A mailbox that sits open on the internet gets mail from every direction: hundreds of senders, each with its own angle, a background hum of half the internet. That is how people picture it. An old archive from the predecessor service says otherwise. Traffic is spread so unevenly that every average misleads: very few senders write a great deal, and most addresses hear from exactly one.
1930 messages, 89 addresses, 75 sender domains
We went through the archive of incoming mail from those early years: 1930 messages to 89 different addresses, sent by 75 different domains. The spread is not wide, it is steep. What actually arrives in an open mailbox is described elsewhere.
Four addresses carry 80 percent
The four busiest addresses received 593, 459, 348 and 153 messages. That is 1553 of 1930: 80 percent of the mail to four of 89 addresses.
The sender side looks the same. Four domains account for 1259 messages, 65 percent: alm.com with 348, facebookmail.com with 319, phpclasses.net with 316 and phpclasses.org with 276. The top group: trade portals, a social network, shops and software makers, senders that never stop once you sign up. 22 domains reached ten messages or more, 27 of the 75 sent exactly one.
54 addresses received exactly one message
The other end of the distribution is the normal case. 54 of the 89 addresses received a single message, ever. Nine received two, four received three. On the sender side it is plainer still: 76 of the 89 addresses heard from exactly one domain. Ten heard from two or three, two from four to ten, a single one from more than ten. A mailbox like that is no marketplace but a line to one counterpart, and that is what a temporary address is for.
For 70 of 89 addresses it was over after one day
The afterlife shows the same imbalance. For 70 of the 89 addresses the first and last message fall on the same calendar day. Two received mail for up to a week, one up to a month, three up to six months, two up to a year. Eleven received mail for longer than a year; the longest gap between first and last message is 746 days.
The average across all addresses is 79 days, and it describes nobody. It comes from 70 cases at zero days and a handful that ran for years. A disposable address has no expiry date the sender knows about: when you stop needing it is your decision, how long the sender writes to it is not.
What the subject lines contain
Counting was by keyword only. 74 of the 1930 subject lines contain a confirmation word such as confirm, verify or activate. 214 contain a marketing word such as newsletter, offer or deal. Eleven concern credentials, words like password, reset or login.
Confirmation mail is the minority, even though it is the most common reason for a throwaway address. Concentration again: the large remainder is ongoing notifications from the few heavy senders, to the same few addresses over years. To get out of series like that, note that the unsubscribe link is not always the best route.
One domain breaks the pattern
hide.me sent 22 messages to 21 different addresses, so nearly every one went somewhere new. That is the fingerprint of somebody signing up with the same provider over and over. Every other domain wrote overwhelmingly to a single address.
What follows from this
- The normal case is one message. An address used for a single signup produces the pattern that 54 of the 89 addresses show: one sender, one message, one day.
- The large numbers come from long-term use. An address left running as a permanent mailbox for an account collects hundreds of messages from one source within years.
- Anything meant to last does not belong in an open mailbox. A confirmation link yes, two years of account notifications no. For that, the more durable ways of protecting an address are the better choice.
How the counting was done
The basis is the table old2016_nre_emails, the archive of incoming mail from getsend.xyz. The period runs from 5 December 2014 to 18 February 2025, but the weight sits at the start: 33 messages to 25 addresses in 2014, 535 to 35 in 2015, 1360 to 41 in 2016 and two single messages in 2025. Between 2017 and 2024 this table holds nothing, it was retired in 2016. So this text describes 2014 to 2016, not current operations.
Only metadata was counted: target address, sender domain, timestamp and subject line. Per address we determined the message count, the sender domain count and the gap between first and last message. The subject lines were machine-searched for keywords and are reproduced nowhere; message contents and individual addresses do not enter the analysis.
What the data set does not support: present-day numbers, addresses that never received a message, and any distinction between wanted and unwanted mail. A marketing email without a marketing word in the subject drops out of the count, so the 214 is a lower bound.