Data Segmentation

Testing and data driven decisions

There’s a lot of my education in the sciences that focused on how to get a statistically accurate sample. There’s a lot of math involved to pick the right sample size. Then there’s an equal amount of math involved to figure out the right statistical tests to analyse the data. One of the lessons of grad school was: the university has statistics experts, use them when designing studies.

Transactional mail

There are a lot of myths in the email space. Things that someone, somewhere said and another person repeated and then another person repeated and all of a sudden it is TRUTH. One of those things is the idea that there is a law defining what can be in a transactional email. Supposedly this law says that 80% of the message must be transactional content while 20% of the mail can be promotional content.
This isn’t really a law. I was even going to say it’s kinda a good idea, but then I stared thinking about it. It doesn’t even really make sense. 80% of what? Size? Space? Bytes? Layout? Do headers count in the 80% or just what’s visible? Does the HTML code count? What makes for “new” content?
Adding promotional content to receipts is great for conversions. It’s a great way to get someone to opt-in to mail. It’s a great way to upsell. It’s great for engagement; that makes it good for deliverability. Senders should include some level of promotional mail in receipts whenever possible.
There are some guidelines I suggest when looking at transactional mail.

Unsubscribe preference centers

I unsubscribe from a lot of opt-in lists around this time of year. I’m generally unbothered by a couple emails a week from companies I’ve purchased from in the past. But, a lot of these companies drastically increase their volume mid-November. I may not be averse to 3 emails a week, but that absolutely does not mean I want 2 emails a day.

Data, data, elections and data

One of the interesting stories coming out of the recent US Presidential election is how much data the Obama Campaign collected about voters, volunteers and donors. Today Politico talks about how valuable that data is, and how many Democrats want to get their hands on it.

Data Driven Email (and other) Marketing

The frequency of emails from the Obama campaign ended up being a talking point for pundits and late night talk show hosts. Jon Stewart of The Daily show even asked President Obama about email directly during his October 18th interview. (Video, email question at the 5:56 mark)

Targeting and Segmentation

MarketingSherpa has a great case study of a retailer that got a 208% higher conversion (purchase) rate for a targeted email sent to a small segment (10%) of their list.

Data Cleansing part 2

In an effort to get a blog post out yesterday before yet another doctor’s appointment I did not do nearly enough research on the company I mentioned selling list cleansing data. As Al correctly pointed out in the comments they are currently listed on the SBL. And when I actually did the research I should have done it was clear this company has a long term history of sending unsolicited email.
Poor research and a quickly written blog post led to me endorsing a company that I absolutely shouldn’t have. And I do apologize for that.
With all that being said, Justin had a great question in the comments of yesterday’s post about data cleansing.

Don't always believe the statistics

Mark Brownlow has a great roundup of how statistics and data can mislead marketers if they’re not really paying attention.

Best time to send email: analysis and discussion

Mark Brownlow (who I don’t think is here in Ams, much to my disappointment) wrote a long assessment of how to determine what is the best time to send email. He walks through the questions and the data that a sender should evaluate when making the decision when to best send email.
I have previously posted about my views on the best time to send email. There is no one best time to send email. In fact, my experience leads me to believe if someone said the best time to send email is at 4pm on Tuesday afternoon then 4pm on Tuesday afternoon would rapidly become the absolute worst time to send email.
It should come as no surprise, then, that I really like Mark’s #4 recommendation.

Best time to send email: redux

Last week I wrote about a study classifying different types of email users. My point is that senders should be very aware of how their users interact with email, in order to provide the best user experience and the most revenue for the sender. If, for instance, the bulk of recipients are daytime (9 – 5 M-F) users, then the best time to email is different than if the bulk of recipients are all the time users of email.
At least 2 different people commented on when the “best” time to send email was, completely missing the entire point of my post. When you send email should be related to when your users are active in their email client. Senders know this, because they can track times when people open and click on links in the email. The data is all there, it just needs to be mined.
Plus, if every sender sent mail at the exact same time, that being the best time to send mail, then it will immediately become the absolute worst time to send email.
Pay attention to your recipients, and not to the internet experts. Listen to what your customers and recipients are telling you. Do what’s best for them, not what’s best for Joe’s Bait and Tackle Shop.

