Aplicando métricas en la inteligencia de negocios.
Tale is a fintech that provides credit cards to people. People may use these cards to pay for stuff, and they have to pay back to Tale when their payment date is due. Tale is interested in increasing the amount of money that people pay with Tale’s credit cards
Answer: First point: What is happening? We must first find out who is acquiring a card and section it by its various available characteristics, in order to find a common pattern with descriptive statistics. We can group the information by type of business that is being consumed, reorder the data and obtain a top (5,10,20, etc) of the type of service/product that people are requesting the most, which would tell us the specific reason why they got their card, the use of it. Now, based on the results, 2 aspects can be chosen: 1. Start a promotional campaign focused on acquiring new customers who focus their expenses on those who observed the greatest use of the credit card, since this being a most marked need in consumers, then we will know that they are going to use the card more frequently (for this it is also convenient to section by type of person, age, credit limit, etc., to better understand the profile of the card user and focus more specifically the promotion campaign, which directly reaches people with similar profiles). Although this seems to be the most viable solution, in reality, acquiring new clients can be extremely expensive in the medium term, so a more immediate solution (without neglecting the first one, since it can be beneficial in the future) would be: 2. Based on the products/types of products most purchased with the credit card, look for credit facilities to make the use of the card more attractive, that is, knowing that, for example, people usually use their card mainly for buy certain articles of clothing, offer reductions in the interest in purchases that belong to products of that category, which will make it more attractive and people will tend to buy more of those products (considering own customers). Now, it is worth asking how much you have to reduce the rate to obtain the maximum profit? For this we can generate a metric divided into 2 (or 3) phases: Lower the interest as a promotion 2 times, this to analyze to what extent purchases increase (if they do) depending on the interest of “x%”. With this, we can find the respective exchange rate and generate a simple linear maximization problem, in order to obtain the maximum benefit at the correct rate (that is, we could lower the rate too much and people use their card a lot but earn very little for each transaction, or lower the rate at the exact point where income is maximized due to the fact that there are a considerable number of transactions at a rate that is not too low). Note that the change in the rate can be applied to the 2 aspects mentioned above, either to acquire new clients or to increase the use of the clients we already have.
Answer: As mentioned in the previous point, to identify the category in which it is going to be decided to invest in a promotional campaign, it is necessary to know the top of the categories that have a greater number of sales, as well as the profile of the client who buys those products. products. We must also consider the number of customers we have with certain characteristics, since it is not necessarily convenient to apply a promotion to a product if it is bought by few people (although it would be implicit that if it is in the top 10, then if it is a product highly purchased by users), for this we must consider both the number of times the product is purchased, as well as its price, since we can have products that are purchased to a great extent but are quite cheap, which would not generate much performance. For example, with the data obtained for this exercise, we can observe the following top categories of best-selling products and the highest amount of income:
Now, there is another metric that we can consider: The rate of change of expenses per product with respect to the given interest rate, since the products that increase your purchase a lot due to the fact that the card issued a lower rate can have a much higher potential. of future revenue growth (may be high-value products/categories of products that people prefer not to buy due to high interest, but are of general interest), which can apply to purchases that are not as recurring, but that This would identify those with the greatest potential. As extra metrics, generate time series graphs, which can be filtered by product/category, and help us identify seasonality, trend, etc., in order to be able to generate a forecast of purchases that certain branches will have in the future (considering similar conditions) through some statistical model (either ARIMA, BATS, Neural Network, etc). With this we can adjust the budget for the campaigns in a better way.