Create a PowerPoint presentation of 10 to 15 slides

The data which is been collected is been provided by an organization that is been doing this for a while. Data reliability means that the data is complete & accurate and it meets the business needs. One of the main objective of data reliability is to ensure data integrity which includes data security, data quality & compliance with the regulated laws. The source from which the data is collected is a regular provider of such systems data. To ensure the data is accurate validation rules has been setup on the data. In order to ensure the data observes regulated laws we checked for diversity and PII information, this is in relation to the associate information. As we expected we also observed seasonality trends during Thanks Giving & Black Friday. We validated the data to confirm that the data is free of redundancy. (Devaraj, 2007).

Data problems

Following are the data problems that we solved

Customer Information:

1. No company shares customer information. Even though we don’t need customer information we require the count of customers in order to capture the total deliveries. For this we looked at the various news articles to gather a rough estimate about the customers.

Date &Time formats:

The data that has been shared didn’t have standardized data & time formats. Also, time zones aren’t clearly called out for some of the records. We corrected the delivery times based upon the fulfillment or sort center time zones.

Product categorization:

We observed some of the products were misaligned to the product categories that they should have been to. We did external research to correct this misalignment. In some cases, product specifications were not clear, this was required to know the package type. We corrected also by doing external research.

Data Relationships

The relationship I find in this data is that there is an impact on the number of customers in achieving a higher profit for the company. Obtaining or maintaining a higher customer could lead to higher profit. This is said to big advantage to let the Nike company decide on how they have to maintain their higher customers in the future.

The data relationships that we are trying to determine involves various entities like customers, orders, product, shipping details. As per the data one customer can place multiple orders and an order shipment can be done by multiple delivery mechanisms. One order can involve multiple order items and each order item can be delivered separately or based upon customer preference all these order items can be combined into one package delivery. So, from data relationship perspective we were able to determine the cardinalities between the entities mentioned above. Customers to order is a one-to-many cardinalities, orders to product is a many to many cardinality, orders to shipment also many to many cardinality. These many to many cardinalities are properly addressed in the database structures.

For example, a customer order can contain an iPad cover and a kid’s toy. These two items will be delivered as a separate package from the separate fulfillment centers. However, if the customer has preferred to deliver them in a single package, then these two items are consolidated to one fulfillment center and from there it will be delivered to the customer. Even the delivery mechanisms can also vary depending upon the number of packages in the order. (SageSeller, 2021).

Missing data

Missing data is one the common constraints in the data request process. In such scenarios we will rely on the external websites to get the information about the missing data. In our case the most common scenario was missing of product categorization and we approached this problem in different ways

1. We referred to the amazon page to get the appropriate product category. For the obsolete products on Amazon, we referred to the external sites and populated the product category. Nearly 8% of the records didn’t have the product categorization.

2. There were situations where the delivery dates were not available however, ordered date was available. In such situations we roughly subtracted two to three days on the delivery dates to arrive at estimated ordered data. Nearly 1-2% of data didn’t have delivery dates.

Summary of Data Samples

The performance dashboard represents the total number of deliveries that are happened along with deliveries that are been missed for previous two weeks data. The variables used from data samples which are been used for this analysis are total number of deliveries, total number of late deliveries, date range from previous 2 weeks to current date, performance for a given warehouse, late deliveries occurred by different employee groups. The above discussed variables from data sample are the observations drawn from the population and we considered a random sample to perform the analysis.

Warehouse Performance:

Daily Delivery Performance:

Delivery rate success %:

Trends

Trend analysis is to determine how the business is performing over a period of time in an intuitive manner. To perform the trend analysis using Tableau reporting tool we used warehouse

performance in comparison with the delivery performance. In few cases, we can observe degrade in the warehouse performance impacts delivery delays. Below is the output:

In the above graphs we noticed that the trend is positive for the total number of deliveries and has been increasing. Positive trend refers to delivering a package to the customer on or before the committed delivery date whereas negative trend refers to delivering a package after the committed delivery date. Hence, there is a relationship between the variables i.e., warehouse performance and delivery of the packages.

Simpson Paradox

In order to avoid the occurrence of Simpson’s paradox, we randomly assigned the treatment conditions to the variables. Also, as it is from collected data sample, we ensured to apply observations carefully on the variables to detect any interesting patterns in behavioral data by disaggregating the sample data to identify subgroups within data whose behavior deviates from rest of the population.

Descriptive Analysis

The Descriptive Statistics summary in SPSS modeler provides the overview of the deliveries measure values by day. The information provides an overview of the mean, median, mode and standard deviation that, together with the minimum and maximum values, can be used to identify outliers, validate the data and analyze the range within which the total number of deliveries vs late deliveries in a given day. As discussed, the metrics for each of measures values number of deliveries and number of late deliveries help us to summarize the overall data. The mean gives the average value of the data counts, Median the middle term, range provides the spread between the minimum and maximum values. This will help the researcher to take just the sample metric value like sample mean, sample standard deviation, Sample Variance to represent the overall population data for testing the null hypothesis.

Boxplots:

There is a relationship between the above variables using the descriptive analysis. and the above plots shows the performances of the warehouse deliveries from the defined time frame. Using the above plots, the company can look into how the total number of deliveries are increasing with the level of customer satisfaction. The fulfillment center can now easily reduce the operational inefficiencies by using above mentioned data thereby they can focus on meeting the customer promised delivery date. In this analysis, Amazon logistics department can make a proper decision in the future to improve their overall performances such as their warehouse, operational efficiency, associate performance.

The supply chain helps the company to reduce the transportation/operational cost and also delivery time to make the product available to end customers. The primary goal is to reduce expenses and increase proficiency through delivery providers, associates, riders etc.

Segmented Data

Data segmentation is the process of grouping the similar data based upon common set of parameters so that it can be used efficiently for specific purposes. In this particular case data segmentation can be performed at the region versus product categorization. In other words which fulfillment centers are dealing with what kind of product categories.

Steps Taken

Here are detailed steps in investigating the data quality issues:

Step 1: Data Scrubbing

Step 2: Validation logic of all required variables & corresponding data

Step 3: Running the descriptive analysis on the collected data

Step 4: Performing the comparing the descriptive statistics of all variables using the SPSS

Step 4: Proving the analysis of generated outcomes

Step 5: Analyzing descriptive analysis & graphical plots

Step 6: Performing and finding the anomalies if any like outliers.

Raw and Summary Data files