No free lunch, but almost:

Food delivery app DoorDash has grown rapidly into the national industry leader, serving all 50 states and outpacing GrubHub, Uber Eats, Postmates, and other players with an estimated 35% marketshare in a rapidly-expanding segment. The company is reportedly preparing for an IPO and has been valued at a remarkable $12.6 billion. 

DoorDash has also been the focus of a sustained controversy over a pay model which misappropriated customer tips intended for workers. Under the pay model the company employed from 2017 until the fall of 2019, each dollar of customer tip did not raise workers’ earnings by a dollar — it simply meant the company paid the worker one dollar less. The company repeatedly defended this wildly unpopular tip-swiping pay model over several waves of public controversy, finally agreeing to change the policy in July, and only implementing the change in September. 

The company’s updated pay model, analyzed in this report, includes minimum pay of just $2 per delivery job. Reportedly, tips no longer directly factor into pay rates. Pay for each job is still set through a black-box algorithm which the company claims accounts for a wide range of various factors; however, workers have no access to the details of how those factors are weighted. They are simply offered a given job which pays a given amount, which they can accept or reject. (However, if a worker rejects too many jobs too frequently, they may receive fewer job offers as a result.) Further, when a job is offered to a worker, the amount of pay from DoorDash and customer tip are combined into a single “guaranteed amount,” with no breakdown available up-front as to how much is pay and how much is tip. This continued opacity has fed much skepticism about whether the company is still somehow factoring tips into pay.

DoorDash has reported that the new pay model has resulted in an increase in what they pay workers, and that gross earnings on the platform averages $18.54 per “active hour”. The company has declined to disclose how much of that gross amount is tips, how much is pay from the company, how many miles workers drive (and thus much much mileage expense they incur), what periods of time they consider to be part of an “active hour”, or other details and costs involved in achieving that reported level of gross earnings. However, we have been able to assemble a substantial amount of pay data directly from workers which provides a unique, independent view into what DoorDash is actually paying.

Our analysis of more than two hundred samples of pay data provided by DoorDash workers across the country finds that DoorDash pays the average worker an astonishingly low $1.45/hour, after accounting for the costs of mileage and additional payroll taxes borne by independent contractors. Nearly a third of jobs actually pay less than $0 after accounting for these basic expenses. Just 11% of jobs pay more than the federal minimum wage of $7.25/hour, after expenses, and only 2% meet the standard of $15 + expenses. Further, jobs with higher tips still tend to include lower pay: the set of jobs with tips of less than $1 pay 1.8 times as much as those jobs with tips of more than $8, on a gross hourly pay basis.

Our analysis shows that DoorDash has deployed a pay model which barely pays workers enough to cover their basic expenses — and sometimes, not even that much. Workers’ only significant earnings comes not from pay by the company, but from customer tips. And customer tips still seem to have an inverse relationship with pay.

The finding that pay is just $1.45/hour after expenses is low enough that it likely creates healthy skepticism. However, consider a job that pays DoorDash’s minimum rate of $2, requires 3 miles of driving from acceptance to the restaurant to the delivery location, and takes a half hour to complete. Accounting for the cost of the 3 miles at 58¢/mile reduces the $2 gross pay to $0.26. The additional 7.65% in payroll taxes paid by contractors reduces that to $0.24. Since it takes a half hour to complete the job, that results in a $0.48/hour rate for our sample job. This would be a perfectly “normal” job in DoorDash’s pay model — and pay almost nothing.

Our figures are also largely in line with DoorDash’s own public estimates. (This may explain why DoorDash has refused to release data on pay from the company, but instead focused on gross earnings including tips.) As stated above, the company recently claimed average gross earnings of $18.54/hour in October, including tips, and not accounting for any expenses. Our sample shows a roughly similar gross earnings of $15.76/hour for a sample collected in November and December. Keep in mind that DoorDash has not clearly explained how they define an “active hour” — for example, whether it includes time spent driving to a restaurant, or only includes restaurant wait time plus delivery time. (Our time figure accounts for all time for the acceptance of a job to the completion of that job.) That said, even if the figures are calculated on an identical basis, if our numbers are in fact too low by $2.78/hour, and if that entire $2.78/hour difference was entirely in pay from DoorDash rather than in tips, we’d still find average pay after expenses of just $4.23/hour. That is hardly an acceptable pay rate for a multi-billion-dollar corporation, and would not alter our basic conclusion.

