Beware the Pepsi Challenge!

Maybe you have seen this in your inbox test results before? Here's the scenario. Perhaps your inbox placement testing tool has a dozen seed addresses for a specific ISP or mailbox provider. For that provider, for the message you sent, ten of seed addresses show inbox delivery, but the other two seed addresses show spam folder placement. Confusing, right? What IS your reputation at this mailbox provider? Does this count as spam folder placement or inbox placement? And what should you do about it? Read on to find out.

This is what I (jokingly) call the Pepsi Challenge. When you see results like this at an ISP or mailbox provider, it’s because the mailbox provider is doing testing, inviting a sample of your (their) subscriber base to provide feedback on your email messages. It’s a taste test, of sorts, where the mailbox provider wants to know if your flavor of email messaging agrees with the palette of the subscribers receiving the mail.

How this works is, an ISP or mailbox provider will place some percentage of your email messages in spam. Not 100%, but not 0%. Maybe 5%-20%. The goal is to see if those sampled subscribers notice, and if so, will they do anything about it? If the mail is truly wanted, subscribers will be more likely to drag it out of the spam folder or click the “this is not spam” (TINS) button to tell the mailbox provider that no, this mail is not spam, and yes, it is wanted. This is a tricky endeavor. After all, how many people review every message in their spam folder? A lot of people will ignore these messages and never do anything about it. But one assumes that the mailbox provider will understand this, and consider it when determining future inbox placement based on some small percentage of feedback. The key being, hopefully that percentage of feedback is greater than zero. If nobody polled seems to care that your email is in the spam folder, that is strongly suggestive to the ISP that perhaps your mail ought to just go to the spam folder 100% of the time, because it doesn’t seem compelling enough to drive anybody to vote it as wanted.

Actively taking action to request feedback about a sender’s email messages isn’t new. Different versions of things like this have been around for many years. Microsoft’s “SRD” mechanism (aka the Spam Fighters Program-- I think that's two names for the same thing) has been around for years and might be the first broadly implemented active feedback mechanism. While implemented a bit differently, it has exactly the same goal: to solicit feedback on whether or not somebody is sending good mail or bad mail; with a goal of using this feedback to decide which senders to block or perhaps push to the spam folder.

So what should you do when you observe the Pepsi Challenge in action? Monitor, first of all. In a lot of cases it’s an early warning system of spam folder placement to come. You could wait and see if that’s going to be the case, or you could be proactive and look to boost your sending reputation by focusing on engagement, either specifically for subscribers at that mailbox provider, or if you really want to take the best possible preventative action, implement it across the board, for your entire subscriber base.

I'm not 100% sure how an ISP or mailbox provider decides to engage the "Pepsi Challenge" process for a sender's email messages. But I have strong suspicion about how it might work (and how I'd do it if it were up to me). I think that a sender with a really great sending reputation is less likely to run into this process -- they're more likely to get solid 100% inbox placement. And senders with a really bad reputation are not likely to see this either -- they're more likely to get 100% spam folder placement. That means that if you're running into spam folder sampling, I suspect that your sending reputation is somewhere between. You're not the worst -- but you're not the best, either. Improving your sender reputation would likely help prevent this scenario, and help you more reliably get mail delivered to the inbox.

Post a Comment