Indo Cabs Report - Final Analysis Essay PDF

Preview

Summary

Final Analysis Essay...

Description

Analysis of Cancellations at a Cab Portal Company October 2018

Executive Summary Part III, Q1 Based on the analysis, IndoCabs does not experience a strong pattern or correlation between the percentage of cancellations of trips and weekday. Three days of the week (Monday, Thursday, and Sunday) experience more cancellations than the others, but there is no strong evidence pointing towards a correlation. Holidays will likely continue to provide peaks in cab demand, and for that IndoCabs must prepare for the demand. It is likely that cancellations on weekdays occur through chance, and a trend in customer behavior cannot be determined from the data. The correlation between the number of bookings and the number of cancellations appear to show a moderate to strong positive correlation. Furthermore, the data points are somewhat close to the regression line, and the r-squared value, though not strong, is still away from zero indicating that the model explains some of the variability of the data around its mean. As such, IndoCabs should implement “surge pricing” for when their bookings increase (e.g. holidays) in order to combat the cancellations and make profits. In doing so, IndoCabs will not have to cancel on the customer and they will target customers who truly need the booking. Finally, IndoCabs experiences cancellations more frequently in the evening hours (4PM - 7PM) than in the morning. One suggestion is that IndoCabs should look into customer behavior in the evenings; perhaps, for example, customers are more likely to cancel their trips in the evening because it is around the time they get off of work and have time to themselves. Given that IndoCabs operates in India, they need to look into Indian culture to determine whether or not customer behavior changes during the evening. On average, trips last 4.29 hours for customers. This number alone is not enough to predict travel time for cab drivers because the mean is affected by outliers. By looking at the median, with a value of 1.32 hours for trip duration, IndoCabs can estimate that the trip duration will be somewhere between those hours. However, the reliability of the data presents an obstacle because the standard deviation was somewhat high. Again, this can be attributed to outliers such as poor road conditions or traffic thereby affecting the values. A quick glance at travel type reveals that the shorter the trip duration is, the more likely cancellations are to occur. For this, it is suggested that IndoCabs hire more cab drivers for “point to point” travel to meet the demand. Interestingly, bookings made via a mobile site have a longer trip duration for non-cancelled bookings as opposed to cancelled bookings. Yet, the opposite holds true for the online booking channel. Mobile bookings experience more cancellations at a rate of 19.40%, so this pattern supports the theory that the shorter the trip duration is, the more cancellations occur. Therefore, IndoCabs should eliminate the mobile site booking channel and focus their efforts on their successful channels like online booking to reduce costs.

1

Average of Trip Duration via Different Bookings Channels for Non-Cancelled and Cancelled Bookings Non-Cancelled Bookings Trip Duration (hours)

Cancelled Bookings Trip Duration (hours)

Mobile Site Booking

4.43

1.79

Online Booking

4.00

4.80

Introduction In this report, I will analyze the issues IndoCabs, a cab portal company based in India, wishes to address. IndoCabs has temporarily suspended their operations due to issues in managing their cancellations, bookings, and other variables. I look at a variety of factors that provide an understanding of their issues: measures of central tendency, trip duration, travel types, booking channel, booking window, and several graphs. The purpose of this proposal is to provide insight and suggestions to IndoCabs about their concerns through a thorough analysis of data and graphs. I propose that they implement “surge pricing”, hire more cab drivers to meet demand, look into the culture of their customer base, and remove the mobile site booking to raise profits and improve their reputation. I end the report with my recommendations and conclusion, important charts to look at, and my data preparation process.

