Jul 31, 2008

Should You Send from a Dedicated IP Address?

We’re getting more questions from customers with large email lists about sending their campaigns from a dedicated IP address, vs. using MailChimp’s "shared" pool of IPs.

They want to know if their deliverability will be any better if they use a dedicated IP.

It’s not a question we can answer with a simple "yes" or "no."

Mark Brownlow has a good overview of all the different points to consider before sending from your own dedicated IP.

At MailChimp, whenever we setup a new IP address (either to add to our overall rotation, or a dedicated IP address for a high-volume customer) it takes time to "break it in." Here’s what we go through…

The problem is, whenever you send lots of email from a fresh new dedicated IP address, ISPs will take notice.

Their filters will say,

"Hmm, that’s a whoooole lotta email coming from a fresh new IP we’ve never seen before. Must be an infected computer on a botnet, sending spam. I’m going to block those emails until they build up a good sending history."

And they watch email volume from that IP.

It can take weeks or months for them to finally come to the conclusion:

"Okay, apparently it’s normal for this IP to be sending so many emails, and overall, we haven’t received too many complaints or spam reports, so let’s just consider this IP trustworthy."

For example, here’s a MailChimp IP address as seen by Trusted Computing’s TrustedSource Internet Reputation Monitor:


This particular IP address has been in use for years, so they’re kind of used to receiving tons of emails from it. See how they watch "deviation in message count" over time?  There are normal "ups and downs" that occur (weekend traffic is generally down), but overall, things are pretty consistent. So our "Sender Reputation" is nice and green.

Now here’s what happens when we recently turned on a fresh new dedicated IP address for a customer who sends hundreds of thousands of emails, almost daily:


As you can see, email volume is non-existent until late July. Then, we turned on the IP and everything goes red, and we look like a "malicious sender." But the truth is, there’s nothing "malicious" about the emails being sent from this IP. They’re actually extremely good (and opt-in of course). You’d laugh at all this "malicious" terminology if you saw the squeaky-clean emails this customer is sending. It’s just A LOT of email, happening all at once, from an IP that never sent any emails before.

To try to offset the high volume from this new IP, we take huge chunks of their campaign and dsitribute those across our shared IPs. Only a small fraction of this customer’s email is actually being sent from this dedicated IP during the break-in period. But as you can see, for some filters, it’s still risky looking. And so you get delivery problems for a while.

So long as the volume stays somewhat consistent, and so long as spam complaints stay within acceptable thresholds, their dedicated IP will make its way to "Trusted." Actually, it’ll go to "neutral" for a while, then trusted.

But just because email volume looks suspicious, it doesn’t mean you’ll be totally blocked.

It’s more complicated than that.

Email Reputation Is Like Your Credit Score

MailChimp subscribes to ReturnPath’s Reputation Monitor. It gives us a birds-eye-view of how our IPs are doing. We like ReturnPath, because they also just so happen to manage the feedback loops for a LOT of the big ISPs out there.

Anyway, we don’t just send all emails from one IP. And we don’t just spread them out equally across a bazillion IPs. That would be stupid.

We have several different "ranges" of IP addresses. Some are "shared" (free trial accounts, & new customers), some are "trusted" (long-time users with good stats), some are "double-opt-in" (people who have never imported any outside list, and have only used our double-opt-in process), and a handful are "dedicated" (high-volume customers who purchased and setup a dedicated IP). The "deliverability" score for each range is different. As you would expect, the "double opt-in" range gets a slightly higher score than the "shared" range. They say the average score out there is 88%.

Now here’s how that same dedicated IP address I mentioned above looks to ReturnPath, compared to a few of our other IPs:


As you can see, the Accepted Rate  (which is basically the number of emails that weren’t rejected by the ISPs they monitor)  is actually higher than a couple of our other "shared" IPs. At least for today, it’s getting a 98.13%. Not bad. But you’ll notice that one of our shared IPs (mcsv15.net) has a higher "Sender Score" (95) than this dedicated IP (80).

