Spam Pre-Tests, Revisited

We have continued to see issues with delivery to Comcast addresses, as well as with a few organizations, such as Universities and Non-Profits which probably use their own spam filters. In my testing on Litmus, I found yesterday that an HTML email which has a large (in this case, ~90%) quantity of text as content, will pass the Barracuda spam filter tests, but not the Postini tests. The very same email will also be delivered to Comcast subscribers. So it would seem a good rule of thumb is to assume that if your email fails the Barracuda spam filter testing, it will not be delivered (or not be fully delivered) to Comcast subscribers.

Also interesting to note: for yesterday's email send, I broke the Comcast addresses into their own list, and sent just the text-only version to those recipients. I had no Comcast delivery issues. These results tell me the problem is not with my actual content itself (i.e., there was not "spammy" wording in the email), nor is it within the links or link structure (i.e., none of the links was malformed, and the tracking automatically inserted in each link by our ESP was not causing problems), as the text version (not spam blocked) had the same links as the HTML version (spam blocked).

If you are suffering with Comcast blocks also, my suggestion to you is this: In the short term, try sending a hybrid version just for Comcast subscribers, similar to the old days of AOL, which combines text with very little HTML... or just send your text-only version. Doing so will ensure that those subscribers will at least receive some version of your message, rather than nothing at all.

Of course, this method is not an ideal long-term solution. Get the deliverability team at your ESP involved as soon as possible - they should have, or should establish, relationships with the correct people in charge at the major ISPs, and may offer valuable insight from past experience as well.

Michele, My ... My!

For some reason, a few months back I began receiving emails from a company called Michele. Other than my middle name being Michelle (spelled with 2 l's), I can't thing of a single common thread between me and this company. I've never purchased from them, and in fact haven't a clue what they sell, despite receiving many emails over that time period.

Anyway, I decided time had come to end the non-existent relationship between my inbox and Michele. This is the most recent email I received, on July 23:



This email has its share of issues, not the least of which being - Where is the unsubscribe link?!?

Can you find it? I couldn't. So, I marked it as spam. Hopefully Michele's ESP has an auto-unsubscribe feature when a complaint code is returned, or I'll keep getting these... since I still can't figure out how I am supposed to unsubscribe. If I get another, I'll reply to the email address from which it was sent, but something tells me it will bounce.

Hiding an unsubscribe link won't get you fewer unengaged subscribers, but it will give you a huge complaint rate and most likely, a blacklist entry. Make your unsubscribe link easy to find and easy to use, so that people who don't want to receive your emails, won't. You don't want indifferent subscribers!

Update: I did, finally, find the unsubscribe link. It is buried at the very bottom (in WHITE on BLACK, shame shame) in mouse type. When I clicked it, it opened a new window with just "Your e-mail address has been unsubscribed" floating in plain Times New Roman (at least it was black on a white background!). I'm guessing it was their ESP (Silverpop)'s default unsubscribe message. What a bad user experience.

Spam Pre-Tests Suck Wind - a Five-Hour Saga of Frustration

Yesterday, I spent five hours (and our development team joined me for 2 of those hours) attempting to get the Spam Score on a legit marketing email down from an original score of 8 to below the recommended threshold of five.

Is it just me, or is that timeframe WAY out of line?

Here's what happened:
The email in question involved a client who runs an auto dealership in California. It was a standard sales and service email marketing piece - current deals, a finance app, terms & conditions... nothing out of the ordinary at all. Due to the nature of the client's business, there were of course the usual red flags in the content (financial wording, "Deal", "Offer" "Special" repeated ad nauseum, etc.), so I did tweak the content a bit prior to loading it into our email deployment solution to be tested.

Upon loading the code into the messenger system, I ran the in-solution spam score testing, which is based on SpamAssassin's algorithms (this makes sense, as SpamAssassin's algorithms are widely used by many filters). That's when things started to go south.

The spam score for the original piece was an 8. An 8! I've done email marketing for over 10 years, and I have not once seen a score as high as an eight for a legitimate marketing piece. I have had scores over the recommended 5 before, but it was always for a reason that was easily rectified, such as a specific, obvious trigger word or phrase. Having done this for so long, I'm pretty familiar with what sets these filters off.

But in this case, the main causes of the high score were codes I'd never dealt with before: FB_GET_MEDS (BODY: Looks like trying to sell meds), and MPART_ALT_DIFF_COUNT (BODY: HTML and text parts are different). The first thing I did (being the easiest issue to address) was to go through and change the ALT tags on all the images in the HTML to match text from the text version. Reran the test - no dice. After about 15 minutes of tweaks, I finally copied a paragraph directly from the text version and pasted it into the bottom of the HTML. Lo and behold, that error was no longer on the report.

But the score was still a 5.1, and unacceptable. Since the 1.5 point penalty for have a low text-to-image ratio was going to stay regardless, since the HTML was so image heavy, I turned to the problem of the Meds rule violation. That rule was adding a hefty 3.6 point penalty, so it had to be addressed prior to sending.

Since there were no mentions of Cialis or Viagara or anything remotely close to that, I figured, well, maybe the section discussing free diagnostic fees was the trigger? I changed it to "free code check fees". No change.

After about an hour of this, I was fed up. So I enlisted the help of the tech team. After explaining my conundrum to the CTO, he asked a couple of the developers to take a look, as well as our HTML programmer. They threw out possibilities, and I continued to test each one.

Our frustration built as we made dozens of content changes, including: removing many of the "special" "offer" "coupon" type phrasing; taking out many of the adjectives (one sentence said "big" and "huge" very close together... perhaps that was causing the problem?; removing % signs and $ signs; even editing the word "enjoy" - just in case. Nothing dropped the score below 5.

Finally, after 3 hours and a missed deadline, we found the culprit: one of terms I was using for Google Analytics tagging, "ViewOnline" was the trigger term. When we changed it to "WebView", the score dropped to 1.5.

My takeaways from this horror show:

1. What the Spam Score tool says the error is, is not always the cause of the violation.


2. SpamAssassin's support truly stinks. I tried checking their Wiki, Googling, searching their site... while I understand that they don't want to reveal the actual trigger words to spammers, I expected to at least find another marketer who shared the same frustrations. But I didn't. So I'm writing one myself!


3. Don't dismiss any seemingly harebrained roots for these violations. Who knew that a section of a GA tag would trip a spam filter, let alone one that dealt with meds?


4. Document, document, document. I wrote down each violation and how we solved it, and am building an internal Wiki for our future emails.

Have your own tale of woe trying to solve spam trap issues? I'd love to hear about it. Please post it in the comments section so we can all help one another out.

Buy a List, Kill a Puppy?

Why in the world any email marketer would even consider purchasing a list these days is beyond me. It flies in the face of any and all best practices for multiple reasons, including: 1, it is about as un-permission-based as you can get and still be legal; and 2, you know nothing about the origin of those names. Adding them to your list, or sending from your IP, is risking your reputation, as well as (most likely) violating your ESP's terms of service.



Check out this video from "EMAPP" (The Email Marketing Association for Puppy Protection):