If you ever tried to segment a list by engagement, you may have wondered about our member-activity rating. I’m talking about the little stars that look like restaurant reviews for your subscribers. I’m not sure how five-star restaurants get their rating, but I suspect it’s a little different than email.

To come up with a member rating, MailChimp tracks open and click data and measures that against your sending frequency. Actually, we get a lot of questions about these little stars and how exactly a two star subscriber to becomes a five star subscriber. To find out how the sausage is made, we’ll need to look at the code behind member ratings.

Our five star system is based on a twenty point activity score, and we weight those activity scores unevenly. An even distribution might be good for my OCD, but loading activity points around four and five stars ensures that high engagement is easily remembered while low engagement is easily forgotten.

Handling Bounces

Have you noticed that new subscribers start out with two stars when they get added to your list? If so, you may be wondering why our activity score goes down to negative five. Don’t worry, your subscribers can’t earn a negative activity score by ignoring email. We reserve the negative range for addresses that soft bounce, hard bounce, or see spam complaints.

Actually, member rating gives us a cool opportunity. When an email address hard bounces, the ISP is telling us to never try sending to that address again. However, hard bounces are not always valid. If your subscriber has a positive member rating and we see a hard bounce, we’ll reset the activity score to zero instead of automatically removing the address. Cool, right?

Okay, I oversold that. It’s not “cool,” but it’s definitely better than the alternative.

Send Frequency?

In order for member ratings to accurately measure engagement, subscribers have to lose points for not engaging with your email. We handle that by decrementing activity scores after you send your campaign. Don’t panic! We don’t decrement these scores every time you send an email. Instead, we factor in your send frequency.

Why does it matter how frequently you send? Well, the law of demand says the quantity demanded and the value of a commodity are inversely related. In this case, your email is a commodity to your subscribers, and regardless of your content, the more frequently you send, the less value any single email has. To put it simply, ignoring a monthly email is a much bigger deal than ignoring a daily email. Here’s the schedule we use when decrementing activity scores…

So there you go. That’s the nitty-gritty on how Member Rating is calculated. Employees must wash hands, and preparation surfaces must be clear of contaminants. Keep in mind, this post is a snapshot of the member rating and activity score algorithms as of the posted date, and it’s entirely possible those algorithms could change going forward. Besides, do you really want to sit down and figure this out? Let MailChimp handle the dirty work.

Tofugu is a self-described “wonky blog about the Japanese language and culture” based out of Portland, but they’re more than that. The three gentlemen behind the site also help people learn Japanese via their online textbook, Textfugu, and their kanji learning application, WaniKani. “Learning a language is so hard,” Tofugu founder Koichi says. “Especially if you aren’t motivated to do so. Once you’re motivated, it’s a lot more fun and you really feel like you’re doing something great.”

To help Tofugu’s customers feel great, Koichi investigated MailChimp’s autoresponders. Inspired, he started crafting simple, motivational nuggets to send to new learners in the hopes they wouldn’t get discouraged. “The content is really varied,” he says. “Sometimes it’s a ‘get off your ass’ sort of email. Other times, it’s a ‘Did you know this is how people actually learn?’ sort of email. I try to write things that would motivate me.”

The first Tofugu autoresponder is sent a week after signup, and the second comes two weeks after that. Koichi says he likes to spread them out so as to not annoy the busy people who are trying to fit a new language into their lives. Also important is the fact that these emails aren’t selling anything—they’re encouraging someone to find happiness in a product they’ve already paid for, happiness in learning something new.

Sometimes, he’ll send an inspirational quote followed by a call to action:

Other times, Koichi presents an easy challenge:

Occasionally, Bruce Lee’s intensity is evoked to help motivate the learner:

How are they working? “We get around a 50% open rate, sometimes higher, which is well above the industry average,” Koichi happily reports. “I also get emails at the end of the autoresponders. Usually they’re sad people asking what happened and when they can expect the next one. Maybe I need to re-read the emails to motivate myself to write more of them for people!”

Koichi says email is the perfect medium for keeping Tofugu’s learners focused. “People read all of their emails much more consistently than they read their Twitter or Facebook timelines,” he says. “Email has allowed us to more effectively reach our learners, and I think it has helped a lot of people to stay motivated and learning, whether it’s with our products or someone else’s. MailChimp has made it possible to do something I wouldn’t have thought to do otherwise.”