Analysis A Look at Trip Durations Part III, Q2 Given that IndoCabs is looking for information about the typical length of trip durations and the variation associated with it, it would be of benefit to them to utilize the median, mean (average), and standard deviation. The median indicates the middlemost number and divides the distribution exactly in half. The benefit to IndoCabs in using the median is that it is immune to outliers; therefore the data for trip duration and booking window will not be impacted by observations that are significantly larger or smaller than the others. One of the weaknesses of IndoCabs was that their current algorithm was allowing customers to request a ride even when cancellation was highly likely to occur because of a lack of drivers. To alleviate this issue, IndoCabs can use the median to evaluate how many drivers they should hire given how long a trip will take for each individual. As such, they will be better prepared with the demand and will not need to cancel on customers, improving their reputation. Their booking windows, too, will be positively affected because IndoCabs can gauge how many days it takes for customers to create a booking and plan accordingly. Likewise, it is important to include the mean of the data for trip duration and booking window because it summarizes a large amount of data into a single value. The mean will give IndoCabs some insight as to how long a trip lasts in hours on average. In doing so, IndoCabs can re-evaluate their current algorithm to better fit customer’s needs and avoid cancellations. In addition, because the median and the mean of the booking window are not too far apart, IndoCabs can create an

2

expectation of when their customers will book a ride. If IndoCabs wants to measure how varied their data is, then the standard deviation is an excellent source of information. Standard deviation is less susceptible to outliers as compared to range, and it measures how spread out the data is from the mean. With this in mind, IndoCabs can determine if their data is reliable or not. By calculating the CV (coefficient of variation), IndoCabs can conclude that they have a high variation, therefore decreasing the reliability of the data and giving them the opportunity to fix these mistakes in the future. Summary Statistics of Trip Duration and Booking Window Trip Duration (Hours) Booking Window (Days) Median 1.32 0.41 Mean (Average) 4.29 1.88 Standard Deviation 12.46 4.74 The Magnitude of the Cancellation Problem at IndoCabs Part III, Q3 Despite claiming to have issues with cancellation, the data indicates that IndoCabs has a relatively low percentage of all bookings cancelled. To put the numbers into perspective, IndoCabs experienced 198 cancellations out of 2313 bookings. This is only 8.56% of all bookings cancelled, and although the numbers can be improved upon, it certainly is not terrible. That being said, IndoCabs needs to change their procedure for cancellations because it is affecting their reputation and causing inconvenience to their customers. If IndoCabs plans to create loyalty within their customers, then their percentage of all bookings cancelled needs to be reduced. Part III, Q4 In terms of travel type, IndoCabs encounters the most cancellations within Travel Type 2 or “point to point” travel. Most likely, this is because this type of travel requires the least amount of time and has a high demand for which IndoCabs cannot keep up with. On average, the trip duration in hours for Travel Type 2 was 1.47. Meanwhile, Travel Type 1 (“long distance”) and Travel Type 3 (“hourly rental”) took 46.68 and 6.13 hours respectively. The stark difference in hours likely contributes to the number of cancellations as there appears to be a trend: the less hours a trip takes, the more cancellations there are. These results are not too surprising because Travel Type 2 has the most bookings and fewest hours, followed by Travel Type 3 and Travel Type 1. By looking at the percent cancelled, the trend is supported because Travel Type 2 has the largest percentage, followed by Travel Type 3 and Travel Type 1.

3

Number of Bookings Number of Cancellations Percent (%) Cancelled Travel Type 1: 96.00 2.00 2.08 Long Distance Travel Type 2: 1827.00 175.00 9.58 Point to Point Travel Type 3: 390.00 21.00 5.38 Hourly Rental Part III, Q5 The cancellation rates for online booking as opposed to mobile booking are considerably lower, suggesting that mobile bookings are more likely to be cancelled. However, there is also significantly fewer bookings made through the mobile channel, with only a total of 134 bookings. In comparison, there are 892 bookings made online, allowing space for a few more cancellations. The difference between these values may be why the percentage for mobile bookings cancellations is higher at 19.40%. Online bookings fare better, with only 12.67% of online bookings cancelled. This implies that IndoCabs should continue with their online booking channel because it is more successful and less prone to cancellations; the mobile booking channel should be shut down in order for IndoCabs to cut costs and have a better focus.

