Digital marketing tech industry continues to fascinate me even though the segment is getting saturated with software vendors of all kinds. Lately, I am spending more time on data science as you can see from my recent posts. However, I feel it is important to discuss the marketing technology spectrum before we discuss how marketing can benefit from data science.
If you follow the growth in marketing technology, it’s easy to get lost in the ocean of new technologies.
There is an application for almost everything and everyday we see a new solution. This certainly creates need for two new marketing roles: marketing technologist and marketing data management. It is possible to lump both skill set into one role, although there are some unique differences in these roles.
Marketing Technology Director (median salary: 150k on Glassdoor) :This term is quickly becoming a buzz word in the industry. I did a quick search on Google Trends for the following terms: marketing technologist, marketing technology, marketing technology job, marketing IT and the only promising one was the keyword “marketing technology job” in terms of the future trend.
Google Trends for keyword “marketing technology job”
Marketing technologist profile looks like a hybrid of a traditional IT person with deep marketing skill set. Ideally, this is a candidate who has to spend most of their careers in marketing dealing with technology, infrastructure, integrations and setup. I don’t think this role requires a person with only core IT background. The extensive knowledge of marketing is extremely important.
Now let’s look at the marketing data management role which I strongly feel is different.
Director of Marketing Data (medium salary: 200K on Glassdoor) : We all know the volume of marketing data is increasing rapidly and there is a great need to make sense of the enormous database. Management of the marketing data is becoming a sizeable challenge for most organizations who rely on business intelligence (BI) or IT. BI and IT have their own priorities and depending on how your organization is setup marketing data management may not be a top priority for them. I have had situations where marketing data was sometimes in the top 5 list of BI and sometimes not.
The need for someone to manage the flow, integration and usage of marketing data is critical. Interestingly this role is also a hybrid of analytics/BI and marketing skill set. An ideal candidate could be an experienced business intelligence person who has to spend most of their time helping marketing teams and understand the practice of modern marketing. A marketing analyst who has strong data skill set can also vouch for the role.
Here is a Google trend view of keyword “marketing data manager”and the future for this role looks promising.
Google Trends for keyword “marketing data manager”
What we have discussed so far is a nice preparation to broaden our thoughts form the traditional digital marketing aspects (SEO, web analytics, marketing automation, SEM, web design/UX).
Now let’s move on and discuss how digital marketing can further benefit from advanced analytics and data science methodologies. When we think of marketing analytics, we generally refer to digital, social or marketing automation analytics.
I think this is great although time has come for marketers to up their game and take a non traditional approach to data analytics. In the rest of the post, we will discuss ten different ways on how you can take advantage of data science. The best part is we will be using completely free tools to perform the analysis.
Tools of the trade
As I said earlier we will be using free tools to perform analysis. The best way to apply data science techniques to dataset is to get familiarize yourself with R programming. I know this may freak out an average digital marketer but our goal is to advance the digital marketing analytics game from basic web analytics to predictive analytics and statistics.
Let me assure you R programming is not as complicated as writing applications in Java or C ++. While R is originated from these languages, the creators have designed the language with statisticians in mind instead of software developers. You can also use Python to do similar data analysis if you are comfortable learning python. Now don’t worry if programming is not your cup of tea. I will also show you how to do some of the analysis using non programming tools. Plus there are many free courses available to learn R in few months if not weeks listed in the resource section below.
Here is a list of the tools:
1. R studio: This is a free package for windows or mac users to run R on your machine. You can download and install the basic R package. The standard R package can be super charged with a bunch of functions for different types if analytics such as text, maps or cart. You can install these functions on ad-hoc basis, and we will discuss this later.
2. IBM Watson Analytics: I wrote a post on Watson few months ago on how Watson can be used to predict the outcome of the cricket world cup tournament. You can create a free Watson Analytics at Watsonanalytics.com. IBM has made it simple for non programmers by developing a predictive analytics solution that does not require programming language.
