Former DD employee here. Can co-sign that the largest ordering segment was low SES millennials. I worked at Caviar which was eventually acquired by Doordash and my biggest shock post acquisition was how many low ticket size orders were going to low/middle income zips. Caviar only operated in big cities and our client base was wealthier. So it wasn't strange to see $200 of Han Dynasty going to the upper west side. But once inside Doordash, I was flabbergasted to see that their median order was a $12 junior bacon cheese burger meal going to a working-class neighborhood in Phoenix (+$10 in delivery fees, maybe a tip, maybe not. Sorry y'all. Doordash's hard data makes it clear that low income people are very bad tippers. $0 tips are WAY more common than you think. Doordash got sued for not giving all tip money directly to dashers. What you didn't see on the outside was that 100% of that money was making it to dashers, but they had some socialist redistribution going on to prevent anyone from ending up at the $0 tip. The lawsuit forced them to stop this practice)
I was so convinced that their business model was doomed to fail-propped up by stimulus checks from modest income housholds- that I sold all of my stock immediately after I left in '21. I greatly underestimated how many people are insanely bad with money.
Yes, there are some sympathetic cases in here of people with kids who want to save time but that's not most of what's going on. This is is driven by the same forces driving up gambling usage: a lack of cultural habits that engender pre-frontal cortex thinking/behavior.
I was definitely shocked in looking at this data to see that the average order was something like $30-35. When my partner and I order delivery it’s almost always a meal that would have cost $40-60 at the restaurant and is $60-90 when all the delivery fees and everything are included. If the order is much smaller, then the delivery fees are going to have to be an even bigger proportion of the price.
But I also notice that the apps usually assume I’m going to make a fast food order. I haven’t quite understood what is the scenario in which ordering fast food delivery makes sense - the big advantages of fast food are speed and price, and delivery undoes those.
I don't know what their numbers are today, but fast food orders were a plurality if not a majority of orders when I was there. I also found this shocking
I'm very confused about the data analysis here. It sounded like you had a bunch of anonymized credit card transactions with amount/date/age/location. And then you did something with census microdata that let you figure out how much that person was making? How does that work?
It sounds to me like the incomes were in or inferred from the original credit card data, and they used the population data to extrapolate to the broader population outside of their credit card dataset. I am also curious about exactly how that worked though.
When I read the headline and the opening anecdotes - about people spending hundreds of dollars on delivery each week while they earn $50,000 a year - I was shocked, and expected the data to show *a lot* of DoorDash transactions. But, in fact, even in the group that uses the service most often the spend is about 1% of income.
That factor should be given more weight in the writeup, or you should explain why you think DoorDash deliveries are only a very small part of the overall volume of deliveries from all delivery services. Based on the DoorDash data, the people in your splashy opening anecdotes are highly, highly unusual consumers - spending 30% of their total income or more on delivery – and so a *very* misleading illustration of the typical pattern.
It is interesting, I do take the point, that people in lower income categories are spending more as a percentage of income than people who are wealthier. One speculation: they can’t afford more expensive luxuries so this is a luxury they allow themselves once in a while. But very few people seem to be driving themselves to ruin like the very poor decision-makers in the anecdotes that lead the story.
If you’re going to write a methodological note, at least tell us where the card data is coming from. Is it from the JP Morgan Institute? Do we know whether it’s a representative sample? Chase has a partnership with DoorDash for the Sapphire Reserve - if we’re disproportionately getting a lot of card data from 30 year olds that have a Sapphire Reserve card, then I question whether we should take these conclusions at face value. “39 million people” can go pretty far to reducing sampling bias, but not all the way.
Another note is that whether food delivery rises with affluence today is beside the point. Matt’s claim is that the only reason people in the bottom half of the income distribution can afford DoorDash at all is because they’re richer than the equivalent cohort from 40 years ago. That does not entail, at all, that richer people today drive more DoorDash sales. It’s easy to tell a story where lower income people have more manual jobs, access to worse ingredients, and irregular schedules such that cooking is much more of a hassle after a long day than it is for higher income people.
This dataset came from millions of people and includes both credit and debit card data. The data was not from a specific credit card provider, and the data provider gave us weights to adjust for skews in the data. Specifically, we used a slice of the MBSH3 dataset described on the Yale DISSC website: https://dissc.yale.edu/data-resources.
I wonder if another word for "people in their early 30s and 40s who don’t make very much" is "people with small children who can't afford paid assistance with either cooking or childcare, two somewhat time-consuming tasks which frequently have to be done simultaneously and are not easy to multitask"???