In essence, the DoorDash pay model appears aimed at having workers deliver food effectively for free in hopes of collecting tips. 

How DoorDash’s pay model works

DoorDash sets the rate for jobs based on a black-box algorithm, which they claim accounts for time, distance, “effort,” and other factors, with minimum pay of $2 per job. (Presumably, the algorithm is designed to offer pay as low as possible in order to get a job completed.) Customers are encouraged to tip when placing an order, though under a recent change customers can also tip when an order is delivered. Default tip amounts are suggested, generally in flat-dollar amounts which vary depending on the size of the bill. Customers typically pay a standard $5.99 delivery fee and a service fee of about 10%, in addition to any tip. (Promotions frequently reduce these charges, the specific rates seem to be ever-changing, and service fees can vary from restaurant to restaurant.) Restaurants typically pay a percentage of the total food order to DoorDash as well.

When a job is offered to a worker, they see a “guaranteed amount” which includes the total of DoorDash pay and any up-front tip from a customer, as well as the restaurant name, and the total distance from their current location to the restaurant to the delivery location. The guaranteed amount is not broken down into separate figures for pay and tip until the worker completes the delivery. In some cases, DoorDash shows a lower guaranteed amount instead of showing the full amount of pay plus tip when a job is offered, only revealing the true, higher total when the job is completed. (This seems designed to encourage workers to accept lower-paying jobs in the hopes they may win the lottery of an “above-guarantee” job.)

Workers have about 30 seconds to accept or reject an offer when it is made available to them. If a job is rejected, it will be offered again to a different available worker, possibly at a higher rate; the company’s pay model is designed to slowly bid up the pay attached to a job until it is accepted.

The data

We assembled 229 job-level pay reports through outreach to workers across the country. For each job, workers used this form to provide the total pay from DoorDash and total customer tip, as well as the time and miles required to complete a job (from acceptance to delivery). All of this data is from November and December of this year, after the company’s new pay model was fully implemented nationwide.

DoorDash does not directly provide workers with all of the data necessary to calculate hourly equivalent earnings. Of course, this necessary information is already recorded by the company via built-in functions of their app (i.e. GPS and time-tracking); it is simply not consistently and transparently provided by the company to workers. For example, DoorDash does provide an estimate of total mileage when a job is offered, but not when a job is completed. (Workers must screenshot those mileage estimates in order to save them.) The app does break down the base pay, bonuses, and tip amounts for a job after it is complete, but no information is provided on total time elapsed or total miles driven from acceptance to delivery. Therefore time and mileage figures are estimates provided by workers, who have every incentive to track this information closely as it is critical to determining how much they are actually taking home after expenses.

Because this data is self-reported and the sample of workers was assembled through campaign organizing work rather than through a random sampling process, it is possible that there are some selection effects in our data. That said, the relative closeness of our gross pay figure to the similar datapoint dribbled out by the company makes us confident in our assessment that we have assembled a good cross-section of data and our aggregate pay rate is indicative, if inevitably not as accurate as it could be if we had access to DoorDash’s entire dataset.

The equivalent hour pay calculation

Gig economy platforms tend to focus on gross earnings — the total amount deposited in a given week, including tips, and not accounting for mileage and other expenses. While this is a way to approach the advertised rates of “earn up to $25/hour” it’s not really an accurate indicator of the amount a worker can expect to have to live on.

The purpose of the equivalent hourly pay calculation is to account for the basic expenses incurred by gig workers and create a baseline of comparison between gig jobs and with conventional employment. This is the basis on which gig workers have made the demand to be paid at least $15 + expenses for all time with an active job.

This simple formula accounts for two key costs not accounted for in gross earnings. The first is the additional “employer share” of payroll taxes which are paid by independent contractors but not employees, which works out to 7.65% of pay. A second and much larger added expense is the cost of driving (including gas, repairs, depreciation, etc). The IRS mileage rate (58¢/mile in 2019) is a standard, widely-accepted, and comprehensive estimate of the full costs of driving a typical vehicle. While this mileage cost would be a large aggregate line item on a corporate financial report if the company owned the vehicles, the magnitude of the expense can be difficult for an individual worker to gauge — it tends to be hidden on a job-by-job or day-to-day basis, but then shows up in a large repair or need to replace a vehicle far sooner than otherwise. 