Here’s a peek at New MailChimp’s campaign creation process and campaign dashboard:

If you’d like to see other previews of what’s coming, we’re releasing a series of blog posts and videos that go into a bit more detail found here: New MailChimp

When it comes to email development, the prevailing attitudes are confusion and frustration. We want to help change that. That’s what drove us to start this project, and now we have an HTML email reference to share with you.

MailChimp’s Email Template Reference is a repository of the things I’ve learned and information I’ve picked up over my years at MailChimp. If you’re a newbie where HTML email is concerned, you’ll find thoughts on design, development methods, and code samples—everything you need to get started designing and building great emails. If you’re an old hand at this stuff, you’ll benefit from the different techniques on display and some new responsive email blueprints.

This reference isn’t something that’s “done” and unchanging, though. We’ll be adding more resources, code, and ideas to it. It’ll be an organic, expansive guide that we’re hoping will kill some of the mystery that still lingers around HTML email. Consider this V1. We welcome any feedback or suggestions at templatereference [at] mailchimp [dot] com. Community is important to all of us at MailChimp, and your input can only make this reference better.

During the redesign, we ran rapid-fire usability tests on commonly used and radically changing features to get immediate feedback. For instance, we wanted to test our Campaign Builder on as many eyes and on as many screens as possible in one day to keep our design schedule on track. We ran around our office with a testing laptop, recording a mix of friends and coworkers as they put the redesign through its paces on laptops and iPads. After several hours of 5-10 minute test sessions, we compiled a prioritized list of bugs, usability issues, and specific changes for our designers to improve. Our UI design performed pretty solidly on standard screen resolutions, but we did find room for workflow improvement on the iPad.

Testing MailChimp on the iPad with Silverback and Reflector

Then we combed our secondary data sources to gauge customer feedback, user behaviors, site analytics, and data from usability tests and surveys we had already conducted (and there were a lot of them). This allowed us to better understand the types of jobs our customers are trying to do, and what features and functions we could emphasize to better empower them.

To aggregate our secondary research data, we use an Evernote for Business account. This allows us to quickly and accurately look at user feedback and previous studies to answer design and development-related questions, like, “What are our customers doing on the subscriber activity profile pages?”

In terms of our redesign, we considered a massive redo of our “View Subscriber” table. But when we combed through Evernote, the results from recent studies on how our users view and use Subscriber Profiles indicated that our redesign idea would disrupt their workflow as they verify subscriber information. They didn’t necessarily need a newly designed view—what they really needed was a better way to sort and scroll on the subscriber table. So until we can study this issue further, we’re going to keep the changes to these minor-but-helpful usability fixes.

New and improved subscriber table

In addition to all that jazz, we ran Card Sorts to verify Information Architecture changes and quick surveys to learn more about our users, like which list statistics are most important in their day-to-day workflow. We learned a lot about our customers’ priorities that way.

Quick survey results for important list statistics

It’s been exciting for the research team to work alongside our designers and developers throughout the process, and we love seeing MailChimp’s new workflow informed by what we know about our users and their workflows. Go Team Research!

If you’d like to see other previews of what’s coming, we’re releasing a series of blog posts and videos that go into a bit more detail found here: New MailChimp

And that’s cool. But what about asking the computer to use your data to decide how it would segment your list for you?

Every time we send a newsletter, we generate interactions. Specifically, clicks are a great source of data for understanding our subscribers’ different interests, and those interests can define segments. In this post, we’re going to investigate how to get Microsoft Excel (or LibreOffice if that’s your poison) to suggest list segments based on past clicks.

This is called data mining or unsupervised learning. It’s called data mining because we’re going to get the computer to dig into your click data and discover some interesting segments on its own. It’s called unsupervised learning, which is a branch of artificial intelligence, because we’re not going to give the computer any past examples of how we’ve segmented. We’ll let it choose what it feels is right.

What can you do with some computer detected segments? Two things:

1. You can dig into the segments it creates to try and understand why. We’re going to use click data in this post, so why would a computer separate these people who clicked these URLs from these other people who clicked these URLs? Cluster analysis can often divide your audience in ways you hadn’t thought about.

2. Even if you don’t understand the why, you can still test out the segments. If I have a list and a computer decides to divide it in two, even without knowing anything about the segments, any A/B test I do on content sent to these two segments will do as well or better when sent at the segment level than at the list level. Sure, both groups might like A better than B. But now you have an opportunity to distinguish a group that may like A better and a group that may like B better.

