Avoiding Late Arrivals on Delta

Overview

Using every flight that left New York City in 2013 (nycflights13), I answered two traveler questions: which NYC origin airport gives the best chance of arriving on time when flying Delta, and which destination airport is the worst for arrival delays.

Show the code

library(nycflights13)
library(tidyverse)
library(scales)

Which origin airport is best to minimize my chances of a late arrival when I am using Delta Airlines?

Show the code

delta_flights <- flights %>%
  filter(carrier == "DL") %>%
  filter(!is.na(arr_delay)) %>%
  mutate(late_15 = arr_delay > 15)

Show the code

origin_summary <- delta_flights %>%
  group_by(origin) %>%
  summarize(total_flights = n(),
    avg_arrival_delay = mean(arr_delay),
    median_arrival_delay = median(arr_delay),
    percent_late_15 = mean(late_15),
    .groups = "drop") %>%
  arrange(percent_late_15)

origin_summary

# A tibble: 3 × 5
  origin total_flights avg_arrival_delay median_arrival_delay percent_late_15
  <chr>          <int>             <dbl>                <dbl>           <dbl>
1 JFK            20559             -2.38                  -11           0.166
2 LGA            22804              3.93                   -7           0.193
3 EWR             4295              8.78                   -4           0.205

Show the code

ggplot(origin_summary, aes(x = reorder(origin, percent_late_15), y = percent_late_15)) +
  geom_col() +
  scale_y_continuous(labels = percent_format()) +
  labs(title = "Percent of Delta Flights Arriving More Than 15 Minutes Late",
    y = "Percent")

Show the code

ggplot(delta_flights, aes(x = origin, y = arr_delay)) +
  geom_jitter(alpha = 0.15, width = 0.2) +
  geom_boxplot(alpha = 0.3, outlier.shape = NA) +
  coord_cartesian(ylim = c(-75, 200)) +
  labs(title = "Individual Delta Flight Arrival Delays by Origin Airport",
    y = "Delay in Minutes")

Show the code

best_origin <- origin_summary %>%
  slice_min(percent_late_15, n = 1)

best_origin

# A tibble: 1 × 5
  origin total_flights avg_arrival_delay median_arrival_delay percent_late_15
  <chr>          <int>             <dbl>                <dbl>           <dbl>
1 JFK            20559             -2.38                  -11           0.166

The JFK airport has the best chance to minimize a late arrival while using Delta. On average Delta overestimates their arrival delay from JFK to anywhere so they let their passengers feel like they are getting in early. This doesn’t necessarily mean that they arrive faster but that they give a generous gap for their pilots to work with in getting from point A to point B.

Which destination airport is the worst airport for arrival delays? You decide on the metric for “worst.”

Show the code

destination_summary <- delta_flights %>%
  group_by(dest) %>%
  summarize(total_flights = n(),
    avg_arrival_delay = mean(arr_delay),
    median_arrival_delay = median(arr_delay),
    percent_late_15 = mean(late_15),
    .groups = "drop") %>%
  filter(total_flights >= 30) %>%
  arrange(desc(percent_late_15))

destination_summary

# A tibble: 28 × 5
   dest  total_flights avg_arrival_delay median_arrival_delay percent_late_15
   <chr>         <int>             <dbl>                <dbl>           <dbl>
 1 SAT             302              8.08                  -12           0.255
 2 DEN            1036              6.01                   -5           0.248
 3 PDX             457              4.25                   -7           0.232
 4 PIT             247              5.91                  -12           0.227
 5 MIA            2902              2.27                   -9           0.215
 6 FLL            2876              2.32                   -7           0.212
 7 PWM             233              1.14                  -13           0.210
 8 ATL           10452              7.42                   -4           0.202
 9 PBI            1450              2.06                   -7           0.190
10 AUS             352              1.41                  -12           0.190
# ℹ 18 more rows

Show the code

destination_summary %>%
  slice_max(percent_late_15, n = 10) %>%
  ggplot(aes(x = reorder(dest, percent_late_15), y = percent_late_15)) +
  geom_col() +
  coord_flip() +
  scale_y_continuous(labels = percent_format()) +
  labs(title = "Worst Delta Destination Airports for Arrival Delays",
    y = "Percent")

Show the code

worst_destination <- destination_summary %>%
  slice_max(percent_late_15, n = 1)

worst_destination

# A tibble: 1 × 5
  dest  total_flights avg_arrival_delay median_arrival_delay percent_late_15
  <chr>         <int>             <dbl>                <dbl>           <dbl>
1 SAT             302              8.08                  -12           0.255

The worst airport for arrival delays would be SAT. This is because the average arrival delay is the highest out of all of the other airports, the median delay is one of the top 3 and the percent of flights that get delayed is the highest out of all of them.