Introduction
Since I was trying to pay back all the money to my relatives and my soon-to-be-ex-wife, I came up with the crazy plan with the aggressive strategy on how to hustler in every single gigs job. I got fired up after reading too many Navy Seal Agent stories, so I disciplined myself with time. I hypothesized that every single minute I must receive an order non-stop. I began the experiment between time versus money. I diversified my time portfolio base on the approach as time was money.
I started researching on my options about the earning money app. I have Doordash, Grubhub, Instacart, Postmates, Uber, and Lyft. I went through every single feature for user experience between each app. I am not going into details about each app because it is still under my research about the LTE, 4G, HTML, Java, and microprocessor. I came up with earning tool as Doordash and Uber, and I chose Grubhub as the back up for late food order. At the same time, I started collecting data about earnings, orders, and time effort. After a while, I sat back, and I began my analysis on Multiple Linear Regression on excel. I used this project as the first pilot product on my portfolio on the way to become an analysis in Data Analytic Industry.
Overview of study
The analysis indicates the correlation between the tips money with 10 cities I delivered food for one month through the Doordash app. I would like to see if there is an impact between the number of restaurants on the amount of tips money. I would like to choose the tips money as the dependent variable and the number of restaurants as the independent variable. The tips money will range from $1 to $10, and the number of restaurants will range from 100 to 1500. For this time, I dig deep into my analysis for multiple linear regression from my Gigs Delivery Strategy toward the earnings. I am going to choose total earnings as the dependent variable as total orders per day, duration and number of restaurants per city as the independent variables. My observation was going for 30 days. The total earning varies from $0 to $200. The number of restaurants will range from 100 to 1500. Total orders per days range from 0 to 30. After setting up condition for this, I collect all my data from Uber earning statement in January, Doordash earning statement in January, and the number of restaurants from Doordash website.
I was going through this analysis and I decided to step back after spending everyday hustling for money before the COVID-19 impact in March. I need to think from the beginning until the end to see if this is worth it to increase my income.
Analysis
I get the number of restaurants from the Doordash website, and I got the record of tips money from the earning section from the apps. Here is the dataset I collected along with the charts I made in Excel:
I ran the Multi Linear Regression on Excel:
For this multiple linear regression, I picked three independent variables as total orders per day, durations, and number of restaurants. I have an equation: total earnings = 5.00(total orders per day) + 8.01 (time durations) + 0.01 (number of restaurants) — 37.67. Therefore, the R square is 0.9476 and the coefficient correlation is 0.973. From the stochastic model, I have u representing for the uncertainties that might affect the total earning of gigs job.
As the number of restaurants is the independent variable and the tips money is the dependent variable, I come up with the linear equation as Tips money = 0.0037 (number of restaurants) + 1.3856. Therefore, I have the coefficient correlation = sqrt (0.5778) = 0.760132 and the coefficient of determination is 0.5778.
From the linear equation and the coefficient correlation, I see the “ideal fit” from multiple linear regression that the total orders per day and duration time correlate toward the total earnings. In addition, two variables total orders per day and duration time does not correlate much to each other because I did not receive many orders even though I worked 6 to 7 hours. It could lead to the bias on delivery strategy that I focused on delivering on the weekend.
However, if I kept that thinking, a lot of drivers would think the same. It might lead to the heat competition between gigs driver. That was exactly what the heat map on Doordash apps or Uber apps wanted the gigs to see when working on the weekend. The heat map could use the heat map to raise the base pay which cost the opportunities for the earnings toward the gigs workers. The misleading from the heat map was to benefit the restaurants and Apps business. For the gigs workers, it only benefits in the short term which everyone calls it “instant-money”. The instant money has high percentage of uncertainties toward the quality of service which make sense when tipstip money is optional in-service business.
Going forward to multiple linear regression, I would put total orders per day, duration times, and number of restaurants as 0, and I have the total earnings as -$37.67. It make no sense if I do not do any delivery orders. However, the intercept could contribute to the uncertainties when doing this business. Because I drive a lot, my car would need to do oil change faster instead of every two months. Especially, Uber has the car leasing program with weekly discount price, and they took advantages of this factor. However, this was the minor factor in this situation because the observation did not account for 30 days. Therefore, I already emphasize as the short-term job.
The main focus was to decide if this is worth it the value I was looking for toward this analysis. linear model between the number of restaurants and the tips money. We could see that the increasing of number of restaurants would lead to the increasing of tips money. However, it does not make sense from the linear equation that the tips money could be $1.39 when the number of restaurants is 0. This could mislead me into the bias decision to focus delivering in the high density of restaurants cities such as San Francisco, Berkeley, Oakland, or Sunnyvale.
