Maximize your sales growth by improving Buyer Propensity

The key is to know what and when your customers want even before they do.

Retailers are always looking to stay ahead of their customers by analyzing their buying behavior and send targeted promotions that are personalized, relevant and at the right time. This is greatly helping retailers achieve higher response rates to promotional offers and as a result of which drive profitability. At the crux of this is advanced analytics that is propelled by Business Intelligence and Big Data Technologies. Predictive Analytics empower retailers to predict When, Where, Why and How customers buy or do not buy a product.

There are 4 key attributes that drive a customer to buy – Intention, Importance, Decision, and Preference (typically brand). Predictive Analytics enable retailers to study and understand these 4 attributes in combination with other relevant attributes of customer behavior by treating each and every customer as an individual and glean insights about their preferences and buying behaviors. Though there are several analytic approaches that enable retailers to stay ahead in their game we will look at how Propensity Models help retailers to increase their sales and revenue.

Propensity Models

Propensity models are enablers for retailers to study their customers buying behavior patterns and predict how they are likely to behave. These specific insights allow retailers to segment customers at a granular level which could be used for sending relevant promotional offers.

Imagine a large retailer having a customer base that runs into millions with each customer having many preferences and attributes. We are now talking of huge volumes of data and throwing in Clickstream data to the mix is a perfect use case for using Big Data technologies as well as creating segments of customers based on their intent, need, brand affinity etc. The ability of propensity analytics is to handle such large volumes of data (both internal and external) and still deliver segments with a higher degree of accuracy making it a highly popular and one of the most widely used analytic approaches in the retail industry.

Typically propensity models identify the relationships that exist between several customers attributes in addition to the variables that have an impact on the overall outcome. As a result, the ability to predict a behavioral outcome by changing the variable values is incorporated into a predictive model. A simple example to understand this could be building a model that would be able to predict a customer’s propensity to buy or not buy or to move from a premium plan to a lesser plan.

By being able to predict behaviors for each individual customer propensity models offer retailers to treat each customer in a unique way. This opens the doors for building a powerful loyalty program, immense opportunities to cross-sell, up-sell etc. and at the same time gives a clearer profile of a customer to the retailer.

A typical example could be that if the electronics retailer that I regularly purchase from has the ability to accurately predict (probably using click-stream data) that I intend to buy a big screen TV, and that I always prefer Sony over other brands but would be happy to pay much lesser to enjoy the same quality offered by Sony will enable the retailer to engage and interact with me by sending relevant offers and ultimately make me buy from them. This will also enable the retailer to increase the ROI from similar campaigns. Leveraging click stream data to build propensity models empowers a retailer to pick the right moment to send out a rebate which will significantly increase the campaign ROI.

Uplift Models

A natural question that comes to our mind would be that why would retailers have to go to the extent of sending relevant promotional offers if they know that the customer is likely to buy that product? Sending promotional offers have a cost attached to it and why not save that? This is where Uplift models come into play as they determine whether such an investment by the retailer is likely to pay dividends.

Propensity models when combined with uplift models allows a retailer to predict what or which of their actions has had a direct impact on the customer’s behavior which triggers the “buy” action. For example, uplift models can tell retailers whether a 15% rebate on a particular product would increase a customer’s propensity to buy the same with a span of 2-3 days. This would allow the retailer to send the 15% rebate coupon only to those customers whose buying behavior is likely to change as a result of having received that promotional offer.

Conclusion

There is a chance that the predictions themselves could become overwhelming unless the predicted results become actionable and ultimately make them realize value. It is important to understand what analytics would work for them based on business needs.

One important aspect that all retailers must think about and be aware of is that when massive data is used for analytics; more decision points will be created. This is good as it would enable them to differentiate and segment customers. Insights from propensity model combined with the uplift model can help retailers in strategizing promotions and discounts which would in turn increase the sales and revenue.