Skip to main content

A-B Testing Intro

Lesson 20 from: Email Marketing Fundamentals Using Mailchimp

Jon Chang

buy this class

$00

$00
Sale Ends Soon!

starting under

$13/month*

Unlock this classplus 2200+ more >

Lesson Info

20. A-B Testing Intro

Lesson Info

A-B Testing Intro

let's get into the world of testing. The most common form is called A B. Testing and that's what we're going to talk about right now. There's also something called multi variant testing zoom out. What is this stuff when you test, what you do is you create multiple versions of the same content and you randomly assign it to people to view. And at the end you'll be able to see which piece of content is performing the best for your overall user base. That's why you have A. And B. A. Is one version B is another version. Multi variant tests. On the other hand it's like an A. B. Test but with more variables. So instead of just A. And B, you also potentially have C. D. E and so forth. It's all up to you and how much you can actually test. So how do you decide how much you can test? Let's start there. It depends on how many people are receiving the email. Just like we decide what kinds of emails we should send. If you have 100 people, you may not be able to test 100 different variations because...

each person would get a different variation. There's a bigger science behind it. But essentially what you're really looking for is a sample size of 200 For each different version. Try to at least have 200 different people who have the opportunity to engage with it. So a B testing, let's start there. What we're going to look at is a series of templates where you can decide where each version goes and how to measure it. This is going to require you to refer back to our analytics part and implement U. T M. Code to track each individual version testing is so important for email marketing. Honestly it's really just important for all kinds of marketing. Digital marketing is really great because it's like the Wild Wild west almost you are able to try whatever but in this world because it's a digital environment, you're able to get results very quickly and also iterate quickly. So A B and testing means you have multiple variables in this sense a good test has a control variable. It's what you will normally send out. B is a version of it that has one particular variable different such as a question. Perhaps you were testing. Well what subject line is going to perform bust a question or a statement. I might also be geographic location. We assume that people in Chicago will respond better to a question than people in new york and then the end is kind of a catch all for multi variant testing. N is the symbol for a variable in general and you can have as many variables as you want. The only catch is that you need a large enough sample size. The sample size really depends but if you're just looking for a single number to go off of then I would say try to have 200 people per variable. So what are the kinds of things you need to measure when you test? So you need the send number. So 200, let's just start there ideally a lot more people just in case some don't actually engage then delivered of all that you're sending all will be delivered because of bounce rates. Perhaps it's a wrong email address or it's an inactive email address or there, they just have a really strong filter. So that depends on each email client. Some email clients filter harder than others and you have open rate, which is the quantity of opens divided by the quantity of deliverers. Why do we use delivered instead of sent? Well, we need to actually pay attention to the numerator and denominator separately. The numerator here is your key performance indicator which isn't open. The denominator in this example is your opportunity size. So if the user never had an opportunity to see the preview text, see the sender name, see the sender email address or see the subject line. Most importantly, then why would we actually measure it? So that's why it's the total number delivered rather than the total number sent. And then finally we have what is arguably the second or even first most important metric. It is the click to open rate, You'll also see something called click through rate. They're kind of synonymous but technically they're actually different. They're only synonymous colloquially click the rate technically is the quantity of clicks divided by the quantity delivered. Where as click to open rate is the quantity of clicks divided by the quantity of opens and in the same spirit that the denominator here is opens because if the user never opened the email then they never had the opportunity to judge whether or not they clicked it because they can't click through. If they don't open the email, that's the basic premise here. So when you're testing these are the four main metrics you need to pay attention to and you'll interchange the metrics based on your primary objective. So if you are testing a subject line, you are testing in looking at open rate. If you are testing images in the email, you then look at click to open rate. So I've laid out a series of examples here and as we go through, you can actually see that we have just general blocks. These are your modules, we've already looked through modules within the email, marketing, dashboard and templates. So this should look semi familiar overall here. You have A and B. Because a B test in a head might have one type of content and be would have a different kind of content. And if I say that we are testing open rate, then you know that perhaps it is sub decline. So what are pro tips? It's run variable at a time? two, you have the large enough sample size three, you were testing concepts not content. This is really important because it relates to one variable, especially a lot of people when they test they'll say well, you know, I just want to test this, these different kinds of videos, which is great. But essentially instead of videos it should ladder up to a bigger idea such as we believe that rich content is better at convincing users to click next. You have data should determine results, not preference. There's something called confirmation bias. And essentially when we see the results that we want, we tend to run with those so try and keep your objective data analyst and email marketing hat on and then use tracking because you actually need to see how people interact past this experience and on your actual properties. So google analytics is still a really great free tool and like we've seen already it integrates with male chimp so easily, literally a click each time. How do you use these templates? This is how I break it out. You have all the modules that you might need and of course play with this however you want. You have this one has a top large module, Then two column modules and the bottom large module. And this is the information that you should generally use. You have a description of the image, the actual call to actions. So one will be order now and then the other one is c full menu and then the copy essentially you don't need anything else unless you get a little fancier. But for this initial example let's assume it's really basic email and these are primary components. So if we have all of this, what are we testing it against? Module placement is a very standard test such as well. We're gonna put the salad up to the left, we're gonna put the sandwich up to the right in this example the blog and social media and we assume that the order of information will dictate whether or not user click through because we know that the primary information that allows them to make their decision is above the fold or what is initially viewable when they open the email. So these examples are 100% for you to be able to complete the upcoming checkpoint. What I'd like you to do now is look at that email marketing worksheet that you downloaded earlier and in it you can see the email testing framework. So it looks just like the slides that I had shown you and it has a little additions like hypothesis testing, variable and overall timing. This aligns with the testing framework you will use within the male chimp dashboard when we stage this. So what I'd like you to do is tear this apart. It's making a modular testing framework is incredibly hard for every type of organization in the world. There's really no catch all because the industry, the type of email, the type of user will all dictate how many modules you will use in their placement. However, this layout is what you saw in the slides, so it just aligns there and that's the reason. First of all we have that hypothesis. This is synonymous with the concept. We are not saying we want this video, we are saying that rich content is a better indication to the user that they should participate. The test variable then would be the actual variables as like a video timing. We haven't really discussed this one as heavily yet, but what you want to do is make sure that all these tests, unless it is a time test, meaning you want to see what time people are best going to engage with. Your email. Time should be constant. In 99% of tests, time should be constant. So make sure that you note when you will actually send this email such as Tuesday morning, this will all go out. Everyone has the equal opportunity to click it at the same time and then we'll be able to say this version was definitively the best version because they all had the opportunity

Class Materials

Bonus Materials with Purchase

Example Data Set
Worksheet

RELATED ARTICLES

RELATED ARTICLES