Comparing 'Apples and Oranges' in Evaluating E-Mail Delivery Rates
Top e-mail service providers Digital Impact and Bigfoot Interactive said separately that their delivery rates were strong in the second quarter. Digital Impact said yesterday that its overall delivery rate for the second quarter was 93.6 percent, up from 91.8 percent in the first quarter. Last week, Bigfoot said its success rate was 92 percent for all campaigns in the second quarter.
However, there is a debate over how e-mail service providers determine their delivery rate, a key metric to marketers concerned that their e-mail is not getting through to customers.
"There are great differences in how vendors are reporting their delivery rates," said David Daniels, an industry analyst with Jupiter Research.
Daniels said the problem begins with the fundamental question of what constitutes an attempted delivery. In determining their delivery rate, some e-mail service providers do not count so-called hard bounces, which result from permanent e-mail failures like a misspelled e-mail address or a closed e-mail account.
Jupiter found that 61 percent of e-mail providers do not count such hard bounces, lowering the gross number of mail pieces sent. The research firm singled out four e-mail firms -- Digital Impact, @Once, Skylist and ExactTarget -- for including all delivery failures when compiling their delivery rates.
Digital Impact, San Mateo, CA, which includes all hard bounces when determining its delivery rate, calls failing to do so "metric manipulation."
"I think it's important the industry talk about it openly and agree how some of these key metrics we use to judge success are calculated," said Dave Lewis, vice president of deliverability management at Digital Impact. "Otherwise, we're calculating apples and oranges."
Jupiter estimates that blocked e-mail cost U.S. businesses $230 million in 2003 and expects it to rise to $419 million in 2008 because of increased e-mail volume.
Cataloging hard bounces is made trickier by the varying ways e-mail receiving domains communicate e-mail failures. Lewis said categorizing bounces requires e-mail service providers to process hundreds of different error message formats from e-mail receivers.
By excluding those attempts for things like incorrect addresses, e-mail service providers boost their delivery rates, Daniels said. The different methodologies used to compile delivery rates makes comparing different providers and in-house systems based on delivery rates nearly impossible.
"You really need to understand the methodology the vendor is using," Daniels said.
Joshua Baer, CEO of e-mail service provider Skylist, Austin, TX, said removing some undelivered mail could hinder marketers in getting an accurate picture of their e-mail campaigns.
"The customer's trying to use this information to evaluate their performance and how they're doing compared to others in the industry," he said. "In that way, I think it's doing them a disservice."
E-mail deliverability firm Return Path, New York estimates that industry-wide nearly 20 percent of all requested commercial e-mail is blocked by ISPs in some fashion. Return Path only counts e-mail delivered to the in-box as mail successfully delivered. E-mail service providers only gauge whether e-mail gets through the gateway.
The exclusion of hard bounces is particularly important since delivery failures, like sending to an unknown address, can trigger further filtering, said George Bilbrey, vice president and general manager of deliverability services at Return Path.
"That's the No. 1 screen many ISPs use in determining if you're a good or a bad mailer," he said.
A further complication is that excluding hard bounces will pollute other e-mail metrics, such as open rates, according to Daniels.
"It would behoove everyone in the industry if we have a set methodology," he said.
Despite the concerns, Bilbrey advises that marketers themselves have the most control over whether their e-mail is successfully delivered.
"Most of the levers are not in the ESP's hands at all," he said. "To use this as a way to choose an ESP is not necessarily the wisest choice."