Online Booking Mobile Booking

Number of Bookings Number of Cancellations Percent (%) Cancelled 892.00 113.00 12.67% 134.00 26.00 19.40%

Part III, Q6 The pattern in the bar chart insinuates that there is no correlation between the percentage of cancelled trips and the weekday. Although Sunday has the highest percentage of cancellations, this may just be due to chance and have nothing to do with customer behavior. Monday and Thursday have similar values as Sunday, so there is no evidence to suggest that a certain weekday affects the percentage of cancellations.

Percentage of Cancelled Trips by Weekday Percentage of Cancelled Trips (%)

12% 10% 8% 6% 4% 2% 0% Monday

Tuesday

Wednesday

Thursday

Friday

Saturday

Sunday

First bar indicates percentage of cancelled trips by weekday.

4

The Relationship between Booking Windows, Cancellations, and Trip Timing Part III, Q7 An r-squared value is a measure of how close the data is to the fitted regression line. When evaluating the r-squared value for IndoCabs, it can be determined that the model explains some of the variability of the data around its mean. If the r-squared value were closer to 0%, this would indicate that the model explains none of the variability of the data around its mean, and the opposite goes for a value of 100%. The points are not especially close to the regression line, but they do display a moderate to strong positive correlation between the number of bookings and number of cancellations. This can be determined by looking at how close the points are to one another, as well as the direction the points are going towards. As such, it can be implied that IndoCabs experiences some correlation between the number of bookings and the number of cancellations.

Correlation Between Number of Bookings and Number of Cancellations Hourly Number of Bookings

200 180

R² = 0.3542

160 140 120 100 80 60 40 20 0 0

5

10

15

20

25

30

35

N=2313. Hourly Number of Cancellations

Part III, Q8 The most informative line graphs are the number of bookings by hour and the percentage of cancellations by hour. The number of bookings by hour provides IndoCabs with a sense of when customers are booking their cabs. IndoCabs can prepare their employees for a small rush during these times (hour 6 to hour 9) so that less cancellations or issues arise. Similarly, the percentage of cancellations gives IndoCabs a broad sense of when customers will cancel their bookings. Hours 16 through 18 reveal a spike in the percentage of cancellations, offering IndoCabs an opportunity to look into what sparks the cancellations between those hours. Perhaps, for example, customers are more likely to remember in the evenings that they do not want to book a cab. The pattern between the number of cancellations by hour and the percentage of cancellations by hour are very similar, but not exact. Both sets of data have similar spikes, but they differ in how long it takes for the spike to decrease. The number of bookings by hour graph does not follow the same pattern, but still grants valuable information to IndoCabs about their peak booking time.

5

Number of Bookings

Number of Bookings by Hour 200 180 160 140 120 100 80 60 40 20 0 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23

Hour of the Day with 0 indicating 12AM and 23 indicating 11PM.

Number of Cancellations by Hour Number of Cancellations

35 30 25 20 15 10 5 0 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23

Hour of the Day with 0 indicating 12AM and 23 indicating 11PM

6

Percentage of Cancellations by Hour Percenatge of Cancellations

25% 20% 15% 10% 5% 0% 0

1

2

3

4

5

6

7

8

9

10 11 12 13 14 15 16 17 18 19 20 21 22 23

Hour of the Day with 0 indicating 12AM and 23 indicating 11PM

Part III, Q9 When the distribution of booking window is split into 0.25-day intervals, the data is more focused and specific. The issue with the 1-day bin size/intervals is that it gives IndoCabs false hope that their whole day is going to provide consistent bookings, when in fact the percentage of bookings decreases as the day goes on. Although both graphs exhibit similar trends in that the percentage of bookings fall as more days pass, the 0.25 bin size/interval gives IndoCabs more information to go off of because it specifies what interval of the day is experiencing the most bookings. This can lead IndoCabs to implement different prices depending on when customers book their cabs, and to even charge a fee if the booking window exceeds a certain amount of days.

