How much do Doordash drivers make? We analyzed 4500+ deliveries in an effort to better understand how workers are paid. Here’s what we found:

Distribution of Estimated Net Earnings / Online Hour ($)

QUICK LINKS:

• Exec. Summary
• Methodology
• Full Analysis
• Conclusion
• GigCompare: Who are we?

EXECUTIVE SUMMARY:

This report was produced by GigCompare. Our mission is to bring greater transparency to the gig economy and help workers answer the questions that are important to them. We set out to field test GigCompare’s methods and technology in an effort to better understand how Doordash workers are being paid. Here are our key findings so far:

1. Hourly Doordash earnings appear to be highly variable and unpredictable. This holds for multiple definitions of pay/hour. While our estimate for median net earnings per Online Hour for this sample is $12.68, we estimate that roughly 50% of workers earned less than that. The U.S. effective minimum wage is estimated to be $12/hour, which suggests that many workers made less than minimum wage during this pay period.
2. Over 50% of earnings for Doordash workers in this sample came in the form of customer tips.
3. Tips, as a % of total earnings, are also highly variable and unpredictable. This variability is likely what drives the overall distribution of total earnings.
4. Doordash makes public claims of earnings opportunities that focus on “Average Earnings” or “Make up to $25 / hour.” These statements implicitly acknowledge the wide distribution of earnings without actually addressing it, as fewer than 5% of the workers in this sample made $25/hour (or more) when accounting for all time worked and expenses.

METHODOLOGY:

We gathered over 100 weekly pay statement submissions from Doordash workers across the United States, representing over 4500 individual jobs done during the summer of 2020. We acquired this sample through digital advertising + direct relationships with workers we’ve met through our research. We calculated hourly pay figures for each weekly pay statement individually, then analyzed the distribution, as opposed to aggregating all stats together before performing calculations.

Some important acknowledgements before diving into the results:

1. The sample size is small relative to the total population, and has selection bias. We make no claims that this report is an accurate representation of the entire population of Doordash workers in the United States. It’s primarily a first step to demonstrate our approach and methodology.
2. For the purposes of defining expenses and minimum wage: we chose to define the U.S. minimum wage at $12/hour, which is close to the effective average in the United States. In future studies, we will anchor results to local minimum wage. However, this data set was not location-specific.
3. We estimated that workers drive 5 miles total for a job end-to-end, on average. This number comes from a prior attempt to answer the “mileage” question and from estimates that workers provided to us anecdotally. Moving forward, GigCompare will ask users to be more specific in their mileage estimates.
4. We use the IRS per-mile rate of $.575 to estimate expenses. In the future, we will attempt to make more precise estimates of expenses using local and vehicle-type averages.
5. Our full rationale for the definitions of time worked and expenses can be found here, but in short, we assume by default that all Online Time is working time, though we excluded statements where Active Time / Online Time was less than 50%.
6. We collected information via pay statement screenshots using GigCompare’s Calculate & Compare tool, and we acknowledge that it’s not impossible to produce fake images. However, we’re small enough that we don’t believe anyone would have taken the time to do so during this test. Long-term, we are working on ways to mitigate this risk.

FULL ANALYSIS:

ROLL THE DICE

Working for Doordash appears to be an incredibly inconsistent and unreliable way to make a living. We found the distribution of hourly earnings, regardless of how it’s defined, to be incredibly broad (figs 1–3).

Figure 1. Distribution of total earnings divided by Online Hours ($). Median: $19.40 STDEV: $5.93. “Gross Earnings” represents all of the money paid to the worker, including tips. “Online Hours” represents all time that the worker was logged into the Doordash platform.

Figure 2. Distribution of Gross Earnings / Active Time ($). Median: $26.73 STDEV: $6.95. “Gross Earnings” represents all of the money paid to the worker, including tips. “Active Time” represents all time that the worker was actively assigned to a delivery.

Figure 3. Distribution of Est. Net Earnings / Online Hour ($). Includes Estimated Expenses. Median: $12.68 STDEV: $5.22. “Net Earnings” represents all of the money paid to the worker, including tips, with total estimated expenses subtracted. Total estimated expenses were calculated using a 5 mile / order average and a $.575 cost/mile for the worker. “Online Hours” represents all time that the worker was logged into the Doordash platform.

Upon further investigation, that variance is due to the fact that over 50% of pay comes from tips (median: 53.33% for this group), and that this percentage is so inconsistent that it likely accounts for the difference in hourly earnings (fig. 4).

Figure 4. Distribution of Tips as percentage of Gross Earnings. Median: 53.33% STDEV: 15.53%. “Tips” represents earnings in the form of tips from customers. “Gross Earnings” represents all money paid to the worker, including tips and pay from Doordash directly.

The pay/job figure (Fig. 5) varied widely, suggesting workers are likely assigned to complete a certain number of jobs that are highly unprofitable. In many cases, they are not given enough time or information to make informed decisions on which jobs are profitable, which results in a vegas-like dynamic where luck plays a larger role than skill in determining hourly pay.

Figure 5. Distribution of Gross Earnings / Job ($). Median: $8.41 STDEV: $2.03. “Gross Earnings” represents the total amount paid to the worker, including tips. “Jobs” represents the number of deliveries a worker made during the period.

ACTIVE VS. ONLINE TIME

How much time on average do workers spend logged into their app in-between jobs? For this sample, the median Active Time / Online Time was 75.5%, with over 70% of workers between 67% and 90%. Workers are not compensated for time and mileage between jobs.

Figure 6. Distribution of percentage of Active Time vs. Online Time. Median: $75.5% STDEV: 10.05%. The % here is calculated by dividing “Active Time,” or the amount of time a worker spent assigned to a delivery, by “Online Time,” the amount of time spent logged into the Doordash app. We excluded statements where the share of Active Time was less than 50%.