Do you seriously think Millennials are the first generation who've had to cook with small children in the house, and without the benefit of a live-in nanny or chef?
(The alternatives being, I assume, some combination of “pay less attention to your kids than you would like”, “cook worse food than you could get on DoorDash”, “unpaid assistance”, and “be extremely exhausted”)
Previous generations in this phase led to the rise of fast food and drive throughs. Millennials are just the ones that came of age at the time the Internet had enabled universal delivery apps.
Someday, the food desert hypothesis will look up, screaming for help. And I will whisper, "You lack conceptual rigor, and any compelling evidence of effect."
Noting this is household income not personal income, I wonder if single people are ordering more delivery? Feels anecdotally true. Somehow feels not worth the effort to just cook for a solo meal at home.
I do wish the data had been broken out by household size! I couldn’t tell if this is families with kids ordering delivery every week, or singles doing it a lot while no one else does, or what.
As someone who used to witness a lot of Doordashing by this exact demographic, I wish someone would break down the times of the transactions. I used to work nights in a call center. We had half-hour lunches. I always packed my lunch, but a lot of my coworkers would get delivery. Why? Because they wanted to enjoy their half hour lunch break instead of spending the whole thing driving to get their takeout and driving it back. And this is in California, where a half hour lunch break is mandatory. I imagine there are a lot of states where lunch breaks are not mandatory, so employees order Door Dash for lunch because if they left to go get lunch, that would be unpaid time and they'd have to leave later to make up the shift. Maybe some of them aren't allowed to leave at all, so their only options are a packed lunch or delivery. I'll say also, most of these people weren't ordering it every day, but they'd do like once a week as a treat, or if they had a really stressful call or something, or if we had to work overtime, which we generally only found out we had to do the day of. I bet a lot of hourly jobs have these dynamics.
I have coworkers who have incredible flexibility and still regularly use it. And it doesn't match the other people I know who do either A plausible reason for some of it though
“average broke millennial orders 18 burrito taxis a year” factoid actualy just statistical error. average broke millennial orders 0 burrito taxis per year. Burrito Taxi Georg, who lives in cave & orders over 10,000 each day, is an outlier adn should not have been counted
Given that these were aggregated figures, I really would not be surprised if spending was driven by "super-users" in these demographics. Anecdotal but I have a friend who orders dinner every night from one of these delivery apps. No one else I know does this - it's maybe once a week if ever.
I think the real issue is that rich people are more in control of which hours they work than poor people. As a salaried project manager, I have pretty good control or at least advance notice over when my lunch takes place and whether I can go out to get it. As a call center employee, I had to coordinate with my other coworkers to make sure the phones were manned before taking a break, and I wouldn't find out whether I was doing overtime until the day of the shift.
I would be curious to see if there's any data to allow you to test my own personal hypothesis that there is a correlation between food delivery usage and limited transportation options (such as not having a car and living in an area with poor mass transit). I'd also be interested to find out if there's any alignment between food delivery usage and rideshare usage.
For fifty years those of us on the left have been overly creative of unsound decision-making among the disadvantaged on pretty much any policy issue. Wild Rube Goldberg explanations and then solutions that won't solve anything. This article does a great job in making those cases and then systematically dismissing them. Lower income people are human like the rest of us and make lots of bad decisions whether it is having kids or spending on sugary food or whatever. We all make bad decisions but if you have money it works out. We need to get away from policymaking built around nanny state thinking (looking at you Liz Warren and all the NGOs). Poor people need more money. Funnel it to them. Simple as that.
I’ve long heard that millennials are the one group who most continued with grocery delivery after the pandemic. I thought of this as older generations having less internet native habits so they gave up when the pandemic ended, and Gen Z having lower household incomes, but I’m not sure. Especially since in this case household income goes the opposite way.
As some other threads suggested, it would be interesting to know household structure here - is this parents who are barely managing with their kids, or is it singles in underpaid work?
Okay but then why is age a bigger factor than income? People 30-44 spend the most out of the group making <$50K, but they also spend the most out of the group making $50K-$100K, and they also spend the most out of the group making $100+K.
This age group didn't grow up with Doordash like the people younger than themselves did; they had to change their behavior as adults. And yet the people even older than them and in the same income bands didn't also change their behavior.
Cohort effects, you see similar things with smoking, alcohol, drug use. Consumer habits tend to solidify when people are 18-25ish. When people in this age bracket were 18-25, everything was subsidized by VC and food delivery was less expensive.
I might’ve overlooked this somewhere, but you’re only looking at the data for *delivery* not *takeout* (which can also be ordered through companies like DoorDash), right?