Let’s get started. But first, some notes:

This post will be a bit longer than most on the blog. I’d encourage you to work along with me in the example spreadsheet (available here for download). If you’re using Chrome and you click the link, Google may try to open the sheet in their Google Drive software rather than downloading. Select “Download Original” from the Google Drive spreadsheet file menu to get the file.

A bit about software: This works in Excel and LibreOffice, but for the love of all things holy, don’t use Google Drive. It won’t work. Google spreadsheets are garbage. Yes, I’m a spreadsheet primadonna, but they just can’t handle what we’re about to do. For the figures in this post, I’m using Excel 2011 for Mac, but this works just as easily in 07 and 10 for Windows. Haven’t tried 2013, but it should be fine. And if you’re using Excel 2003, that’s just sad.

Pulling down some input data

I’m going to work through an example of this technique using click data from The Next Web, which is one of the most influential tech blogs on the web. If you’ve never been over to TNW, check it out. They send blog updates through MailChimp each day, so their account is a data scientist’s idea of a playground. Of course they gave us permission to use their data here.

Now, why am I using click data instead of open data? Open data is, after all, more prevalent. Cluster analysis depends on the reliability of these interactions at an individual level, and since some mail clients can generate false opens (Apple Mail) while others don’t (Gmail), it’s best not to use interactions that are dubious indicators of individual interest.

So in order to cluster email addresses using click data, I’m going to pull down into a spreadsheet clicks from the last three months of daily emails. In MailChimp, you can export these clicks url-by-url by logging in and viewing the clicks under the subscriber activity section of each report. Alternatively, you can pull this data via the MailChimp API or by hitting the SQLite copy of your data created by Hairball. If you export CSVs from the MailChimp reports section of the website, you’ll need to open them up and copy paste them into a single spreadsheet.

For The Next Web, I’m going to lop off the long tail of the data and consider only the 40 best performing articles out of those they’ve sent in the last 90 days. What I end up with is a sheet in Excel that looks like so:

Note that in this list of who-clicked-what I’ve purposefully obfuscated the email addresses to protect TNW and their readership. I haven’t obfuscated the URLs, because I’m sure they’d be happy for you to click them. Also, I’ve removed the tracking code from the URLs (anything after the & in the URL), leaving just the original link.

Creating a matrix of click data

The first thing we’re going to do with this data is create click vectors for each email address. A click vector is just a list of 1s and 0s for an email address where a 1 means that they clicked a URL and 0 means they didn’t.

For instance, for my email address, let’s say I clicked google.com and news.ycombinator.com today but not excite.com (which sadly still exists). Then I could write that as [1,1,0] for the URLs in that order.

To do this in Excel, I can just highlight the first two columns of data in the “TNW_40″ tab I showed in the figure above and create a pivot table (Insert–>Pivot Table on Windows, Data–>Pivot Table on Mac) from them.

In the pivot table builder interface, I can place my email addresses in the rows, URLs in the columns, and counts of the URL instances by email in the values section. This gives the builder setup and table shown in the picture:

This is a pretty awesome presentation of click data. The differences in the rows expose the individuality of each address, and as we scroll through the table, outliers pop out faster than Rick Steves at an Insane Clown Posse concert.

For instance, there’s row 2483 (see below) where the email address clicks nearly everything sent. Turns out, the email address maps back to a domain without a website (just receives mail apparently) whose name servers point back to a marketing intelligence firm. They’re likely registering bots to popular email lists to scape content or measure inbox placement. Gee, thanks for that guys. Just FYI, your bot missed a few clicks.

Getting started with 2-Medians clustering

Let’s set up a sheet where we’re going to divide our list of clickers in 2 using a technique called k-Medians clustering using cosine distance. Scary name, but it’s not that bad.

Essentially, we’re going to create two archetypal click vectors, “cluster center 1″ and “cluster center 2,” and each email address will be assigned to whichever of these two clusters has the nearest center. These assignments are our two segments.

OK, so let’s create a tab called “TNW 2-Med” and copy-paste values of the pivot table into the top left of the tab. Let’s also add blank rows in row 2 and 3 to the sheet for our two cluster centers (right click row 2 and select “Insert” to insert a blank row).

This gives us the following sheet:

These cluster centers start off empty (empty cells are 0s in Excel) meaning they’re both defined by no clicks whatsoever.

