In order to plan a marketing strategy, you need to know your customers. By knowing who to target and who to avoid, your are able to greatly reduce the advertisement cost and increase ROI. Regarding your advertisement, your customers are primarily divided into four categories as explained in the following figure.
Among your potential customers, some people have already decided. In marketing science, they are called Sure Thing if they plan to buy your product even if you don't advertise to them, or they are called Lost Cause if they are definitely against your product. The third group of people is called Sleeping Dog they don't like to be disturbed by an advertisement to the extent it has a negative effect on their decision.
Advertising strategy is all about the fourth group or Persuadables. We are interested in knowing which customers buy our products if they receive an offer.
There are various modelling practices that can predict who will be persuadable customers. We believe uplift modelling is the most promising one. In uplift modelling, we build a model which predicts the causal influence of the action by comparing two datasets: The first dataset, called treatment dataset, applies when an action is taken. The second dataset, called control dataset, applies when no action is taken.
Currently, the most popular approach for predicting response rate is logistic regression. We believe it is theoretically flawed because we compare two models from two different universes: one in the treatment group and another in the control group. This also means we are measuring the modelling errors twice . In our recent work, we used other nonparametric approaches that actually try to directly predict the uplift. Particularly, we used a decision tree based approach which resulted in a very accurate learning algorithm.
In this approach, we can predict the difference that a marketer's actions will make on the behaviour of customers, comparing them to prediction of customer actions that would have happened without the marketing advertising. In this regard, marketers can determine if advertising for a particular customer is worth it or not.
These approaches provide clear customer segmentations and make sure that organizations are targeting the right audiences. It both reduces campaign costs and increases marketing campaign ROI.
It is crucially important for marketers to know if they are targeting the right audience in their advertisement. This knowledge can be extremely effective in reducing their costs and producing revenue.
By using intelligent analytics, and particularly uplift modelling, companies make sure that they invest in the customers who respond positively to an offer and avoid those customers who have already made their decisions.
Our uplift model tells us how much more likely your customers are to purchase if they have received an offer as opposed to not receiving an offer. Moreover, it tells us what are the factors and segmentations that help us to determine our marketing strategy. In the future, we will talk more about this approach and the key details that you should pay attention to when building your models.