ESTIMATED NET ONLINE EARNINGS/HOUR

The median Est. Net Earnings / Online Hour was just a hair above the effective minimum wage estimate at $12.68 / Online Hour (see fig. 7). While this number could be used to argue that Doordash is paying above U.S. effective minimum wage, that argument would be misleading. If we assume a $12 minimum wage, nearly half of the Doordash workers in this sample are making less. Gig apps are intimately familiar with this variability, which is why their marketing focuses on statements like “make up to $25/hour,” which implicitly acknowledges the wide distribution without actually addressing it. Indeed, in this sample, fewer than 5% of workers made $25/hour or more when accounting for all time worked and estimated expenses.

Figure 7. Distribution of Est. Net Earnings / Online Hour ($). Includes Estimated Expenses. Median: $12.68 STDEV: $5.22. “Net Earnings” represents all of the money paid to the worker, including tips, with total estimated expenses subtracted. Total estimated expenses were calculated using a 5 mile / order average and a $.575 cost/mile for the worker. “Online Hours” represents all time that the worker was logged into the Doordash platform.

This ad is currently active on Facebook. Note the “up to $25/hour.” In our sample, we estimate that fewer than 5% of workers earned this much after expenses.

CONCLUSIONS, NEXT STEPS

Much of the debate around gig worker pay has focused on average earnings, when perhaps it should be focused on what appears to be a large % of the workforce who are making less than minimum wage, after expenses. The findings of this first report by GigCompare should be treated as directional, and we intend to produce more authoritative analysis of Doordash and other gig apps in the near future. It would be trivial for Doordash, or any gig app for that matter, to produce a similar report using a full set of data, location-specific minimum wages, and worker-specific expense estimates. We invite them to do so.

Until then, GigCompare will continue to reassemble this puzzle.

Stay tuned!

QUICK FOOTNOTE ON PROP. 22

The debate around gig pay is playing out before our eyes in California, with a proposed ballot measure written by gig apps that would exempt it from employment obligations and set “pay standards” for app-based workers. One of the elements of the app-written proposition includes an “earnings guarantee,” which sounds like a minimum wage, but is actually just a pay floor formula tied to Active Time. It’s possible to retroactively estimate whether the pay floor would have compelled an app to raise pay for a worker using an old pay statement.

Prior work has been done to estimate what this “pay floor” would mean for workers, on average, and it’s far from encouraging. We’ve built a simulator directly into GigCompare for California workers to test with their own prior earnings statements, and we’ll report on aggregate findings soon.

GIGCOMPARE: WHO ARE WE?

GigCompare was founded in the Spring of 2020. Our mission is to bring greater transparency to the gig economy and answer the questions that are important to workers:

1. How much do gig workers really make per hour, after expenses?
2. How does this number differ between apps and in different cities? For example, who is paying the most in Seattle this week, and how much are they paying after expenses?
3. How predictable (or unpredictable) are earnings on a day-by-day basis for individual workers on each app?
4. How are all of the above changing over time?

We’re far from the first to ask these questions, and the sense of urgency surrounding them has only increased during the pandemic. The answers put forward to-date have incited debate, and ultimately have fallen short of providing timely value to workers. In short, the answers only exist in two places.

Think of all gig economy pay data as an assembled jigsaw puzzle:

1. Gig companies can see the full picture, and they employ teams of data scientists to study it and make decisions that align with their interests. They do not share aggregate pay data externally willingly.
2. Each gig worker has a single piece of the puzzle in the form of their earnings statements. On its own, each puzzle piece is relatively useless. But if gig workers were able to collectively reassemble the puzzle, they too would be able to see the full picture and use it to make better decisions.

To date, it has been challenging to reassemble the puzzle. It’s labor-intensive and expensive to gather data, most methods rely on workers to invest significant time to decipher and submit their own pay statements. As a result, the most robust answers to our questions have come from studies funded by gig apps themselves or are limited to a specific point in time and space.

What’s the alternative?

We believe that answering these questions effectively will help workers make better and more informed decisions, and ultimately put pressure on gig apps to do more for their workforce. We also believe that an independent third party who makes this information freely available and useful to workers in real-time will have an easier time collecting it in the first place. The operative word here is independent: GigCompare will never accept money from a gig economy app in any form.

The first step was to make it as easy as possible for a worker to anonymously submit their piece of the puzzle, and give them a good reason to do so. At the same time, we need to be able to trust the accuracy of the submitted information.

Existing pay calculators and submission forms (an early iteration of GigCompare included) rely on users manually copying information from a pay statement into a web form containing 10+ fields. Not only is this time-consuming, the temptation to skip fields and/or estimate the values is real and completely understandable. Indeed, the first iteration of GigCompare’s calculator saw more than half of the submissions contain numbers far too round to be realistic.

The solution to both of these challenges was to develop a way to extract information directly from pay statements. This reduces the time it takes to submit a puzzle piece to less than 60 seconds, and reduces the chances of inaccurate data submission. Over the past month, we’ve built this functionality for Doordash, Instacart, Uber, and UberEats pay statements, with others in development. Which brings us to the results of our first few weeks of data collection.

We pursued this first report partially as a field test for GigCompare’s newly developed Calculate & Compare tool. Processing 100+ pay statement screenshots and producing the numbers used in this report took only a few minutes, as all of the extraction and calculation processes are fully-automated. On average, it took a worker less than 90 seconds to complete the Calculate and Compare process. We’ll be automating even further, with the goal of providing free access to real-time analysis of all gig platform pay trends to GigCompare contributors.