Now, in columns AP and AQ to the right of each click vector row, let’s measure the distance between each email address and the two centers. In the end, each address will get assigned to whichever center is closer.

In order to understand our distance calculation, it’s best to first understand how to measure the cosine similarity between two click vectors. Cosine similarity is calculated as:

# of click matches between center and email address / (sqrt(email address’s total # of clicks) * sqrt(cluster center’s total # of clicks))

On the top of the fraction, we get points for every matched URL between the two vectors. On the bottom of the faction, we divide by the product of the square root of one click count times the square root of the other click count. When the two vectors are identical, this calculation is equal to 1 meaning “completely similar.” When the two vectors have no matches, this similarity value goes to 0. If they only overlap on some but not all, then we get a value in the middle.

The cool thing about this calculation is that it focuses on clicks rather than non-clicks, because let’s be honest, when someone clicks they’re showing more intentionality than when they don’t. There are a multitude of reasons for not clicking something (non-interest, didn’t understand the title, didn’t see the email, etc.), but a click is most often due to interest. At the same time, the denominator of this fraction penalizes vectors that click everything, essentially watering down the intentionality of every click (like that bot up there).

Since this is cosine similarity, that makes cosine distance simply:

1 – (# of click matches between center and email address / (sqrt(email address’s total # of clicks) * sqrt(cluster center’s total # of clicks)))

Yep, we just subtract the value from 1 to flip everything around. That way, when you’re identical to something, you have a distance of 0 that grows as your clicks begin to differ.

Then in cell AP4 in Excel we can calculate the distance between the first email address and the cluster in row 2 by writing out this distance formula as:

=IFERROR(1-SUMPRODUCT($B4:$AO4,$B$2:$AO$2)/(SQRT(SUM($B4:$AO4))*SQRT(SUM($B$2:$AO$2))),1)

The SUMPRODUCT of the two binary vectors just sums up matches. The whole distance calculation is wrapped in IFERROR to account for the fact that if a center were set to all 0s instead of creating a divide-by-zero error, we’ll just specify a maximum distance of 1. Note that I’ve used absolute references (“$”) so when I copy the formula to another row, the email address changes but the cluster center’s row stays put.

Similarly in AQ4 we can measure the distance to the second cluster center:

=IFERROR(1-SUMPRODUCT($B4:$AO4,$B$3:$AO$3)/(SQRT(SUM($B4:$AO4))*SQRT(SUM($B$3:$AO$3))),1)

In the next cell over, AR4, we can calculate the closest distance of the two centers as:

=MIN(AP4:AQ4)

And then in AS4 we can assign the email address to the closest of the two clusters using the MATCH formula:

=MATCH(AR4,AP4:AQ4,0)

This just spits out a “1″ or a “2″ depending on which of the two columns has the same value as the minimum.

We can then copy these cells, highlight AP5:AS2634, and paste them to send the calculations to the other email addresses:

Every distance starts at 1 (as far as it gets), because our centers are both zeroed out.

Let’s label cell AU1 as “Total distance” and below it in AU2 let’s total up the min distance column (AR) as:

=SUM(AR4:AR2634)

Initially, this sum is as bad as it gets: 2631.

Optimizing our two centers

This is where we get segmenting! We need to find the best 0-1 values for our cluster centers on rows 2 and 3. To optimize this problem we’re going to use Solver. Solver comes with Excel, but you have to turn it on. (2007, 2010, Mac 2011) Once solver is added in, it can be opened from the Data tab in the Excel ribbon.

Solver allows us to specify an optimization problem for Excel to solve. In this case, we want to find the best 0-1 vectors for our cluster centers that minimize the total distance in AU2. So opening up Solver, we specify AU2 as our objective and make sure we select “Min,” because we want to minimize distance. We’re going to change our cluster center cells $B$2:$AO$3, and we’re going to add a constraint that these cells remain binary.

To add a constraint, click Add, give the cell reference ($B$2:$AO$3) and select “Bin” from the drop-down before hitting OK. The solving methods given are algorithms for finding optimal solutions to optimization problems. Our particular problem is non-linear (the decisions are used in the distance calculations inside of denominators and square root calculations), so we can either select “GRG Nonlinear” or “Evolutionary” as our algorithms to solve. If you use the evolutionary solver, bump the maximum time without improvement up from 30 seconds to something more reasonable, like 5 minutes.

