Whenever I discuss bounces with marketers, someone inevitably asks about hard and soft bounces. Virtually every information request that crosses my desk asks whether we split bounces this way. A straw poll of deliverability professionals indicates they all struggle with the same issue: disagreeing with the concept of hard and soft but having to live with it.
The reason for this preoccupation is the psychologically neat and intuitively obvious differentiation between temporary and permanent bounces. Unfortunately, the real world is nowhere near as tidy; using these artificial distinctions can be very harmful to your metrics and lists.
There’s no agreement on hard and soft bounce definitions. The e-mail delivery protocol, SMTP (define), classifies failures as transient or permanent. Some people consider a soft bounce to be a transient failure and a hard bounce a permanent failure. This is an excellent, accurate definition, but it doesn’t address future deliverability for an e-mail address, which is what marketers usually try to determine with the hard/soft distinction. Also, many e-mail systems don’t record transient failures unless they persist for several days, then most are pretty much permanent.
Other people suggest that a soft bounce is a condition such as a full mailbox. Yet such situations are often indicated by SMTP permanent failure codes, as ISPs consider these to be permanent. For example, Yahoo’s delivery instructions specifically cite “mailbox full” as a permanent condition that should cause removal of an address from a mailing list.
There are numerous other definitions that vary in precision and clarity. This disagreement on nomenclature results in great confusion, with marketers talking at cross purposes, but it doesn’t cause direct harm. The problem occurs when decisions are made based on these definitions.
Many e-mail metrics are based on delivery rate and are where the confusion starts:
- Deliveries = recipients – bounces
- Open rate = (opens * 100)/deliveries
- Click-through rate = (clicks * 100)/deliveries
- ROI = (revenue – cost)/deliveries
Are bounces defined as all bounces or only hard bounces? Does this include transient failures? Does it include just permanent failures (where the precise definition of what constitutes permanent varies by provider)?
What matters for the metrics isn’t an address’s future deliverability, but whether the e-mail was successfully delivered. The hard/soft distinction is unnecessary for metrics.
Hard/soft is simply too blunt an instrument. Failures don’t split neatly into temporary and permanent. Some may be permanent unless corrective action is taken. A Sympatico block, for example, is permanent, unless you contact the vendor to get it removed. Other failures may be temporary but become permanent if you take the wrong action. An AOL temporary block will expire unless you do something wrong, such as failing to reduce complaints or bounces. Both cases can occur with blocklistings. Transient errors (such as “unable to connect to server”) may be permanent.
Furthermore, bounce reason codes aren’t universally accurate or descriptive. A database failure can result in a “user unknown” error for perfectly valid addresses. Some sites will return with “system error” for a blocklisting, while others return it for a temporary configuration problem. Some may even return “user unknown” for a blocklisting.
Reliance on automatic classification can cause good addresses to be rejected through overly aggressive hard-bounce culling while bad addresses remain on a list due to failure to cull soft bounces. Leaving bad addresses on a list can be disastrous, resulting in high bounce rates and spamtrap hits as ISPs repurpose addresses that have been undeliverable for a long time.
For these reasons, you must put aside convenient yet misleading concepts like hard/soft and look beyond to the untidy reality that is e-mail delivery. Your marketing metrics and your list hygiene will benefit from the simplification and the complexity this entails.
Until next time,
Derek is off this week. Today’s column ran earlier on ClickZ.