3. BigML: BigML is another popular tool to run prediction without using a programming language. It has a simple point and click UI which allows anyone with limited statistic knowledge to predict like a professional.
4. Excel: Now excel is not free (Libre Office is a free alternative), but I am sure as a digital marketer/analyst you are already an excel junkie, so there is no new investment required.
So let’s get started with the analysis. We will discuss ten different use cases, and I have lumped them into four different categories:
I have divided the post into two parts for the ease of reading, and in this part 1 we will cover the content marketing prediction for four content types.
Category #1. Content marketing: You will be able to predict which content has the potential of going viral or becoming popular, including the following content formats:
3. Pintrest images
Category #2. Marketing success: You can predict the success of the marketing campaigns, including (coming soon in part 2)
5. Digital order prediction
6. Paid search optimization
7. Banner optimization
8. Email optimization
Category 3. Customer Analysis: You can analyze customer data for churn, growth and retention (coming soon in part 2)
9. Churn prediction
10. Customer segmentation and clustering
Let’s get started with the first category.
Category#1 Content marketing: Traditionally creating a winning piece of content is the very artistic process even in the most sophisticated marketing organizations. It starts with a brainstorming session involving content writers, subject matter experts and social media teams. The group may or may not use some basic web analytics. The outcome of this meeting will be a list of topics the content writers can use to write the content.
We are going to chance this and add a very scientific approach to creating a successful content piece. How can we do that? We can do that because we will be using the power of statistical analysis to allow us to predict the key areas that leads to the popularity of content.
To predict the popularity of the content (blog post, tweets, video, images) we will use IBM WatsonAnalytics so make sure you create a free account right away. To get started with predicting which content assets have the chance of becoming popular or viral we need to prepare the data.
First, you need a list of your historical assets that have already been successful, or you need some third party list of successful assets. You can also use an intern or someone from Odesk (upwork.com) to collect a list of all blog post or tweets for your topic that has been popular in the last 6 months. Another option is to use tools like Buzzsumo.com or Google Chrome Scaper plugin.
You will need all the key content information such as the title, category, name, abstract depending on the type of asset list you are putting together. You just need the title, abstract, and some other information. Believe me once you complete this step it will be massive time saver, and you can utilize your content resources efficiently. If you are predicting the popularity of a blog post, your final .csv file should look like this.
Content prediction csv file
In the file above the dependent variable will is the “Popular” column and the criterion used is Popular = 1 if the content has more than 1,000 social media shares. The independent variables are something that influences the dependent variable, i.e. headline, category, abstract, wordcount, and snippet.
Next, upload the csv file to Watsonanalytics.com. Watson will give you a score and from my experience a score or 60+ is good. Then you will be asked to define the dependent variable which in our case is the “Popular” column.
Select the prediction criteria on Watsonanalytics.com
Finally, you can run your prediction and here is how your prediction screen will look like:
WatsonAnalytics prediction dashboard
Watson allows you to customize the number independent variable you would like to use to improve the quality of your prediction. You can use one, two or all independent variables ( I always go with all as this is more accurate in most cases when using Watson).
In the example above, Newsdesk (category), wordcount and section name have the most influence on the content popularity of the blog posts. The combination of the section name and the blog category has 90% predictive strength, which is pretty good.
Now you can also identify the key themes for each of these independent variables. When I drill down to category, I notice technology is a top category, and 800 is the optimal wordcount.
WatsonAnalytics prediction table
Similarly, you can identify the winning titles/headlines and other details, or you can use the process to predict the popularity of tweets, social updates, Pintrest images and videos. Forget brainstorm meetings as with this type of analysis you can advance your content game significantly. You now have predictive power in your hands to improve your content marketing results.
In the part 2 of this post, we will use R programming to predict marketing success and BigML.com to predict customer analytics, including churn analysis. Subscribe to my blog so you don’t miss the next release.