How to Derive Insights from a Mess of Data >

In a recent blog post, we explained just how important reporting is to any marketing campaign. All the numbers and data collected from marketing efforts tell a story, but often there is too much noise to find the important information. In email marketing specifically, hundreds of thousands of data points are created with each mailing. Not all of this data is actually useful and it is near impossible to derive strategies from it when it’s still just a jumble of numbers.

Given the importance of understanding the data, the sheer amount of data each campaign creates and the limited time we have to make decisions, it can seem like there is too much data coming in too frequently to find insights. In reality, this big data problem is growing rather than shrinking as data points are created with every human interaction.

There is no simple solution to this problem and all possible solutions are going to be multi-step processes. At BrightWave, we adopt the following process in order to understand raw data.  


Step 1: Clean the Data

First and foremost, all newly collected data needs to be cleaned. This means that there should be no missing values, no duplicates records and that everything is in the same format. For example, all states should be stated in the same notation. Rather than having one record say ‘Georgia’ and another say ‘GA’ and yet another say ‘G.A.’ all should be changed to one common format. The consistency, continuity, and clarity of the data hugely impacts the resulting marketing strategies.


Step 2: Find Relationships and Trends

Once the data is prepared, we begin to manipulate it to find new insights. The simplest way to find relationships between data points is to display the data in different ways. Plotting two variables against one another is an easy graphical representation to find trends, groups, and outliers. Graphs can also show month over month data changes. Tables are another visual way to sort the data for deeper understanding. Among other functions, these tables can be sorted to find the largest or smallest value very quickly.


Step 3: Calculate Statistics

After using graphs and tables, formulating descriptive statistics is the next step in data understanding. Statistic metrics such as the mean, median, and mode, as well as the range and standard deviation, can be useful in understanding the data set. These numbers give a benchmark against which you can compare new data points. For example, If we know that the mean open rate is 25% and our most recent mailing had a 35% open rate, we know that something about the most recent email was more interesting to customers.


Step 4: Build Advanced Analytics Models

Finally, we run more advanced analytic tools on the data. These tools can include regression, clustering, decision trees, and many more data mining algorithms. With more advanced models, the data becomes much more powerful. These tools are the first step towards machine learning which makes the analysis of the data even easier.


Step 5: Repeat!

Of course, the data collection and processing is an iterative process that doesn’t end after a specific point. There are always new models to run, new observations to find, and new strategies to form. Every time you receive new data from your marketing efforts is an opportunity to gain new insights and make new improvements to your next campaign.


While this process is by no means easy, it does provide a clearer image of the end customer. Once we know the customer better, it is easier to market to them more efficiently. Our best advice to finding these insights? Pull the data and start processing.


Leave a Reply
(will not be published)