Distribution of Booking Window in Days (1-day intervals) 90% 80%

% of Bookings

70% 60% 50% 40% 30% 20% 10% 0% 1

2

3

4

5

6

7

More

N= 2313. First bar indicates percentage of booking windows less than or equal to one day.

7

Distribution of Booking Window in Days (0.25-day intervals) 40% 35%

% of Bookings

30% 25% 20% 15% 10% 5% More

7

6.5

6.75

6.25

6

5.5

5.75

5.25

4.5

4.75 5

4

4.25

3.5 3.75

3 3.25

2.75

2.5

2.25

1.75 2

1.5

1

1.25

0.75

0.5

0.25

0%

N= 2313. First bar indicates percentage of booking windows less than or equal to 0.25 day.

Part III, Q10 It appears as though booking window does not impact cancellations to a significant degree. While there is some variation between the non-cancellation and the cancellation graphs, it is not substantial enough to say that there is a correlation or impact. For example, the non-cancellation and cancellation graphs for the 1-day bins hardly show a change besides on days 4 and 6. As before, the 0.25-day bins is more useful in communicating these insights because it does not generalize the data; rather, it provides a focus for each 0.25-day interval. The 1-day bins are useful when a generalization or summary of the data is desired, but IndoCabs should not take risks given their current situation. Consequently, IndoCabs needs to look at specific data in order to grasp the impact of their variables. In each scenario, the frequency of booking windows that are more than 7 days is larger than the data from day 2 to 7, indicating that customers create their booking early but do not start their trip until many days later. Information like this can be useful to IndoCabs when they want to evaluate how much to charge, when they will be busy, and more.

8

Distribution of Booking Window for NonCancellations in Days (1-day intervals) 90% 80%

% of Bookings

70% 60% 50% 40% 30% 20% 10% 0% 1

2

3

4

5

6

7

More

N= 2115. First bar indicates percentage of booking windows less than or equal to one day.

Distribution of Booking Window for Cancellations in Days (1-day intervals) 90% 80%

% of Bookings

70% 60% 50% 40% 30% 20% 10% 0% 1

2

3

4

5

6

7

More

N= 198. First bar indicates percentage of booking windows less than or equal to one day.

9

Distribution of Booking Window for NonCancellations in Days (0.25-day intervals) 40% 35%

% of Bookings

30% 25% 20% 15% 10% 5% More

7

6.5

6.75

6.25

6

5.5

5.75

5.25

4.5

4.75 5

4

4.25

3.5 3.75

3 3.25

2.75

2.5

2.25

1.75 2

1.5

1

1.25

0.75

0.5

0.25

0%

N= 2115. First bar indicates percentage of booking windows less than or equal to 0.25 day.

Distribution of Booking Window for Cancellations in Days (0.25-day intervals) 60%

40% 30% 20% 10%

More

7

6.75

6.5

6.25

6

5.5

5.75

5.25

4.5

4.75 5

4

4.25

3.5 3.75

3 3.25

2.75

2.5

2.25

1.75 2

1.5

1.25

1

0.75

0.5

0% 0.25

% of Bookings

50%

N= 198. First bar indicates percentage of booking windows less than or equal to 0.25 day.

10

Recommendations and Conclusion In conclusion, the analysis presents multiple recommendations. First, IndoCabs should prepare for the demand of holidays and bookings by hiring more cab drivers or implementing “surge pricing” to promote profits and reduce cancellations. The benefits from this are that IndoCabs will target customers who need their service the most and their reputation will improve because they will not need to cancel on the customer. Secondly, IndoCabs should look into the culture of their customer base, specifically Indian culture. In doing so, they can determine why customers decide to cancel their bookings in the evenings. Thirdly, IndoCabs should hire more “point to point” or Travel Type 2 drivers because this is where the bulk of cancellations occur. At last, IndoCabs should follow through with dropping their mobile site booking channel because it is creating unnecessary costs and scheduling issues, ...