Ashwin asks, “I have data around me related to campaigns – emails, phone calls, and survey data – so how should I start analyzing it?”
First, let’s recall that marketing data science is about the application of the scientific method to data. Our first priority, then, is to use the data we have, techniques like KPI mapping, and some exploratory techniques to build the hypothesis we want to test. Watch the video for full details.
Can’t see anything? Watch it on YouTube here.
Listen to the audio here:
- Got a question for You Ask, I’ll Answer? Submit it here!
- Subscribe to my weekly newsletter for more useful marketing tips.
- Find older episodes of You Ask, I Answer on my YouTube channel.
- Need help with your company’s data and analytics? Let me know!
- Join my free Slack group for marketers interested in analytics!
What follows is an AI-generated transcript. The transcript may contain errors and is not a substitute for watching the video.
In today’s episode Ashwin asks, I have data around me related to campaign emails, phone calls and survey data.
So how should I start analyzing it? Nope.
First let’s recall that the the entire point of marketing data science is to apply the scientific method to data right to marketing data, in particular for doing marketing data science.
And therefore, we need to start the scientific method, the scientific method begins with a question, then you define the problem and the data you need.
And you build a hypothesis that building on that hypothesis is the first start.
So before we can go any further, we need to understand the question and we need to define the the data we’re going to need.
So if you have all this data laying around, that’s great.
What is the problem What is the question that we’re trying to ask it trying to answer? Are we trying to find out what drives revenue, drivers leads, builds, loyalty reduces complaints, there’s all these different outcomes that we want to know.
most marketers have at least some role in creating demand that results in either lead generation or sales of some kind of revenue targets.
So that’s the place that you normally would start and you do an exercise that we call KPI mapping, where you take that ultimate end goal and then map backwards.
The steps you need to take in order to find a metric that is within your measurement sphere.
So you might care about revenue and revenue comes from in a b2b example, closed one deals in the b2c example would be you know, consumer sales.
And from there okay, what drives that number? Well, if closed one deals are created by Open deals, sales are preceded by say shopping carts, filled shopping carts.
And you just keep working your way back until you get to a metric that you have control over.
You have visibility into and you understand what influences it.
And from there, you define what other data do you have? And do you need any other data.
So let’s say your b2b marketer, your key metric, the KPI that you probably care about really is going to be that lead generation.
And if you’re a b2c marketer, you’re probably care very much about shopping carts filled on the e commerce side or maybe people who walk in your store, whatever the case may be, find that number.
And then you build a hypothesis.
If you say that The green button on our website would perform better as a red button.
And you believe that a change to the red button would give you an increase of 5% in your shopping carts filled.
Now you have a provably true or false statement, right? That is a singular condition that you can go and test.
And that’s that right there is the heart of data science is I’ve come up with a hypothesis.
Now I need to use the scientific method to test it out to prove or disprove that statement.
To go and change the button color, you collect data, you analyze it and you find out the red button has no lift sales, okay? You’ve disprove them hypothesis.
So the question is when you have all this data, emails and phone calls and survey data, what is it you’re trying to prove? Chances are and this is something you get to in that different Fine stage early on, you’re not going to use all that information all the time, you may only use some of the pieces of data, you may want to do a deep dive into a particular data pools if you have all these emails, great, what influences open rate of your emails, time of day, day of week, domain name of the email you’re sending it to.
So in that defined stage of the scientific method, you’re going to be doing a lot of what’s called exploratory data analysis where you’re going to engineer out all those features that essentially would be like extra columns in the spreadsheet from the data to try and come to a point where you can develop a hypothesis.
So the question is, what drives emails? Is it the length of the subject line? Is it who it sent from Think of all the characteristics of an email that just sitting in your inbox, all the different ways you could analyze it? And then, being based on that corporate hypothesis, you hypothesize that sending emails on Sunday results in a 15% greater open rate than sending email on Tuesdays.
And it has to about point you can begin the testing process because you have a provably true or false singular condition that you’re testing.
It’s important to say that that singular condition is vital for good testing, because if you’re trying to test more than one variable, same time, it gets really messy, it gets complicated you can be done.
There’s a technique called multivariate testing, but you need really good software for it and you need a statistically relevant sample size of your of your population.
If you have, say, an email list of, say, 1000 people, you’re gonna have to test with about half of that list just to get any kind of margin of error, that’s reasonable.
You may just have to test the whole list.
And so it becomes if you have a small sample size, it can become very, very difficult.
So Decide what you’re trying to test.
Right? build that question, define it, do your exploratory data analysis, come up with your hypothesis? Very often what will happen is that the first time run, you do go through even just an exploratory phase, you’re going to end up with like 10 or 15 different hypotheses, you’re going to end up with a whole bunch of stuff.
And the question for you will then be, okay, well, what is the most important and that’s that that’s the reason why we do the KPI mapping early on so that you can figure out what is closest to the end of the chain of evidence for revenue, right? If you’re testing audience metrics, but your boss cares about revenue, you and your boss is so far apart on on different metrics that while your investigation and your proof may be valuable, it will not have the impact that your Boss is looking for what your company is looking for your stakeholders.
So make sure that you use that KPI mapping to prioritize either the most important thing to the company, which is pretty much always going to be in revenue, at least in normal business, or you prioritize the thing that’s most broken thing that is just working lease.
Well, that’s, that’s where you’re going to spend your time.
So figure that out, figure out how to figure out what it is that is the priority to test it and then go with that really good question.
Challenging lots to unpack lots of experimentation do that’s the other thing with data science, you’re doing a lot of experimentation.
So give yourself some time.
Give yourself some resources and, and be patient.
As always, you have fought questions please leave them in the follow up box below.
Please subscribe to the YouTube channel on the newsletter.
I’ll talk to you soon.
What helps solving your company’s data analytics and digital marketing problems? This is Trust insights.ai Today and let us know how we can help you
You might also enjoy:
- How to Set Your Public Speaking Fee
- The Evolution of the Data-Driven Company
- Best Practices for Public Speaking Pages
- How to Set Your Consultant Billing Rates
Want to read more like this from Christopher Penn? Get updates here:
Get your copy of AI For Marketers (2019 Edition)