This gives the solver setup pictured below:

Pressing solve, we can watch Excel try things out until the distance falls to an optimal level. Now, since this is a non-linear problem, Excel can’t prove its solution is optimal. Feel free to run the problem again, maybe using the other algorithm. In my runs, I ended up at 1832:

Interpreting the results

As we scroll through the cluster assignments in column AR, we see that some emails go to cluster 1 and some to cluster 2. All right. Why?

One way to understand this is to count up the most popular URLs within each cluster. Scroll to the bottom of the sheet, and label rows 2635 and 2636 as “Count 1″ and “Count 2.”

Let’s start in B2635. In this cell we’d like to sum up all the clicks for our first URL by cluster 1 email addresses if indeed this URL is part of the cluster center definition. To do that, we use this formula:

=B2*SUMPRODUCT(B4:B2634,IF($AS4:$AS2634=1,1,0))

We have B2, which is whether or not the URL is part of the archetypal cluster vector, times the sum of all clicks of the URL from those whose assignment is to cluster 1. Now, unfortunately, this formula doesn’t run in Excel as-is, because the IF formula is operating on an entire range of values. This only works if we make the whole thing an array formula. To make that happen just press CONTROL + ENTER (CMD + RETURN on a Mac) when you put in the formula. This will make curly braces appear around the formula and now everything will work just fine.

The formula in B2636 is the same except we check the center for cluster 2 (B3), and in the IF statement, we look for emails assigned to cluster 2:

{=B3*SUMPRODUCT(B4:B2634,IF($AS4:$AS2634=2,1,0))}

Note that I’ve used absolute references on the cluster assignment column AS, so we can copy these two cells, select range C2635:AO2636 and paste them to get all of our counts. This gives the sheet shown below:

Now, let’s create a new tab called “2-Med Interpretation”. First thing first, let’s paste our URLs in the first column. Since they’re in a row on the previous tab, copy the URLs and then paste-special-values making sure to check off the “Transpose” option in the Paste Special dialog box. In columns B and C, do the same for the count values from the previous tab. This yields:

Applying Auto-filtering to the 3 columns and sorting by the counts in the first cluster, the top stories are all “shareables” (general human interest) or Google/Android related. In contrast, sorting by the counts in the second cluster, the top four stories are all iOS related. See the screenshot below:

Pretty cool right? One thing that TNW could try based on this is A/B testing subject lines per segment where Android vs. iOS is played up where relevant (such as in the AppMachine post).

Adding more clusters

And we don’t have to just do two segments. We can add another cluster center to the clustering sheet, tack on a third distance column, modify the solver formulation, and get some more counts at the bottom of the sheet. In the downloadable example workbook, I’ve added a “TNW 3-Med” sheet that does just that. So how do things change with three clusters?

Well, one cluster that the model chooses to separate out is a new “shareables” article cluster, defining a group of email addresses that mostly click the general interest stories (5 of the top 6 articles are “shareable” with the outlier being one about Google glass). See the screenshot below:

We could keep going. Vanilla Excel Solver allows for 200 decision variables and 200 constraints on non-linear problems. With 40 URLs and 3 clusters, we’re clocking in at 120 decision variables and 120 binary constraints. So we could go up to 5 clusters on this problem. The fewer the URLs you consider, the more clusters you can do. Note that the number of email addresses on your list does not affect the size of the problem.

Getting this data back in

Once we have some clusters we’d like to act on, we can copy/paste values of the email addresses and their assignments to a new sheet:

Auto-filtering the assignments, we can copy the addresses from the cluster we’re interested in. Then we can navigate to the Lists section of MailChimp, and under the List Tools menu we can create a static segment of our list by pasting this cluster of addresses into the static segment box:

We can save that segment off and then send whatever we want to it!

Wrapping up

OK, so I know that wasn’t the easiest worksheet you’ve ever put together, but I hope it demonstrates the power of your own campaign data, which can be mined by the computer even when it may be too large for you to understand with your own two eyeballs.

Feel free to try this out, and reach out if you find something interesting in your own data. And if you think this stuff is pretty awesome (that’s probably about two of you at this point), I’ve got a whole book on spreadsheet data science coming out this fall. Sign up for my newsletter to be notified about the pre-order.

Good night. Happy clustering.

