In more detail
There is no handshake and no retransmission, so UDP has little overhead and low delay. Its header is just 8 bytes: source port, destination port, length and checksum. Each datagram stands alone and may arrive out of order, twice or not at all; over IPv4 one datagram can carry at most 65,507 bytes (65,535 minus the 8-byte UDP and 20-byte IP headers). It suits traffic where a late packet is worse than a lost one: DNS lookups, video calls, online games, streaming and VPNs. Applications that need reliability add it themselves, as QUIC (the basis of HTTP/3) does.
Why it matters
Because there is no connection, firewalls and NAT track UDP flows with timeouts, and an idle VPN or call can be cut when they expire. The lack of a handshake also lets attackers spoof the source address, which is why open UDP services are abused in amplification attacks.
Example
A DNS lookup is one UDP datagram to Port 53 and one datagram back, a single round trip. Over TCP the same lookup first needs the three-step handshake. dig example.com uses UDP by default; dig +tcp example.com forces TCP.
Common confusion
TCP guarantees order and delivery at the cost of delay; UDP guarantees neither, which is exactly why real-time traffic prefers it.