Bounce Handling Best Practices

Last week I was doing some research on bounce handling best practices. My main questions were these:

• Is marking hard bounces as undeliverable after the 3rd consecutive hard bounce within one year the right way to treat hard bounces? Personally I think 3 is too many.
• Should we set different thresholds for hard bounces, soft bounces and spam complaint bounces? If yes, what should that threshold be?

Today, I'd like to share my findings with you:

Matt Vernhout referred me to his blog where he had just posted a piece on bounce handling. In this article he explains the difference between a hard bounce and a soft bounce:

• A hard bounce (or permanent failure) is one which is not likely to be resolved by resending the message in the current form. English translation: "Do not try to deliver this message again as it currently exists".
• A soft bounce (or transient failure) is one in which the message as sent is valid, but persistence of some temporary condition has caused abandonment or delay of attempts to send the message. English translation: "Try to deliver this message again, but later".

and he gives some examples of each and links to more information about bounce codes. Read more here.

In response to my question on the Email Marketer's Club forum he share some more of his wisdom:

• Does the system differentiate between "571 I don't like your content" and "550 User unknown" - or do these both count as the same thing toward your disable threshold? The first should have the opportunity to try again based on business logic like you speak of (3 hard bounces in one year), while the second one should be removed immediately.
• Soft bounces are a little different these are usually temporary errors and should be built around business rules, but take into consideration - 1 soft bounce followed by a success should reset your counter to 0.
• Spam Complaints should be removed immediately from that mail stream. These should also be reviewed and measured against some metrics (such as opt-in source, time on list, # of messages) to see if there are potential relevance or recognition issues for these members. These should also be considered as an opt-out from your emails - only resuming after a valid opt-in. Keeping these number below 0.1% is imperative. Check this by measuring against domain volume and not total volume (i.e. AOL sent/AOL complaints < 0.1%).

I also posted my question on the EmailRoundtable and got some more insights:

Stephanie Miller responded that "one hard bounce and off the file is the best rule.  Hard bounces are just that - hard.  There is no one there and won't be".

Dennis Dayman agrees with Stephanie: "At 33% address churn per year, your list may contain a high number of undeliverable records if it hasn't been recently mailed or too much time has elapsed between data capture and mailing. We suggest removal on the first hard bounce. There is no reason to attempt delivery a second and third time. With the fundamental change in anti-spam methodologies going from primary being content based filtering to more reputation based filtering these days, it becomes more important to ensure removals of hard bounces quickly."

When it comes to soft bounces he says "you need to examine the specific failure types to determine the likely causes:

• Resend records that fail for temporary conditions such as mailbox full
• Invalidate records only if the same result is received on subsequent mailings over at least thirty (30) days. Look for 3-5 times
• Contact the recipient for replacement or corrected email address or flag for Address Recapture either by postal mail or another emailaddress from ECOA programs like what ReturnPath has"

Finally, when it comes to dealing with complaints, Dennis says: "The number in your reporting should include both complaints received from established feedback loops (AOL) and formal abuse complaints filed with system administrators that come back as bounces.. Managing your complaint rate is important because they trigger ISP blocks and blacklisting, and are a key component to your reputation as a legitimate email sender. Since only a handful of ISPs offer complaint feedback loops, you should use this data as a proxy for customers you may not be hearing from..."

Bill Kaplan added to that: "I'm in agreement with Stephanie that "one hard bounce and you're out" is the best rule.  This should also be the rule for spam complaints. ISPs basically watch three list hygiene thresholds in deciding whether or not to block your email: a) bounce thresholds, b) "This is Spam" complaint thresholds, and c) spamtrap/honeypot email address thresholds. "Better safe than sorry" is your best course of action these days if you want to maximize your deliverability.  Or, as we instruct our clients, "When in doubt, throw it out.""

Thanks to everyone for responding to my question! If you have something to add, feel free to post it in the comments or on the Email Marketer's Club.

Source: EmailKarma, Email Marketer's Club, EmailRoundtable

Comments

[Maarten Oelering]

As the author of email marketing software I know how hard it is to give general guidelines on bounce management. To decode, analyze and classify the hundreds of different bounce mail formats is a complex task. Furthermore, the bounce code or description in the bounce mail is sometimes inconsistent or just incorrect.

The famous two categories, "hard" and "soft", for which many use different definitions, is vague concept and a generalization which should not be used to base any business logic on. On the other hand, trying to be very specific about the bounce reason is also problematic. When the bounce mail is not clear or misleading, invalid addresses may bounce forever or valid addresses may be thrown out of the list.

I think the best "guideline" is to properly filter out all the invalid recipient addresses (mailbox unavailable, user unknown, invalid recipient, etc.), and remove them from the list after one mailing. All other bounces with a code "5" (not delivered after one or more retries) should be removed from the list as soon as the following conditions (which are also specified by the ISIPP) are both met:

- at least three consecutive bounces have occurred, and
- the time between the first bounce and the last bounce is more than 15 days

By "generalizing" this way, a good balance is achieved between the risk of getting into deliverability problems, and the risk of misinterpreting inaccurate bounce mails. For more background I would recommend to read the articles on bounce management by Derek Harding on ClickZ.

[Tamara Gielen]

Thanks for the insights Maarten. It sounds pretty tricky indeed!

[DJ Waldow]

Tamara -

We just posted on this issue on Monday - http://blog.bronto.com/bronto_updates/2007/08/31-flavors-of-b.html

dj at bronto