a nice big metal thing when reception isn’t good already (see below). And even then it’s still awkward geometry; you don’t have a ton of space, and are trying to detect something moving just a small distance, in a radial fashion (axis around the mailbox door hinge).
As an aside, I might mention that, for the umpteenth time I wish someone would make sensors that are separate from the rest of their unit. Here, it would be wonderful if there was a teeny-tiny open/close sensor (or accelerometer) that was on a very thin wire 2’ long or whatever (but easily extendable becuase it’s, well, a plugin sensor, thank you very much). Meanwhile the actual device including antenna could be put some other better access / broadcasting place. Sometimes it really is a huge inconvenience that companies always insist on putting all parts into one inseparable package. There are also good reasons for everything to be in one package, of course. I just wish that not every single company, every single time, insisted on it.
So anyway, instead of trying to use the MPS’s Open/Close, I simply stuck it to my wooden mailbox pole, up in the little V alcove of a little buttress beam supporting the mailbox. It’s hidden from view and rain here (so far, fingers crossed). I used 3M VHB #4920 double-sided tape for a solid yet non-permanent bond (to catch the movement). I also have it oriented right-side up, which is to say, I stuck the white top of the MPS to the underside of the mailbox beam. This way, any moisture that might by chance get there will tend to roll down the sides and maybe only tend to moisten the join there. And it would never pool inside the white cap (which it would do if I stuck the gray bottom of it to the underside of the mailbox beam).
.
These ST multipurpose sensors accelerometers are incredibly sensitive (just a little tap is enough) and always signal, when tested inside my home. In fact I had to send one back because it was super over sensitive… not only would it go off with people walking down the hall when it was attached to a wall near there, but it even started to go off spuriously. Maybe it was just detecting Hawking radiation particle annihilation.
So the ST Multisensor was great for sensitivity and sending a signal. But its Zigbee signal was problematic, apparently. It is about 50’ open air from mailbox to front of house, then another 40’ or so to my ST hub. If I plugged a Samsung SmartThings Outlet into a wall receptacle on the front wall of our house (inside the house) as a relay, reception was extremely spotty. This is behind the brick wall front of our house. But if I rest it on the window (wood frame) on the inside of the house with old-school metal screen on the window, it does much better (85%). Maybe it would be 100% if I took down the screen or actually put the Outlet outside (but then it’d be exposed to elements). It has done fine with occasional Atlanta GA rain and humidity for several months, so mounting it under the beam seems to be okay. Definitely not covered by warranty though, I bet. 
.
If my MPS doesn’t detect a mail delivery, at this point I can’t be sure if it’s because of reception, or because the mailbox simply wasn’t shaken much. If I put mail into it very lightly, it will not go off. If I deliberately bop the mailbox a few seconds later, it will almost always go off. So, in ideal testing conditions, the actual amount of acceleration of the mailbox does matter. (Remember it’s mounted on 2-sided tape which probably damps it a little. You could screw it on but that’s more work and not completely easily reversible.) The good news is that mailmen are not in the habit of lightly and gently taking their time to close the mailbox door. Most of the time it seems to work, but not always.
For the record, I also tried a Z-Wave Aeon gen5 Multisensor 6. Its Z-wave can go much farther (of course), but its accelerometer is clearly not as sensitive as the MPS’s. It was simply not good enough.
I also posted accelerometer data captured by @krlaframboise’s Simple Event Logger over in this thread, if anyone wants to see what it looks like.
I also once asked if the orientation of the MPS matters (to optimize detection) but the answer seems to be No, it’s equally sensitive in all directions. That thread also has some clues as to the chip used, and accelerometer sensitivity. (Thanks @tgauchat!)
For the record, the ST Multipurpose Sensors have a setting “Do you want to use this as a garage door opener”. This is irrelevant to using it in my particular accelerometer setup; it doesn’t matter. Maybe the garage door has it sense a change in angle, I don’t know. (If it does, that could be used in interesting ways.)
I hope this setup helps someone. Detecting mail was one of the two main reasons I decided to “go smart home” in the first place, but in reality (if you read through the lines) it was highly problematic. In fact, it was a complete fail initially (you can’t really use Open/Close, and the accelerometer is a rather hidden property). And the distance was a real problem, next to the metal mailbox.
So perhaps this will give someone else hope.