For the first peek at New MailChimp, we’ll look at the restructured navigation and search options. Account notifications are now displayed in the nav, and you can switch to Pro mode to simplify the layout, or tap the hamburger icon to expand and collapse menus on the iPhone. Search is in every part of the nav, too, so you can search for subscriber details, campaigns, lists, and reports from anywhere in the app.

Here’s a look at New MailChimp’s navigation, search, and responsive design:

If you’d like to see other previews of what’s coming, we’re releasing a series of blog posts and videos that go into a bit more detail found here: New MailChimp

We did a lot of traveling to meet with customers, and spent hours in interviews learning how people use MailChimp. We compiled hundreds of pieces of feedback from customers and our support team, and conducted surveys with thousands of users. All of this helped us see some places where MailChimp was falling short, but more importantly, it helped us see bigger trends. We realized people are still doing the same kind of work they always have, but there’s been a shift in how they get that work done. Most people are trying to do more with less. They have a ton of responsibilities to address by the end of the day, many are accountable to a boss who tracks their work, and there are still just 24 hours in the day. They’re using mobile devices to get stuff done during what would otherwise be idle time. 9-5 just doesn’t cut the mustard anymore.

With limited time and resources, teamwork is more important than ever. 35% of our customers are working collaboratively today, and we see this statistic increasing steadily. People are passing the baton to colleagues when they’re unable to complete a task. They’re collaborating in order to do better work and get it done quickly.

So the requests for a mobile experience were just symptomatic of a bigger change. We’re all feeling extra pressure to get more things done these days, and because we can’t work any harder, we have to start working smarter. That’s why we’ve created a brand-new MailChimp that will hopefully help you do just that.

MailChimp will switch to a new look and enhanced features in June. We’ll show you our work over the next few weeks on the blog, so you can get familiar with the new look and understand why certain things changed. We want to help you work smarter, so you can focus on your customers and fans—not on our software.

Three goals have guided our design process. First, we want to make MailChimp ubiquitous, so it’s within reach whenever and wherever you need it. We also want to make the workflow more efficient, shaving seconds off of tasks, so you can get your work done and move on with life. And finally, we want to help teams collaborate so they can make better work and get consensus quickly.

Ubiquity

Though mobile devices are outselling PCs these days, they’re not replacing the desktop experience—they’re extending it. People start a task on one device and pick it back up on another. This new behavior pattern requires a different kind of design thinking that’s focused on consistency, optimization for different contexts, and clear continuation of workflows.

New MailChimp is responsive. Whether you access it from your desktop, laptop, or tablet, the layout will reflow to fit your screen. Although fitting an entire app as sophisticated as MailChimp into a tiny smartphone screen isn’t ideal, you can still fire off a campaign in a pinch. We’ve redesigned MailChimp Mobile to extend the desktop experience to smartphones, so you can manage your lists and track campaigns when you’re away from your desk.

New MailChimp is always ready to go whenever and wherever you are. For us, making MailChimp ubiquitous isn’t just about making it “mobile-friendly.” It’s about making it work in all contexts.

Efficiency

MailChimp’s new design is the product of countless iterations over many months. We love great design, but not if it comes at the expense of functionality. We started redesigning MailChimp by removing any visual elements that weren’t essential to getting things done. Extraneous design elements add cognitive load that gradually erodes usability and learnability. Strong information hierarchy and clarity about what can be done on a page help new users and veterans alike move more quickly through the app.

Ok, now I’ll take off my designer hat and share a few examples of specific changes that are going to shave time off your workflow.

We had to touch every page in MailChimp for this project. That gave us the opportunity to question previous design decisions to see what could be made more efficient.

The setup page in the Campaign Builder has a lot of options, but people rarely do anything more than set a campaign name, subject line, and reply-to email. So we’ve collapsed all the advanced options, keeping them within reach if you need them, but getting them out of the way of the typical workflow.

We also spent a lot of time making improvements to the subscriber list view, one of the most visited pages in MailChimp. We’ve made it easier to scroll through columns while still seeing email addresses, and now less important columns can be temporarily hidden from view, helping you focus on the information that’s most important.

We even made small efficiency improvements to the navigation. New users see an icon and label in the global navigation that guide them to the app’s key sections. Once you’ve got your bearings, you can simply collapse the nav bar, showing icons you’ll now recognize. That creates more space for you to do your work, which improves work speed.