Our dataset covers any purchase from DoorDash, Grubhub, Postmates, and Uber Eats, so that would include both delivery and takeout orders; however, takeout is likely a small share of the business these companies do.
Aren't the prices higher regardless? It seems silly to use doordash for takeout (unless you have a gift card and don't want to wait an hour for your food.)
"it took ... 30 minutes to review Claude’s work. If we had done this whole project without the use of AI, it probably would have taken us closer to 10 hours. (Rest assured that we carefully reviewed all of the code and analysis that Claude produced.)"
30 minutes of review is not plausibly enough to carefully review an analysis which would take you around 10 hours. Especially given ai code + methods is often extremely verbose.
Former DD employee here. Can co-sign that the largest ordering segment was low SES millennials. I worked at Caviar which was eventually acquired by Doordash and my biggest shock post acquisition was how many low ticket size orders were going to low/middle income zips. Caviar only operated in big cities and our client base was wealthier. So it wasn't strange to see $200 of Han Dynasty going to the upper west side. But once inside Doordash, I was flabbergasted to see that their median order was a $12 junior bacon cheese burger meal going to a working-class neighborhood in Phoenix (+$10 in delivery fees, maybe a tip, maybe not. Sorry y'all. Doordash's hard data makes it clear that low income people are very bad tippers. $0 tips are WAY more common than you think. Doordash got sued for not giving all tip money directly to dashers. What you didn't see on the outside was that 100% of that money was making it to dashers, but they had some socialist redistribution going on to prevent anyone from ending up at the $0 tip. The lawsuit forced them to stop this practice)
I was so convinced that their business model was doomed to fail-propped up by stimulus checks from modest income housholds- that I sold all of my stock immediately after I left in '21. I greatly underestimated how many people are insanely bad with money.
Yes, there are some sympathetic cases in here of people with kids who want to save time but that's not most of what's going on. This is is driven by the same forces driving up gambling usage: a lack of cultural habits that engender pre-frontal cortex thinking/behavior.
I was definitely shocked in looking at this data to see that the average order was something like $30-35. When my partner and I order delivery it’s almost always a meal that would have cost $40-60 at the restaurant and is $60-90 when all the delivery fees and everything are included. If the order is much smaller, then the delivery fees are going to have to be an even bigger proportion of the price.
But I also notice that the apps usually assume I’m going to make a fast food order. I haven’t quite understood what is the scenario in which ordering fast food delivery makes sense - the big advantages of fast food are speed and price, and delivery undoes those.
I don't know what their numbers are today, but fast food orders were a plurality if not a majority of orders when I was there. I also found this shocking
I'm very confused about the data analysis here. It sounded like you had a bunch of anonymized credit card transactions with amount/date/age/location. And then you did something with census microdata that let you figure out how much that person was making? How does that work?
Income was from the card transaction data, not from the Census microdata.
What did the census microdata add then?
It sounds to me like the incomes were in or inferred from the original credit card data, and they used the population data to extrapolate to the broader population outside of their credit card dataset. I am also curious about exactly how that worked though.
When I read the headline and the opening anecdotes - about people spending hundreds of dollars on delivery each week while they earn $50,000 a year - I was shocked, and expected the data to show *a lot* of DoorDash transactions. But, in fact, even in the group that uses the service most often the spend is about 1% of income.
That factor should be given more weight in the writeup, or you should explain why you think DoorDash deliveries are only a very small part of the overall volume of deliveries from all delivery services. Based on the DoorDash data, the people in your splashy opening anecdotes are highly, highly unusual consumers - spending 30% of their total income or more on delivery – and so a *very* misleading illustration of the typical pattern.
It is interesting, I do take the point, that people in lower income categories are spending more as a percentage of income than people who are wealthier. One speculation: they can’t afford more expensive luxuries so this is a luxury they allow themselves once in a while. But very few people seem to be driving themselves to ruin like the very poor decision-makers in the anecdotes that lead the story.
If you’re going to write a methodological note, at least tell us where the card data is coming from. Is it from the JP Morgan Institute? Do we know whether it’s a representative sample? Chase has a partnership with DoorDash for the Sapphire Reserve - if we’re disproportionately getting a lot of card data from 30 year olds that have a Sapphire Reserve card, then I question whether we should take these conclusions at face value. “39 million people” can go pretty far to reducing sampling bias, but not all the way.
Another note is that whether food delivery rises with affluence today is beside the point. Matt’s claim is that the only reason people in the bottom half of the income distribution can afford DoorDash at all is because they’re richer than the equivalent cohort from 40 years ago. That does not entail, at all, that richer people today drive more DoorDash sales. It’s easy to tell a story where lower income people have more manual jobs, access to worse ingredients, and irregular schedules such that cooking is much more of a hassle after a long day than it is for higher income people.
