Hard bounce vs soft bounce: which one to worry about

Every bounce is either permanent or temporary, and the difference decides what you do next. Hard bounce vs soft bounce sounds like trivia until your bounce rate climbs and you have to decide whether to clean the list, slow down, or change domains. This is how to tell them apart and which one to worry about.

5xx and 4xx: the actual difference

Every SMTP reply starts with a three-digit code. The first digit is the whole story (RFC 5321):

• 5xx is permanent. The server is refusing the message and won't take it if you try again. That is a hard bounce.

• 4xx is temporary. The server couldn't take it right now, and trying again later may work. That is a soft bounce, and your sending server normally retries for you.

The enhanced code that follows (like 5.1.1) says why, using the classes in RFC 3463: .1. is about the address, .2. the mailbox, .4. the network, .7. a security or policy refusal. Our bounce code library explains each code and whose problem it is.

Hard bounces: examples and what to do

• 550 5.1.1: the mailbox doesn't exist. A typo, a deleted account or a guessed address.

• 550 5.2.1: the mailbox is inactive.

• 553 5.1.3: the address is malformed.

• 552 5.2.2: Gmail's full-and-inactive mailbox, sent as permanent.

What to do: remove the address. Do not retry. Yahoo's guidance for permanent errors is the same: take the address off your list. One hard bounce is nothing. Many from one list is a bad source, because mailbox providers read a lot of mail to addresses that don't exist as a sign of a careless sender. Stop that list before it becomes a reputation problem, and verify addresses before the next send. High email bounce rate? Burned domain or bad list shows how to read the pattern.

Soft bounces: examples and what to do

• 452 4.2.2: the recipient's mailbox is full. Their problem, temporary.

• 450 4.2.1: Gmail says the recipient is receiving mail too quickly, so it's deferring yours.

• 421 4.7.28: Gmail is rate-limiting you. This one is about you, not them.

What to do: let the retry happen. If the same address soft-bounces across several sends, stop sending to it, since an address that never accepts mail isn't worth the risk. A full mailbox alone isn't a reason to give up on someone; reach them another way. One soft bounce on its own is noise.

When a soft bounce is really a block

The 4xx label can hide a sender problem. A temporary code is often a provider telling you, politely, to slow down.

Take 421 4.7.28. It's technically temporary, but its text says Gmail has detected an unusual rate of mail from your IP or domain, and other versions name the domain in a link. Mail may be delivered on retry, but Gmail is telling you what it thinks of your sending. A run of these across many recipients at one provider is a throttle, not a set of full mailboxes.

How to tell:

• Many recipients, one provider, same code: that's about you, not them.

• A 4.7.x code (security or policy) is a different animal than a 4.2.x mailbox code.

• Soft bounces that start the day you raised volume point at your own ramp.

Providers also don't always follow the classes cleanly. Gmail sends 552 5.2.2 as permanent for a full mailbox, where the RFC prefers a temporary code. Read the text after the code, every time. For the reverse case, where a permanent code means the sender is the problem, see 550 5.7.1 and 550 5.7.26.

Hard bounce vs soft bounce: which one to worry about

Hard bounces first, because they're the ones you cause and the ones that pile up into reputation damage. Soft bounces deserve a look when they cluster. Do the sums yourself: bounces divided by emails sent, split by hard and soft, per sending domain. If you want to dig into what reputation does with all this, read domain reputation.

In SendingOS, every bounce is sorted into hard, soft or block, with its reason (unknown recipient, mailbox full, spam block, authentication policy, DNS failure) and the SMTP code shown on the conversation. The thread is labelled Bounced, so you can filter the inbox to just the bounces and pull dead addresses fast.

Working out your numbers

Don't lump bounces into one figure. For each sending domain over the same period, work out two rates: hard bounces divided by emails sent, and soft bounces divided by emails sent. Say you send 1,000 emails and 15 hard-bounce while 10 soft-bounce: that's 1.5% and 1%. (An example with round placeholder numbers, not a typical result.) The hard rate tells you about the list. The soft rate, split by provider, tells you about pace and about the other side's mailboxes. If the hard rate jumps after you add a new list, the list is the cause. If the soft rate jumps at one provider the day you raised volume, your ramp is.

What not to do

• Don't retry a hard bounce. It adds another bounce and tells the provider you ignore feedback.

• Don't delete every soft bounce on sight. You lose real people with full inboxes or a bad afternoon at their server.

• Don't read "550" as one thing. It covers a dead address and a spam block alike. The enhanced code and the text decide.

• Don't switch domains at the first block. Find out whether it's your list, your authentication or your volume. SPF, DKIM and DMARC is the first thing to rule out.