When I set duration times and number of restaurants as 0, I have the equation as the total earnings = 5.00 (total orders) -37.67. When using the scatter plot for simple linear regression, I could see the total orders are linear toward the total earnings when looking at the scatter plot in figure 4. It is understandable that the more orders I receive would help make more money even though the value of order might be low.
The next part is to verify the quality of my stochastic model by going into four factors of linear regression:
I am going to dig deep into the linear regression for the better strategy in delivering:
In order not to make a bias decision and a smart strategy, I would need to consider the other factors in this analysis in order to create better linear regression in this stochastic world of data. I would tackle my analysis on four factors:
1. Linear in Parameter
2. Random Sampling
3. Sample Variation in the Explanatory Variable
4. Zero mean of the error term conditional on the independent variable
The stochastic model and the data would help me make better decision to either continue or not from this gigs business. Even though there are uncertainties u in my stochastic model, I was trying to minimize the impact of randomness for this gigs business model such as car maintenance, traffic, and especially the pandemic COVID-19. When the pandemic started impacting on traffic, I saw the extreme drop of orders from both food and humans, so I had to switch to do Instacart. However, the earnings were not enough for me to cover the car leasing, so I had to return the car in the middle of March.
1. Linear in Parameter:
Base on this analysis, I had to schedule my time extremely for every single in order to fit in this model. I had to secure the full-time job from 7:30 to 4:30pm during the weekday, and I created the range of 4 to 6 hours every day to do delivery job. At the same time, I see the trending of earning statement after two weeks along with the driving hours. From the earning statement, I comply with my schedule along with the earning goal. The number of orders and the hours were the controllable factors for me toward this model, and they were linear toward the goal I was setting. As I made the list of uncontrollable factors such as numbers of restaurants, lunch time frame, dinner time frame, and the business complex where there are restaurants, mall, bars, I included the dinner time frame and numbers of restaurants in this analysis. From there I would like to see how much the uncontrollable factors could impact my model.
Since the uncontrollable factors have low impact on the analysis, I would have to shift my priority toward the controllable factors more than them, but I will not ignore the uncontrollable factors. The world is complicated and spinning. If I ignore uncontrollable factors, I will have made bias decision toward operating business.
When looking at the scatter plot, I see more dots falling on the line, and I only see four dots deviated from the line. Those dots would require more data for the variation on the scatter plot, and those dots belongs to the San Francisco, Oakland, Milpitas, and Richmond. At first, I assumed San Francisco and Oakland would tips me a lot since there are a lot of restaurants in the area, but it turned out they tips just like the same in Milpitas though the number of restaurants is less 3 times comparing to other cities.
I was wondering what other factors would drive me more tips money when my assumption about delivering in the big cities like San Francisco and Oakland were wrong. It made me to think about the other factors such as the type of food, the distance of delivering, size of the order, and the time of delivering.
From there, I could see the limit of my linear regression, and I need to weight each single factor base on the histogram of frequency for each factor.
1. Random Sampling:
For this multiple linear regression analysis, the data was observed over 1 month as I work every without any day off. At some points, I could not drive back after finishing those hours, so I had to sleep in the car during the winter season. Because it was the winter season, I assume that a lot of people would orders a lot of food more ordering Uber rides. I had to be an open mind toward the weather since I was trying to study people’s behavior toward the weather and the traffic. On the nice day, everyone wanted to get out, so they did not want to order food. I might have to drive to other cities where there is the demand for food or Uber rides.
In this delivery gigs business, delivering food or humans have the same value in term of cost, distance, and delivery time. Whenever the nice weather came, I had to drive to other cities, so I could have driven to 10 different cities in the Bay Areas in order to meet the earning goal. My MLR analysis might work well in the winter season, but it might vary in the summer and fall season. I had to keep my mind open and data continuously.
Because I was thinking about the tips money and the number of restaurants in each city, I already realize that I miss other certainties that might correlating to the true value of delivering food. However, I need to consider weighting each factor in order to fit in the stochastic modeling because some factors I could not control, and I might be crazy try hard on something that I could not control. That is why the histogram of frequency for each factor would define which one to fit into the stochastic model. For example, the tips money might be the main incentives for service workers to work hard, but it does not mean it could be the main driving force to define the earning for the gigs worker. For gigs worker like Doordash, I should focus on receiving as many orders as possible in order to maximize the earning before the peak hours. It is the game of making choice, so I need to be careful when building the delivering strategy.
