Today the amount of available marketing data is huge. And thus, people tend to get confused about how to use data the right way.
Using data efficiently is the basis for data-driven marketing. And there are tools like Analytics maturity model that help companies assess how advanced they are with their marketing analytics. How do you track your KPIs? Do you use data to make predictions of the future?
To help you move forward with your analytics, we want to give you examples of different levels of questions that can be answered with data. In a nutshell, data can give answers either about the past or the future. In this blog post I will present simple KPI summaries (which show what has happened in the past) and move on to more advanced analytics (which aims to predicts the future) in my next blog post.
Traditional business analytics measures the business through descriptive summary statistics and relevant KPIs, such as average lead cost rate or conversion rate, which tell us what happened in the past and what is the current state of the business. These KPIs are monitored through automated reports and dashboards, which help you to track down how the business is going with questions like: How many of the customers convert? How does the conversion behave compared to the previous quarter?
If you work as an analyst you may also want to drill-down into more fine-tuned subpopulations and examine the data from multiple different angles. One tool for this is to draw more complex visualization, for example compare sales versus profit by country, or ask questions like: Does the conversion behave differently between female and male customers? What is the conversion rate among young adults in our top selling region? How does profit and sales behave within product category and sales quantity?
Multi-plot grids help you to gain insight into interesting relationships and phenomena. Above is an example of how one can examine the data from multiple angles using multi-plot grids. We have visualized sales, profit and quantity broken down by product category.
This stage utilizes the power of diagnostic statistics to have a more sophisticated understanding of the past: you can drill-down, examine distributions, draw reference lines like medians, calculate trends and forecasts with confidence intervals. Testing hypotheses helps you to put KPIs into context.
Testing hypothesis: Is the increase/decrease on conversion due to a random variation? Is the change truly significant?
Examine distributions: What is the typical conversion and how much does it deviate?
For example, the value of a single purchase varies. Most probably the purchased item is going to be affordable but sometimes also more expensive items are bought. This means that the data distribution is left-skewed. Hence, the mean does not perfectly represent a typical value of a single purchase. Below you can see a sales amount distribution which is left-skewed and therefore does not perfectly represent a typical sales amount.
Calculate correlations: Which characteristics correlate with conversion?
In the correlation matrix below we can see that the shipping cost correlates with the amount of sales.
Examine trends and forecasts: Is the conversion within our assumption? Based on past events, is the value something we would expect?
How well does your company utilize data to analyze the past? Can you see patterns in the data and how do you draw keys learnings from it? If you are ready to move towards more advanced analytics and predict the future with data read our next blog post where we will explain how data and machine learning can be used for better decision making.
Mirva is a marketing data professional in Tieto’s Customer Experience Management business unit. Mirva helps brands and organizations develop data-driven digital and customer engagement transformation that enables businesses to adapt to the era of the customer. Mirva has a strong understanding of business processes and visual analytics, combined with theoretical background in computer science and mathematics.