How could that be? And what’s a Sender Score mean, anyway? We have no friggin’ clue.

Here’s ReturnPath’s definition of SenderScore:

This score is derived from a proprietary Return Path algorithm, and represents your overall performance against metrics important to both ISPs and recipients of email. Sender Scores are on a scale of 0 to 100, with 0 being the worst, and 100 being the best score. Like a percentile, your Sender Score tells you how you compare to other email senders. For example, a score of 70 indicates you are measured as better than 69 percent of all mailers, and worse than 30 percent of mailers.

As you can see, their algorithm is super secret, so we have no way of knowing exactly how our Sender Score is derived. But the number is pretty important, because more and more ISPs and spam filters are using ReturnPath’s scoring system:

Receivers use the top level scores, and (for receivers who contribute data) rules based on specific data elements as inputs to their mail filtering decisions. A simple implementation may be to reject all mail with a volume above some lower limit and a Sender Score beneath some level they deem acceptable. A more complicated implementation may be to whitelist IP addresses with Sender Scores above some threshold but blacklist senders who have recently hit a spamtrap, or who have generated very high complaint rates.

And here’s a list of some of the ISPs and spam filtering companies who use ReturnPath’s Sender Score Certified rankings. Long story short, the score is mysterious, but the score is important, because it’s used by ISPs to filter incoming mail.

So why does our dedicated IP address get a lower Sender Score? The answer could be in that "Problems" column: "Mailing Consistency." Over time, the email volume from this IP will appear more consistent, and the score will go up (so long as the sender doesn’t goof up with a campaign, and generate a lot of spam complaints).

Getting a high volume of emails delivered is complicated. A colleague here at MailChimp describes it like "cell phone dead zones." Sometimes, you get a dropped call. People have come to expect a certain level of reliability, but they aren’t shocked if they drop a call in the middle of a desert.

Email receivers are all constantly trying to find newer, smarter ways to filter out spam. A dedicated IP address is not a silver bullet. Today, at least. It’s true that a dedicated IP address will isolate you from any "bad apples" that might be on a shared pool of IPs. But if you’re using a decent email service provider, there shouldn’t be very many "bad apples" on that shared system anyway. Hopefully, they have some kind of review process that keeps bad apples from even joining (at MailChimp, our human review team scans about 30 different criteria to determine if a user is safe enough to send). And so you can piggy back on that shared IP range, because there’s "strength in numbers." Then again, you may be a financial institution doing some intense authentication, and you absolutely positively need a dedicated IP. If you have the patience (and volume) to break one in, you might get better inbox acceptance rates. If you also have the budget to get the IP SenderScore Certified, you can get whitelisted by some ISPs and spam filters. But if your volume isn’t consistent, or if it’s sporadically high, your overall reputation score may suffer, which could result in delivery problems. Some people (especially those who are new to email marketing) are better off using a shared IP, while they build up their lists and hone their good email habits.

It’s a lot like your credit score. There are tons of factors that credit agencies look at to determine your score. And there are different credit agencies with different scoring algorithms. A person with a ton of money in the bank can have a terrible credit score, just because they don’t own a credit card, and never take out any loans, and always pays with cash. There’s simply no history to reference. And when that person is finally ready to get a credit card or loan, their low credit score will hurt them. Eventually, they’ll get a credit card, and build up a strong history. Maybe even get a great credit score. But if the person tries to buy too much stuff at once, it might trigger a block at the credit card company, resulting in a really awkward moment at the cash register.

ISPs and receivers do the same thing. They can’t peek into your office or look at your financial records to determine that, "Okay, this dude seems legit, so let’s whitelist him." They have to look at your email reputation, which is composed of your email activity, like: volume consistency, abuse complaints, server setup, and content.

See, I told you this was going to be a complicated answer.