It is worth emphasizing that we have used an extremely conservative calculation of hourly equivalent pay. It does not account for the cost of paying for basic public insurance benefits like workers compensation or unemployment; does not account for paid sick time or other time off; and does not address other work expenses beyond mileage and payroll taxes.

We count all time and all miles from when a job is accepted to when it is completed as time worked for the purposes of this calculation. Note that some apps only report time & miles from pickup location to delivery.

Correlation between tips and pay

Tips have been an ongoing area of controversy for DoorDash. While the company insists that tips are no longer factored into pay, many workers and customers are skeptical of this claim. In part this is due to the company’s many months of obfuscation about how their pay model previously did in fact account for tips, and the fact that they continue to defend this past practice. And in part this is due to the ongoing opacity of the company’s pay model, which makes it impossible to truly verify what is and is not included in the calculations.

Further, even if the company is no longer explicitly misleading customers and workers about its tip-handling practices, the company’s current pay model does have a mechanism which would nonetheless enable the company to pay less to workers on jobs where customers tip more:

• Because pay is calculated via a black-box algorithm instead of being based on time and miles or some other set of transparent factors, there is no way to verify what a job “should” pay.

• Workers are not given a separate breakdown between DoorDash pay and customer tip when a job is offered; they only see one combined “guarantee” which blends tip and pay.

• If a job offer is rejected because the guarantee is too low to be worth doing, the company’s pay model will offer that job again at a higher rate. 

• Because a) a lower tip means a lower guaranteed amount, b) rejected jobs will tend to be those offering lower guaranteed amounts, and c) rejected jobs are re-offered at higher guarantees by raising DoorDash pay, that means d) jobs which are rejected due to low tips will tend to end up paying more, while jobs with higher tips will be accepted despite lower pay from DoorDash.

Our dataset suggests this may in fact be happening: while the pattern is not perfectly sharp, we do find that overall, there is an inverse relationship between what customers tip and what DoorDash pays — that is, jobs with higher tips tend to have lower rates of pay from DoorDash than do jobs with lower tips. In fact, as the table below shows, jobs with higher tips tend to include less gross pay from DoorDash per hour, less gross pay per mile, and less net pay after expenses.

This inverse relationship between customer tip and DoorDash pay is particularly notable because in a more straightforward pay model, the opposite relationship ought to hold true: we would otherwise expect that jobs with higher-value customer orders would include more pay from DoorDash as well as higher tips.

While this mechanism of reappropriating customer tips is more complex and indirect than in the company’s previous pay model, the end result seems to be similar: higher customer tips result in lower pay from DoorDash.

Conclusion

DoorDash was the first major food delivery app to pay workers not directly for time, miles, pick-ups, deliveries, or another transparent basis, but rather via a black-box algorithm. They first developed this approach in 2017, and many other apps have followed.  

DoorDash was also the first major food delivery app to directly substitute tips for pay, and the last to relent and reverse this wildly unpopular scheme. 

Now it appears they may soon be the first to drive down pay to effectively zero after expenses. While customers and restaurants pay DoorDash handsomely for use of the platform, workers are left to deliver food essentially for free, with net earnings coming almost entirely from customer tips rather than company pay.

It further appears that DoorDash may have discovered a new way to use customer tips to substitute for pay, by training their algorithm to offer higher-tipping jobs at low pay, while bidding up pay on lower-tipping jobs.

This is alarming, but none of it is a wild conspiracy — it is the company’s short-term economic incentive to pay as little as possible for each delivery, as well as the company's stated goal. As reported in Fast Company, DoorDash’s “ideal world” is that the company “would pay its contractors almost nothing – relying on customer tips to cover most of the fee quoted for a job.” 

This does in fact seem to be exactly what our data shows they are achieving.

Left unregulated and excluded from basic labor standards, DoorDash and other gig economy companies will tend to exploit their vast resources and data to drive workers’ pay down to zero, turning after-expenses earnings into an optional bonus paid directly by the customer. It’s all the more reason why we need new laws that set pay standards for gig economy companies like DoorDash: $15 + expenses, tips on top, pay transparency, and more.