Part two: Data Visualizations and Interpretations
On-time Arrival performance
Firstly, we analyzed the data about airline on-time performance. A flight is counted as “on-time” if it operated less than 15 minutes after the scheduled time shown in the carriers’ Computerized Reservations Systems (CRS). All the data are based on gate arrival times. Punctuality is a very important indicator when we consider our partner airlines. We analyzed the annual on-time rate for carriers over the last three years and the three-year average on-time rate.
From the bar chart, we can tell that the higher the bar, the better the on-time rate of the carrier. We can see clearly from the picture that Endeavor Air has the most punctuality rate while Allegiant Air has the least punctuality rate. According to the bar chart, we think it is better to work with an on-time rate of over 81%. From the top six airlines, we decided to exclude Hawaiian Airlines since the plane route of Hawaiian Airlines is relatively few and is mainly for routes to Hawaii. The remaining five airlines are all under consideration. Let’s take an insight into these top 5 carriers.
From the box plot, we can clearly see that the box of Delta Airlines is comparatively short which suggests that the overall percent of on-time arrival is high. Endeavor Air and Republic Airways have comparatively tall box plots, so we can conclude that their punctuality is variable. PSA Airlines, Southwest Airlines, and PSA Airlines also have a comparatively short box plot. PSA Airlines and Southwest Airlines also have the comparatively short box plot but for PSA Airlines the median, upper and lower quartiles are lower than those of Southwest Airlines. Therefore, from the box plot, we can conclude that Delta Airlines is our best choice and Southwest Airlines is our second choice. Finally, we decided to do the outlier analysis for these two airlines to see which airlines perform the best.
In the outlier analysis, we compared Delta Airlines and Southwest Airlines’ average on-time performance with and without outliers. From the scatter plot we can see that the range for Delta Airlines with outliers is from 60-95% which is much smaller than that of Southwest Airlines. After removing outliers, the difference in range is narrower but still, the range for Delta Airlines is smaller than that of Southwest Airlines. From the recent two years, the range for Southwest Airlines tends to be more stable but the performance is far worse than that of Delta Airlines. Therefore, we can conclude that for the fight delay indicator, Delta Airlines is our first consideration. Then we need to use other indicators to make our final decision.
Airports
According to the monthly on-time arrival and departure percentage data by airports nationwide by the US department of transportation, we generate two visualizations by the scales of percentages on both arrivals and departures. The reason why it is essential to our agency is that we can easily find out the on-time ratio of some major cities which our customers are more willing to travel to. By doing so, we are able to create more travel routes without concerning the time consuming that our customers waste at the airports and the aircrafts to some extent. For departures, the darker the color is, it is more likely that planes would take off on time, while for arrivals, we applied a reverse way on the icon marks that the lighter the red color is, the airplanes lands on time, which is a more understandable way for the management to view the graphs. For example, if we would like to plan a route between east and west coasts, we would more likely to recommend our management to choose a origin with not too small population, but has a really good aviation management, like Providence, RI or Hartford, CT in new England area because they are light green in the first graph. On the other hand, to choose a destination, we can choose San Jose, CA or Sacramento, CA on the west coast. There should be more elements for our final decision on choosing the origins and destinations; however, in this case of the graphs, those are the one crucial element for us to create new traveling routes.
Mishandled Baggage
Passengers travel with baggage most of the time. Especially in the air travel industry, it’s hard to separate the user experience of onboard baggage delivery away from the total user experience. With that said, we decided to include mishandled baggage in our research.
As for the mishandled baggage, we are mainly looking at three things here in this section - number of bags enplaned, number of bags mishandled, and number of bags mishandled per 1000 passengers. These three sets of data are closely related to each other, however, bags mishandled per 1000 passengers is the most important indicator because it reflects the performance of each airline.
For the data in this section, our tableau demonstration file is recommended for users to use as it’s built for the purpose of providing our clients a better and interactive experience to access the data and understand our findings.
Bags explained
According to the data, the top five bags enplaned airlines amount to a total of seventeen airlines are Southwest Airline, Delta Airline, American Airline, United Airline, and SkyWest Airline. The top four airlines have taken over 60% of the entire market share left about one third of the market shared to the other fourteen airlines.
Bags mishandled
We found that American Airlines and Southwest airlines have always been in the top two positions in terms of mishandling baggage and followed by United Airlines and Delta Airlines. The rank of bags mishandled is very similar to the rank of bags enplaned. We figured that the number of bags mishandled is deeply related to the number of bags enplaned, the number of bags mishandled goes up and down along with the number of bags enplaned.
Bags mishandled per 1000 passengers
Bags mishandled per 1000 passengers is the most important indicator in this section because it directly compares each airline’s performance regardless of the noises such as airlines’ market share and size, etc. In this segment, larger numbers indicate poorer performance.
In 2018, there were four airlines that had a number greater than the average which were Envoy AIR, ExpressJet, SkyWest, and American Airline. In 2019, there were ENVOY AIR, American, MESA, Republic, PSA, United, and SkyWest that had numbers greater than the average. In 2020, the rank is the same as 2019 but with a closer number between the seven airlines.
