Unrefined data is not any different from unrefined oil. Anyone with a moderate budget these days can collect huge amounts of data, but collection in itself should never be the end goal. The key is to extract meanings out of the collected data.
At the pinnacle of this data refinement process sits what we simply call “analytics”. If you work with data or marketing, I’m sure you hear that word at least a few times a day. However, it is apparent to me that everyone uses the word with different meanings. That is really unfortunate as it further confuses marketers who are already lost in a sea of data, and worse, it unnecessarily complicates matters for everyone who tries to rescue them.
For clarification and simpler dialogues among all involved parties, allow me to illustrate 4 major types of analytics:
- “Business Intelligence”, or BI reporting: This is all about knowing what is going on, the more real-time the better. What is the past 30-day response rate and conversion rate for a specific campaign? How did the response curve move through time? When is the optimal daypart for email blast or ad broadcast? What is the ROI? What channel and product offering is the winning combination? This can be in a form of dashboard reporting or any other conventional reporting, but many call it simply “analytics”.
- “Descriptive Analytics”: How many different segments of buyers are we dealing with? Where are they, and what do they look like? How do high value customers differ from barnacles? What are they interested in? How about income, age, number of children, occupation, and regional break down? At times we may use terms such as profiling, segmentation, or clustering, and they fall under descriptive analytics.
- “Predictive Analytics”: This becomes useful when we start asking questions in future tenses. Who will respond to this campaign, and for what product and through what channel? What are the potential values of each customer and prospect? Who will stop subscription of your service, and when would that be? When it comes to predictive analytics, we need carefully structured statistical models, which will return “scores” that define likelihood of customers behaving a certain way in the future. In terms of complexity, this is the most demanding type of analytics, where trained statisticians work with specifically designed marketing databases with all kinds of custom variables.
- Optimization: This type of analytics also requires a complex type of modeling, where “what if” type of questions are answered. What if we spend more money on mass media than on direct channel? What would be the most optimal combination of marketing spending that yields maximum return? What would be the ultimate ROI? This type of question is typically answered by marketing agencies, and it involves econometrics modeling. This type of analytics calls for different types of data in comparison to typical predictive modeling for 1-to-1 marketing, but the whole process is also called analytics.
So, before marketers jump into a big data discussion and shell out a great deal of money, I suggest coming up with a list of questions to be answered in the first place. After all, analytics is all about making the best of the data that we have, and there are plenty of sources near and far, small and big.