In the not-so-distant past, the standard methodology in marketing analytics was to input raw data variables into statistical programs to generate marketing insights. This approach was a core element of what is referred to as CRM 2.0.
For example, customer data variables were uploaded into cluster analysis programs to identify market segments. Customer purchase data plus demographic or firmographic data were read into regression programs to estimate customer lifetime value. New lead datasets were compared to current customer characteristics to score leads or estimate their future value.
The common denominator for CRM 2.0 platforms was the raw data variable. Each analysis started from scratch with raw data. This consumed considerable time and resources. What if you could skip the raw data starting point and start marketing analysis with smart data? This is one of the main features of CRM 3.0 platforms: they let marketers work with pre-analyzed “smart variables,” or marketing signals.
Marketing signals help marketers leap into the world of conducting highly personalized conversations in the moment. They allow you to sense customer activity and, with the help of signals, select the next best thing to “say” to the customer.
Smart Dummy Variables
An analogy for explaining the creation of marketing signals is the dummy variable. Their value is well established among sales forecasters. Forecast accuracy for a string of monthly sales numbers can be increased by simply adding an on/off promotion variable. A forecast based on sales plus a promotion dummy variable typically produces far more accurate sales forecasts than a simple forecast based on sales only. In other words, the promotion dummy variable is also a marketing signal.
Promotions are highly complex. Was the promotion a simple price discount or a buy-one-get-one free offer? If it was a price discount, was it described as cents off or percent off? How deep was the discount? Which products were covered by the discount? What steps were required to qualify for the discount? Where did the promotion run? How long did the promotion run? And so on. All of these questions have precise answers and are relevant to an analysis of promotional effectiveness.
CRM 3.0 platforms help the marketer synthesize vast sets of complex raw data into a smaller number of signals. The first step in the synthesis is moving from raw data to descriptive signals. One descriptive signal could be all of the features that together describe a particular promotion. Or, it could be that one descriptive promotion signal covers multiple sets of raw data relating to the final price paid by the customer. Then other sets of raw data could cover product, geography, duration, ease of redemption, and so on.
CRM 3.0 platforms can work with company-defined descriptive signals or they can analytically determine the optimal combinations of raw data into descriptive signals. Once the descriptive signals are set, the hunt for insight is now more efficient and effective because the analysis-to-response cycle skips raw data and starts with descriptive signals.
The second type of signal is predictive. They can be thought of as an aggregation of various descriptive signals. A predictive signal can be a simple combination of certain descriptive signals, or it can be the result of a more complex statistical process, such as weightings, regressions or other statistical treatments of descriptive signals. The table below illustrates descriptive and predictive signals for a price promotion.
Signals for a Price Feature
Next Best Conversation
Certain combinations or levels of predictive signals allow marketers to trigger new conversations for individual customers that have a high probability of success in the moment. The next best conversation for the above set of descriptive and predictive signals could be to immediately show a price discount for a moderately priced product in the product line currently being examined by the customer. CRM 3.0 platforms can automatically trigger next best conversations like this when the necessary signals converge.
Today’s most advanced CRM 3.0 machine learning and artificial intelligence systems do all of the above automatically, simultaneously and continuously. They seek and integrate raw data, find new descriptive and predictive signals and update current ones, and determine which next best conversations have the highest probability of success in the moment for individual customers.