3. Sample Variation in the Explanatory Variable:
When I look at the summary output for the regression analysis, there are still standard errors toward the total orders and duration times even though they are more linear than the number of restaurants per city. The standard error for total orders is 0.88, and the standard error for duration times is 1.32. The uncertainty contributing certain percentage of my total earnings is the tip money which is optional in this service industry. They could be the reason causing the deviation of points on the scatter points between (total earnings vs total orders) and (total earnings vs duration times). Even though tip money is optional, the unspoken rule in earning for gigs workers is tip money after taking orders. Because gigs workers cannot put price on their own delivery service, the delivery cost might look high to customer sometimes, but it was not still enough without tip. For example, it costs a customer $10 for 10 mins delivery toward Doordash app. When I finish the delivery, I only receive $7 including tip, and Doordash took 30% my delivery earning. If I take 10 orders like that, I only earn almost $60 because the base pay varies by time. That is why service workers say tip money is appreciated.
For the total earning impact, tip money plays certain percentage on the total earnings, but not only me but also gigs workers cannot control the tip money demand toward the customers. Because I cannot control the demand of tip money, I had to be flexible in term of duration times and total orders so I can reach my earning goal. When I decide to change my working hour, it is an opportunity cost for me. I will reach the earning goal, but I will be tired or hungry or be running out of gas when driving home. If I get stressed on the tip money, I should not do this gigs job. However,
Even though I was trying everything I could on delivery time and quality of service, I might not get more tip money without asking my customers, and I could not say out loud toward them. That is the trade off I must accept.
The deviations from the dots representing San Francisco, Milpitas, Oakland, and Richmond do not provide the full picture of the driving force of tips money. The scatter plot only shows the ideal picture for the general Doordashers to go delivering food in the big cities with abundant of restaurants. In addition, heat map on the Doordash app also support that ideal picture when providing high base pay in those cities. However, if abundant of Doordashers gather those major cities, they would receive less orders even though the tips money might great sometimes. Because of the tricky heat map on the Doordash map, I had to be very flexible when delivering food that made me drive all over 10 different for 6 hours because waiting for the good tips money would only waste my time when waiting for the orders in the big cities like San Francisco, Berkeley, and Oakland.
4. Zero mean of the error term conditional on the independent variable:
For the uncertainties toward doing this gigs business, I made a list of uncertainties that might impact on the total orders and the duration times such as my car batteries died, the shortage of staff of the restaurants, the confusing direction from the customers toward dropping points, the unexpected car accident causing the traffic, I have to stay late for the full time job, etc. Those uncertainties only happened once during that month, and the uncertainties were extremely unpredictable. If those uncertainties happened during the delivery time, I would take a break or extend my hours in order to meet the earning goal. Looking at the big picture, the uncertainties only impact the earning. In addition, the uncertainties would affect my decision to continue to work or not. Toward the total orders and the duration times, the uncertainties could not have impacted on them, but it could indirectly affect them through my decision mind.
Therefore, if I become stressed toward the uncertainties, I should not do this job. Gigs workers need to have courage to do this job because they also must face the high risk and high reward during the rainy season. They could get paid extremely high with a lot of tips during the rainy hours especially dinner time. However, they would have to face with abundant of slippery roads and car accidents on the roads.
Doordash or Uber is aware of those uncertainties, so they have a backup plan by calling another gigs worker to come over to pick up the orders. For gigs with accident, they provided discount car towing and insurance for them. In the ends, the orders are going on in the uncertainty’s scenarios. However, the major impact was toward the earning but not total orders or the time duration. Total orders and duration times are independent variables, and they could be adjusted anytime depending on the person. At the end, this gigs job is a game of choice.
In Summary:
For credits, I appreciate the great development of technology on the Doordash app and Uber app, and I was able to collect data from my earning statements. From this model, I was able to pay 30% of my debt toward everyone especially my soon-to-be-ex-wife. When I was about to expand the model into different angle, the COVID-19 has impacted greatly on the global scale, and the shelter in place order has impacted the major percentage of service workers especially to Uber driver and Lyft driver. I suggested them being flexible by doing Instacart and Doordash that may help a small portion of their earning. At this time, I decided to step back to decide what I want to do in my life because I deferred Pharmacy school for one year already. However, I am not sure if I wanted to become a Pharmacist in future, and I am going to start with my why first not the pros vs cons. I have better figure it out SOON because life is too short.
I realized the value and how worth it was toward doing the gigs job. However, I did not see it for a long term even though I earn almost equivalent to my full-time job, but I could not figure out what I would like to do the rest of my life. I get too distracted to all the subject surrounding me, and I do not know what I am good at. Because of those reasons, I am not the king of money yet, but I am going to “Skin in the game” at every single subject for my Ultra learning.
P/s: I am thankful to all the author who wrote “Skin in the game”, “Ultralearning”, “Extreme Ownership”, “Atomic Habit”, and “Power of Habits”. I appreciate any readers leave the comments and feedback toward this article. I am still mastering my writing skill and my new habit.
