Replication factor (durability tier)
Object storage keeps more than one copy of every file, spread across different machines. The replication factor is simply how many -- two copies, three copies. More copies means the data survives more machines dying at once, which is the difference between "a server failed" and "a customer lost their wedding photos."
Every extra copy is real disk you have to buy and bill for. Three copies of a terabyte is three terabytes on your floor. So you match the replication factor to how much the customer would mind losing the data: a hobbyist's cache can ride on one copy, a backup company's archive needs three. Picking too low to save money is how you end up in a data-loss horror story.
Detailed explanation
DurabilityTier maps to replication factor: Standard=1, High=2, Mission=3. A bucket survives N-1 simultaneous host failures. Storage is billed at logical_GB x replication_factor, so the cost scales linearly with the copy count. Real S3-class systems use erasure coding to hit the same durability at lower storage overhead (deferred here).
Copies only help if they land on independent failure domains -- different hosts, ideally different racks or AZs. Three copies on one rack die together when the breaker trips. The strongest version of this is multi-AZ replication: copies in physically separate buildings, which is the only way to honour the top durability/availability promises.