These improvements just scratch the surface. Many small improvements to efficiency add up to minutes saved with each session, and hours over months and years. Though we can’t add more hours to your day, we can help you spend less time on things that aren’t important.

Collaboration

Over the past few releases, we’ve been steadily adding features to help teams work together. Multi-user accounts let you share your account with colleagues while keeping a handle on permissions. And collaborative campaign building lets you work in tandem as you write and design. We’ve also added comment threading to campaigns, so you can give feedback from inside a campaign or by replying to a test email from any device.

Change is Hard

We know that change is scary for a lot of people, and New MailChimp might take a little getting used to. Since rethinking MailChimp on a grand scale could be disruptive to millions of people, we’ve made an effort to avoid changing things for cosmetic indulgence alone. We really want to make our product serve your needs better. Digital lives are changing, work habits are changing, and we think MailChimp should change with you.

Starting in June, we’ll roll out New MailChimp as an optional upgrade for four weeks, so you have plenty of time to try it out when you’re not under deadline. We hope you like what you see, and we’re listening to your feedback.

If you’d like to get a head start on what’s coming, we’re releasing a series of blog posts and videos that go into a bit more detail found here: New MailChimp

One year ago today, MailChimp’s Lead Engineer went rogue and formed a little startup inside our company called Mandrill. In one short year, Mandrill has grown to over 55,000 active users. That’s 100X the total number of users they had in May 2012. Since launching, the amount of email they’re sending every month has grown 100X, and revenue has grown 200X. We’ve begun the process of tightly integrating Mandrill data into MailChimp, and Mandrill is innovating like–well, like a bunch of rogue MailChimp engineers. Every week it seems they’re launching new features, dropping prices to infuriatingly disruptive levels, adding servers in more continents, and shaving milliseconds off delivery times.

We considered getting a congratulatory cake for the Mandrill team today, but I’ve learned that if it’s anything “offline” it’s not real to them. Instead, we put together this little comic strip: Purfickt Bahnana

Mandrill’s still got a way to go before it matches MailChimp’s 3 million users and 5 billion emails per month level, but something tells me they’ve got some tricks up their sleeve to make their graphs turn all exponential on me. I’m told Chad, the man who is out to destroy MailChimp from within (there, I said it) will be posting a one year retrospective on the Mandrill blog soon. That’s a good blog to follow if you’d like to learn more about transactional email, or destroying MailChimp from within. Happy Birthday, Mandrill!

Let’s start with the new “Comments” tab in the drag-and-drop editor:

Amazing things happen under that little tab.

Just Reply to Comment

While you’re working on a campaign, open the new Comments tab:

You’ll notice a blue button to “send a test email” and start a conversation.

Your test emails can include a personal message now:

Which will appear at the top of your test email inside a gray bar:

Your recipients can simply reply to the test email:

and then their comment will automatically appear in the comment panel back inside MailChimp:

Completed Tasks

Did you notice the “Mark as Completed” link in the screenshot above? Yep, to help organize some of the comments and feedback you’ll get, you can mark them as completed, and we’ll move them to the bottom:

We’ll even add a little “check” icon, to validate what a productive person you are. Seriously. You deserve a raise.

Comment Notifications

If there are members of the team that should be notified every time a new comment is posted, you can specify them here:

Select the members that have been given access to the account (this pulls from the multi-user access levels we released back in v8):

Every time a new comment’s posted, they’ll get an email. In the future, we’ll offer the option to send notifications via our native MailChimp Mobile app.

The Way Teams Actually Work

As companies get larger, their email campaigns turn into team projects. And the teams are diverse. You have designers, writers, marketers, and management all working together to create content, select customer lists, approve messaging, and then some brave soul has to push that final “send” button. This dream team is almost never in the same room at the same time. It’s almost impossible to get them all to review an email at once. So that’s why we focused on building for asynchronous collaboration. In other words, co-workers reply via email whenever they have the  time—from work, from home, or on the go.

Inline Collaboration Too

Of course, if they so happen to be in the office at the same time as you, they can also sign in to MailChimp and comment back and forth with you (like a hovering art director) in real time:

Neapolitan now lets you know when others are currently editing.

The editor now shows you who is currently editing a campaign.

Their edits happen in real time and are color coded, so you can see what they changed:

Either way you choose to work together, MailChimp works the way you work. We’ve got even more tricks up our sleeve coming in the next few releases—ones we think will help teams design and collaborate around their campaigns even faster.