Data Analytics for Digital Marketers - 2.0 (The basics)

Hope you all had a relaxed Thanksgiving!

Now that you have the SQL basics under your belt, we will learn to drive insights from historical data and predict performance which in turn will help us make decisions. The industry term for this process is Data Driven Marketing.

I will discuss this process in 3 steps:

Analyze: Analyze historical data from organization as well as market and provide insights.

Predict: Forecast or predict performance based on historical data.

Decide: Make decision i.e. take action to invest in relevant marketing programs and campaigns.

Data driven marketing is an iterative process. i.e. once you reach the third step of “Decide” or “Take Actions”, you go back to step 1 and repeat these steps again to address changes(market as well as organization) and optimize. One can write a book in each of these steps but today I will dedicate couple of paragraphs to each.

Analyze:

The very first step of data analysis is data exploration. In most companies, your data engineer will have the data in reporting database for you to analyze. In some scenarios though, especially in startups, when you start a new marketing campaign you may have to work with your data engineer to ensure proper tracking of the campaign level insights(acquisition,revenue,retention).

Once we have the data ready, we summarize or aggregate data to answer some questions that is important for our business. We want to know how much revenue has been generated on a monthly or quarterly basis. We will also want to know how much does a customer spend on average and so on.

When a company is launching a new product, they will be interested in knowing revenue share of the new product with respect to total revenue or how this product revenue compares to the other products by the same company. It is also very important to compare the performance of the product with respect to the industry and its major players. For example, if the product market is growing at 30% and the newly launched product is growing at 15%, they may want to revisit their marketing campaigns given the new product itself is competitive. These questions will vary depending on the business objectives. And the quantitative measure of these key business objectives are known as Key Performance Indicator.

Predict:

Forecasting or predicting is the process of estimating future performance based on historical data. As we already know, it is impossible to predict a future event with 100% certainty; be it weather, election, company revenue or customer retention. So, what we do is make an estimate, calculate likelihood or probability of an event outcome. There are various statistical algorithms already available to address different scenarios and one needs to understand these and use them accordingly. Many of these statistical algorithms are highly complex and can read data minutely that is impossible for a human eye to catch. Based on these readings an algorithm discovers a pattern and use that pattern to predict future events. This process of reading, discovering and estimating data is popularly known as Machine Learning.

Machine Learning is at the core of intelligent data analysis also known as Data Science, a term very recently popularized, thanks to data revolution. It is a field of study that combines various traditional disciplines (Mathematics, Statistics, Computer Science, Business) along with Industry specific knowledge to extract insights from data.

Decide:

The sole purpose of the previous steps (Analyze and Predict) is to guide us to make decisions and drive future investments and campaigns in a more analytical way rather than solely depending on our intuition. For digital marketers, the decision is to build a marketing strategy to channel marketing investments towards different marketing vehicles in rewarding markets and products. Popular digital marketing vehicles include advertising, email campaigns, promotions, incentives, website, social media as well as online customer support community.

In the coming weeks, I will discuss each of the above 3 steps for a specific marketing operations and how it can benefit from data driven marketing. I will start with customer segmentation.

Customer segmentation is act of grouping customer into groups of individuals that are similar in multiple ways relevant to marketing. One group, for example, can be women in mid thirties who spend more than $1000 couple of time. We use unsupervised machine learning model to form clusters. This technique is known as clustering. Once we find a way to name these clusters and find what works for them, we can market to each of these groups separately. As we acquire new customers, we can use supervised classification models to assign new customer to these groups.

That is all for now. I will be back in a couple of weeks with my next post.