If we look at the average of the recent three years of their rates, ENVOY AIR has the worst performance in this segment and Allegiant Air has the best performance. However, for the airlines that are performing better than the industry average, Southwest and Delta are standing out to us because they manage to constantly deliver good services considering the size of their business.
American Airline, MESA Airline, Republic Airways are the most frequent names we see in the table. From the perspective of the performance in mishandled baggage, we suggest avoiding these airlines listed above if baggage service weighs heavy in your business value.
Consumer incidents and Complaints
Incidents
Based on the definition by The International Civil Aviation Organization (ICAO), an incident is an occurrence associated with the operation of an aircraft, except for an accident. Data related to incidents are crucial, for it can help our customers avoid cooperation with airlines that are more likely to be involved in in-flight issues. The monthly incidents of airlines from July 2018 to June 2021 were graphed. The histogram above clearly shows that most airlines' monthly incidents are within 100 times. However, American Airlines and United Airlines are notable exceptions. American Airlines usually have more than 100 incidents per month, and United airlines even have more than 1,000 incidents in some months. In short, these two airlines have higher numbers and extreme value in incidents.
Complaints
A dissatisfied customer can cause a company to lose many potential customers. The airlines that are prone to receive complaints from passengers indicate that in the future, passengers who purchase these airlines’ tickets through a travel agency will also have a higher probability of complaining to the travel agency about their travel experiences and even asking for compensation. When dissatisfied with the service quality and the travel agency's reply, some consumers are likely to spread negative comments regarding it. Assuming that on average, each dissatisfied customer complains about the travel agency to four acquaintances around him or her, the company is likely to face the problem of customer reduction.
​
According to the tree-map of Air Travel Consumer Complaints, United Airlines is the enterprise that received the highest average number of complaints from passengers in the last 3 years, with an average of 418.5 complaints each month from 07/2018 to 06/2021, followed by American Airlines, Frontier Airlines, Delta Airlines, Spirit Airlines, Southwest Airlines, Jetblue Airways, and Alaska Airlines in descending order. In the tree-map, southwest airlines, JetBlue Airlines, Frontier Airlines, and Spirit Airlines are low-cost airlines. Their average complaints are more than other airlines. The outcome is understandable because low-cost airlines tend to charge passengers more extra fees, which might easily result in complaints from passengers. However, both United Airlines and American Airlines have more complaints than those low-cost airlines. From the perspective of the top three airlines in the world - American Airlines, Delta Airlines, and United Airlines, this chart shows that Delta Airlines performs better than the other two airlines in terms of consumer complaints.
We also use the monthly Consumer Complaints data from July 2018 to June 2021 to draw the violin plot. The plot with log function in the x-axis clearly shows that United Airlines has the widest range of customer complaints per month, and its number of customer complaints is almost never fewer than 75 per month. What's worth noticing is that though American airlines' number of consumer complaints is relatively concentrated, there are still almost 25% of the months with more than 500 complaints per month based on the interquartile range of the violin plot. The other airlines showing better performance in this category are Skywest Airlines, Allegiant Airlines, JetBlue Airways, Southwest Airlines, and Delta Airlines, for each of their median value is within 100. In other words, United Airlines and American Airlines have terrible performance in customer service.
By combining the average data of Incidents and the average data of Consumer Complaints that occurred on the day of Incidents over the three years, we use the scatter plots to display data associations. This approach is taken to identify which airlines are likely to have better service quality. Compared to Frontier Airlines, Delta airlines performed better because they have a small gap in incidents but a relatively large gap in complaints based on the scatter plots above. Delta airlines receiving 136 complaints and 47 incidents is significantly better than Frontier airlines, American Airlines, and United Airlines, which has around three times as many complaints and incidents as Delta airlines.
Regression Analysis
From the previous reports, we would like to pick both Delta Airlines and Southwest Airlines as potential partners. More specifically, Delta can offer more route options and availability on booking tickets, so we run a nonlinear multivariate analysis on the satisfaction score of Delta Airlines and try to find out what are the most important factors affecting customer satisfaction.
First of all, we run a regression analysis based on all regressors we can get from the database.
Reflection on Regression 1
As we pick a 95% confidence interval, Incidents, which has a 0.3963 p-value should be dropped since it would affect our final outcome. Also, as an explanatory, we use log function on mishandled baggage since it has a much larger scale rather than other regressors. Then the results and plot is shown below.
Reflection on Regression 2
This is not quite reliable since most of the points are on the left hand side and the trend does not make sense that more complaints generates lower satisfaction scores. As a result, we should have to drop the outliers. The regression models we generated are shown below.
Reflection on Regression 3
After we generated another data frame removing all of the outliers, we found that mishandled baggage no longer satisfies the requirement of p_value<0.05, so we need to drop this regressor. Finally, we have the regression model shown below.
Reflection on Regression 4
As we can observe from the above picture, four factors affect satisfaction score significantly, which are departure on time, arrival on time, complaints and incidents. All of the p values for the independent variables are less than 0.05 and the p value for the whole function is extremely low. Among all the regressors, departure on time and arrival on time are crucial for customer satisfaction. However, for incidents, it has a positive relationship with satisfaction score, which does not make sense. It may account for the inaccuracy of data or some other factors we have not been able to find, so we chose to exclude it as well. Furthermore, we come up with a graph showing all of the regressors and satisfaction scores of flight customers.
