Understanding the Factors Influencing the Timing of Product Purchases: Analysis of Key Variables

1. Introduction

One of the most important questions that retailers like 家乐福 沃尔玛 but also Scotiabank Banco Sabadell Banco Bci Santander Banco Ripley Chile among many others, ask themselves is not only whether or not a person will buy a certain basket of products (yes/no), or what will be the product they will buy after buying certain products (distance), but to know WHEN you will buy the product. This will mean that the definition of supply, positioning, stock, price, and promotion strategies, among other implications, can be defined with more information.

In this case we have worked with a particular dataset. This dataset has been provided by a real company and contains information necessary for conducting an analysis on the time until a purchase of a certain basket of products.

The dataset consists of 10 variables, each of which contributes to understanding the factors influencing the timing of a purchase. In this analysis, we will explore the components of each variable and examine their potential impact on the time to purchase.

For that, we will use advanced machine learning models and microeconomics principles to understand when a certain basket of products is purchased.

2. Variable Components?:

1. Age: This variable represents the age of the individuals. Age can be a significant factor in determining purchasing behavior, as consumer preferences and needs often vary across different age groups.
2. Gender: The gender variable indicates the gender of the individuals. Gender can play a role in shaping consumer preferences and purchase decisions, as it may influence product preferences, brand loyalty, and shopping behavior.
3. Income: The income variable reflects the income levels of the individuals. Income is an essential determinant of purchasing power and can influence the affordability and willingness to make a purchase.
4. Marital Status: This variable captures the marital status of the individuals. Marital status can affect purchasing decisions, as the needs, priorities, and spending patterns of individuals may differ based on their marital status.
5. Location: The location variable represents the geographic location of the individuals. Geographical factors such as culture, availability of products, and local market dynamics can influence purchasing behavior.
6. Purchase History: This variable indicates the individual’s past purchase history. Previous purchasing behavior can be an indicator of future purchase intentions, as individuals with a higher purchase history may exhibit greater brand loyalty or a higher propensity to make repeat purchases.
7. Online Behavior: The online behavior variable reflects the level of engagement of individuals in online activities related to shopping. Online behavior can influence the time to purchase, as individuals who are more active online may have access to a wider range of products, information, and promotions.
8. Interests: This variable captures the specific interests of the individuals. Individual interests can influence purchase decisions, as individuals are more likely to purchase products or services related to their specific areas of interest.
9. Promotions and Discounts: The promotions and discounts variable indicates whether individuals are exposed to promotional offers and discounts. Promotional activities can influence the time to purchase by creating incentives for individuals to make a purchase sooner or by influencing their decision-making process.
10. Customer Experience: This variable represents the individuals’ past experiences with customer service. Customer experience can impact the time to purchase, as positive experiences may enhance brand perception and increase the likelihood of repeat purchases.

By analyzing these variables and their components, we aim to gain insights into the factors that contribute to the timing of a purchase. Understanding these factors can assist businesses in developing targeted marketing strategies, optimizing promotional activities, and improving customer satisfaction to enhance the overall sales performance and profitability.

ggsurvplot(survfit(Surv(time,status) ~ OnlineBehavior + Gender, data = df)

ggsurvplot(survfit(Surv(time,status) ~ OnlineBehavior + Location, data = df)))        

bst <- xgboost(data = sparse_matrix, label = output_vector, max.depth = 4, eta = 1, nthread = 2, nrounds = 10

importanceRaw <- xgb.importance(feature_names = sparse_matrix@Dimnames[[2]], model = bst, data = sparse_matrix, label = output_vector)

xgb.plot.importance(importance_matrix = importanceRaw, main = "Variables Importance - SHAP", xlab = "Importance"))        

3. Economic and Statistical Analysis

In the figure above, the variables that have the most weight to define the moment of a purchase are:

1. Age: It has a positive coefficient of 0.022528, which indicates that the customer’s age has a significant influence at the time of purchase. A higher value for age suggests that the customer is more likely to make a purchase at a later time.
2. PurchaseHistory?: Has a negative coefficient of -0.01586, suggesting that more frequent purchase history is associated with earlier purchase time. That is, customers with a more active purchase history tend to make purchases at an earlier time.
3. OnlineBehavior?: Has a positive coefficient of 0.009331, indicating that customer online behavior such as searches, clicks, and purchases on the retail website have an influence on the time of purchase. A higher level of online activity may be related to an earlier purchase time.

These three variables present more significant weight coefficients in the model, which implies that they are important factors in determining the moment of a purchase. However, it is important to note that the relative importance of these variables may vary in different contexts and for different data sets.

4. What about PromotionDiscount??

The variable “ PromotionsDiscounts “ has a negative coefficient of -0.02363. This indicates that promotions and discounts have some influence on the time of purchase, but in this case, the negative coefficient suggests that the presence of promotions and discounts is associated with a later purchase time.

This could be interpreted in different ways. For example, promotions and discounts could be strategically designed to encourage purchases at specific times, such as periods of low demand, or to drive sales of slower-moving products. Therefore, customers may wait for these offers to appear before making their purchases.

It is important to note that the relative weight of the “ PromotionsDiscounts “ variable in the model may depend on other factors and the specific retail context?. In some cases, promotions and discounts may have a more significant impact at the time of purchase, while in other cases, other factors may have a stronger influence. It is necessary to evaluate the model in conjunction with other analyzes to obtain a more complete and accurate picture of the variables that affect the moment of purchase.

About these models and how to apply them to the design of data-based products, will be my participation next September 11 in NYC at the "Customer Analytics Summit". I leave the link here