A Detailed Analysis of Food Delivery Aggregator Data

Introduction

A detailed analysis of food delivery aggregator data offers critical insights into consumer preferences, delivery trends, and market dynamics. By examining data from leading platforms like UberEats, DoorDash, and Grubhub, businesses can uncover patterns in ordering behavior, peak delivery times, and popular cuisines. This analysis enables companies to optimize their operations, enhance customer satisfaction, and make informed decisions that drive growth. With the rapid expansion of the food delivery market, understanding these data-driven insights is essential for staying competitive and meeting the evolving demands of consumers in a fast-paced industry.

Recently, we shared how to extract data from food delivery websites and process and mold it into a visual map with python. Let's discuss further to resolve a few queries, as shown below.

Based on the input address shown in the map below, We filtered them into different categories based on delivery fees in different colors. As predicted, we found that the higher the delivery fees of the restaurant, the highest the delivery. As indicated in the map, you can see a circular boundary with orange shading. Beyond that boundary, the delivery fee is greater than 5 USD. But, there are a few points beyond the limit where the delivery fee is less than 5 USD.

To understand the factors influencing these outliers fees in a better way, we will separate points outside the orange boundary and compare their features to those points inside the boundary. To do so, we will choose the topics that create this orange boundary and use them to draw a polygon in the script. Then, we will check the facts in the datasets and their location to analyze further.

We observed the general trend in an outlined area with red shading, where delivery fees increase with the input address distance. But, the points outside the mapped polygon don't follow the pattern (black points are an exception), and we can consider them outliers. We need to convert coordinate points for every restaurant into a geometric point to compare features of QSRs inside & outside the polygon and find if these points are inside or outside.

In the dataset, our function determines whether the restaurant is inside the polygon, and the last column shows that. To study the characteristics of the restaurants in both categories, we will separate the dataset into two groups.

The most impactful factor on the delivery fees seems to be the distance between the restaurant and the delivery location. We will consider this feature to begin.

Most restaurants inside the polygon are situated 1 to 3 km from the input address, and their delivery fees range from 0 to 7 USD. Apart from this, restaurants in the 4 to 7 km range have a probability of delivery charges below 5 USD.

From the analyzed data, we observed that most of the restaurants located outside the polygon have zero delivery fees, which means the delivery distance needed to be more accurate. We saw the delivery fees of 7 USD for the restaurants within the polygon. This reflects that restaurants outside the created polygon have considerably fewer delivery fees than inside restaurants. Moreover, analyzing the potential impact on delivery fees, we selected a group of columns with our assumptions and the output of the correlation matrix measuring the strength of relationships of values.

df.corr()['deliveryFeeDefault'].sort_values(ascending=False)[2:]

Per the correlation matrix, delivery by Scoober is the main factor with a strong correlation with the delivery fees. There is a possibility that other factors could have a weaker correlation because of the existence of outliers. Therefore, we will further investigate by plotting the data.

After plotting the selected data column groups, we see all the restaurants in the polygon plotting. The columns' price ranges and new have the same distribution. The slightest order value also varied more in this portion considerably. But no restaurant outside the polygon used Scoober delivery, and the sponsored restaurant distribution looks more balanced.

Let us check the restaurant distribution that uses Scoober delivery, where we will plot two plots parallelly using the below function. This will help us to compare both groups easily.

The boolean value shows whether a restaurant is using Scoober delivery or not likely plays a significant role in delivery fees of restaurants, as you can see in the shared red range in our study earlier. The reason is that restaurants are making agreements with lieferando to use Scoober delivery services, where Scoober will work as a food delivery aggregator. To further investigate the relationship between delivery fees and Scoober, we can categorize the data for the restaurants associated with Scoober delivery and study the link between delivery fees and other factors.

Here, the distance is related to delivery fees on a positive note, where the output has grown from .15 to .75

It is worth noting that the order's base value has NaN, meaning the order value is constant or missions. Let's analyze these restaurants again for the case where lieferando doesn't use Scoober delivery for food shipment, considering that these restaurants are using their delivery platforms

Here, we couldn't see any improvement with the distance remaining constant at 0.20, and the minimum order value was 0.4, way higher than the distance.

When our team plotted the least order values for Huston-based restaurants to check whether they use Scoober for delivery service, we got the results below.

It looks like lienferando has a base value for orders of 8 USD for each restaurant that offers scoober delivery service. It may be because of the agreement between Scoober delivery and restaurants. It is the same as another food aggregator Wolt as they also have a base order value of 8 USD for every restaurant.

Our team created histograms to compare selected columns on whether restaurants used Scoober delivery. This helps to quickly overview the notable variations between the two groups we created for analysis.