Best time to send marketing email

Pages and pages have been written about the best time to send email. Marketers spend significant amounts of energy discussing and researching the best time of the day and the best day of the week to send email. I have long thought that these discussions do not put enough attention on individual end users and how the recipients interact with email.
Researchers recently developed a model for email user behaviour that splits email users into two classes “e-mailaholics” that send, and presumably read, email all the time and “day labourers” that send, and presumably read, email during standard business hours. There is very little transition between groups, 75% of users stayed in the same usage group over the 2 years of the study.
What does this mean for senders? Senders need to know know how their recipients use email and which user group recipients are. By analyzing clicks and opens, senders can classify recipients and use that data to send mail that is more relevant and better targeted.
h/t arXiv blog at Technology Review

How to devalue your mailing lists

This morning I got spam about college basketball – Subject: Inside: your ESPN Tourney Guide. That’s anything but unusual, but this spam got through my spam filters and into my inbox. That’s a rare enough event that I’m already annoyed before I click on the mail in order to mark it as spam.
Wait a second, the spam claims to be from Adobe. And it’s sent to a tagged address that I only gave to Adobe. Sure enough, it’s Adobe and ESPN co-branded spam about college basketball sent to an Adobe list.
Down at the bottom of the email there’s a blob of tiny illegible text, in very pale grey on white. Buried in there is an opt-out link: “If you’d prefer not to receive e-mail like this from Adobe in the future, please click here to unsusbscribe“.
I’d prefer not to receive college sports spam from anyone, including Adobe, so I click on it and find a big empty white webpage with this in the middle of it:

Palpable ennui

Put any group of senders together and the conversation invariably turns to discussions of how to get email delivered to the Inbox. There is an underlying flavor to most of these conversations that is quite sad. Many senders seem to believe that the delivery of their email is outside of their control and that since the ISPs are difficult to reach that senders are stuck. The ennui is palpable.
I am here to tell you that nothing could be further from the truth!
Senders are not passive victims of the evil ISPs. In 99% of cases, delivery problems are fully under the control of the sender.
Mail being deferred? Mail being blocked? Mail being delivered to the bulk folder? Senders do NOT NEED TO CALL THE ISP to fix most of these. Tickets do not need to be opened nor do personal contacts need to be employed. You can resolve the vast majority of problems with data you already have.

Customer support surveys

I have seen a lot of companies attempt to send out customer support surveys by email, only to fail dismally. Generally, the intentions of the companies who do this are good, but the executions are appalling. Companies have found any number of ways to invite epic fail to call, including mailing to non-customers, mailing to the wrong person at a customer company and mailing to former customers.
Mailing to non-customers generally happens when companies sort abuse and support mail through the same ticketing system. Good customer support (tell us how we did) turns out to be rotten complaint support. The failure here is multifactorial, but revolves around not understanding the difference between customer support mail and abuse complaints. Abuse is not, usually, mail from your customers. More often mail to abuse is from non-customers. While it may seem like a good thing to follow up with abuse complaints to find out if the person is satisfied, generally someone who complains about spam does not want more mail from a company. The fix it to change the selection process for surveys. Survey customers not complainers.
The second failure is more common with enterprise vendors. Generally the vendor will have multiple contacts at company but send a single survey out to all contacts at the customer. Take an average website that provides statistics about web or email performance. A company establishes an account there, and then provides a logins for customer support people, a manager or two and maybe an outside consultant. These people are all using the same site, but are possibly using different parts of it. The consultant can give some feedback on the API and data access, but is not the right person to ask about pricing, packages or overall usefulness and value for money. Management can provide feedback on pricing and value for money but probably has never logged into the website, despite having a working account. Customer support can provide feedback on the user interface and overall usefulness of the site. Knowing who is who at the customer and who is the right contact for different surveys can be tricky, but it is always better a company to appear to be acting purposely.
Finally, some companies send out surveys to anyone who has ever registered for a website, or game or product no matter how long ago that registration was. They send mail to the person who registered for a website but has not logged in for 6 months, or 12 months or even longer. The recipient may have even taken positive action to close an account, such as discontinuing payments. And, yet, the company still mails them a customer satisfaction survey. If the recipient is not paying for the product, if the recipient is not logging into the website then they are no longer a customer. Sure, there are times to reconnect with old customers, and it can be done well. However, what I am talking about is the survey that is clearly designed to be answered by current users and customers.
The sad thing is, I have received customer satisfaction surveys in all of the above categories in the last 6 months.
If you as a sender, are going to use customer satisfaction surveys, do it in a thoughtful and purposeful manner. Do it in a way that brings value to your company and to the people you are surveying. If you do not, you risk higher complaint rates. Remember, people who are not your customer or who are a former customer are probably more likely to hit “this is spam” then to answer your survey. Like any mail you send, make sure you know who your audience is and have a mental model for how they will treat your mail. Do not just grab all available addresses and mail them. Do some analysis of your customer base before you mail and mail them surveys that apply to them. You will get fewer spam complaints and probably more and more accurate survey responses.