This dataset came from millions of people and includes both credit and debit card data. The data was not from a specific credit card provider, and the data provider gave us weights to adjust for skews in the data. Specifically, we used a slice of the MBSH3 dataset described on the Yale DISSC website: https://dissc.yale.edu/data-resources.
I wonder if another word for "people in their early 30s and 40s who don’t make very much" is "people with small children who can't afford paid assistance with either cooking or childcare, two somewhat time-consuming tasks which frequently have to be done simultaneously and are not easy to multitask"???
Do you seriously think Millennials are the first generation who've had to cook with small children in the house, and without the benefit of a live-in nanny or chef?
Well, no, just the first generation for whom DoorDash is a viable solution and, depending on what you value, plausibly better than the alternatives!
(The alternatives being, I assume, some combination of “pay less attention to your kids than you would like”, “cook worse food than you could get on DoorDash”, “unpaid assistance”, and “be extremely exhausted”)
Previous generations in this phase led to the rise of fast food and drive throughs. Millennials are just the ones that came of age at the time the Internet had enabled universal delivery apps.
Bean burritos FTW.
Cheap and easy
Someday, the food desert hypothesis will look up, screaming for help. And I will whisper, "You lack conceptual rigor, and any compelling evidence of effect."
Noting this is household income not personal income, I wonder if single people are ordering more delivery? Feels anecdotally true. Somehow feels not worth the effort to just cook for a solo meal at home.
I do wish the data had been broken out by household size! I couldn’t tell if this is families with kids ordering delivery every week, or singles doing it a lot while no one else does, or what.
A pot of soup/stew can last the better part of a week. Much my cooking is for multiple meals.
As someone who used to witness a lot of Doordashing by this exact demographic, I wish someone would break down the times of the transactions. I used to work nights in a call center. We had half-hour lunches. I always packed my lunch, but a lot of my coworkers would get delivery. Why? Because they wanted to enjoy their half hour lunch break instead of spending the whole thing driving to get their takeout and driving it back. And this is in California, where a half hour lunch break is mandatory. I imagine there are a lot of states where lunch breaks are not mandatory, so employees order Door Dash for lunch because if they left to go get lunch, that would be unpaid time and they'd have to leave later to make up the shift. Maybe some of them aren't allowed to leave at all, so their only options are a packed lunch or delivery. I'll say also, most of these people weren't ordering it every day, but they'd do like once a week as a treat, or if they had a really stressful call or something, or if we had to work overtime, which we generally only found out we had to do the day of. I bet a lot of hourly jobs have these dynamics.
I have coworkers who have incredible flexibility and still regularly use it. And it doesn't match the other people I know who do either A plausible reason for some of it though
“average broke millennial orders 18 burrito taxis a year” factoid actualy just statistical error. average broke millennial orders 0 burrito taxis per year. Burrito Taxi Georg, who lives in cave & orders over 10,000 each day, is an outlier adn should not have been counted
Given that these were aggregated figures, I really would not be surprised if spending was driven by "super-users" in these demographics. Anecdotal but I have a friend who orders dinner every night from one of these delivery apps. No one else I know does this - it's maybe once a week if ever.
>But also, poor people might be more likely to be short on time or unable to afford other splurges.
Like the food desert hypothesis, this is more of a shibboleth than anything. Rich people (almost by definition) work more hours than poor people.
I think the real issue is that rich people are more in control of which hours they work than poor people. As a salaried project manager, I have pretty good control or at least advance notice over when my lunch takes place and whether I can go out to get it. As a call center employee, I had to coordinate with my other coworkers to make sure the phones were manned before taking a break, and I wouldn't find out whether I was doing overtime until the day of the shift.
If only there was some way to bring lunch to work.
I literally go out to eat for lunch only a couple of times a year.
I would be curious to see if there's any data to allow you to test my own personal hypothesis that there is a correlation between food delivery usage and limited transportation options (such as not having a car and living in an area with poor mass transit). I'd also be interested to find out if there's any alignment between food delivery usage and rideshare usage.
For fifty years those of us on the left have been overly creative of unsound decision-making among the disadvantaged on pretty much any policy issue. Wild Rube Goldberg explanations and then solutions that won't solve anything. This article does a great job in making those cases and then systematically dismissing them. Lower income people are human like the rest of us and make lots of bad decisions whether it is having kids or spending on sugary food or whatever. We all make bad decisions but if you have money it works out. We need to get away from policymaking built around nanny state thinking (looking at you Liz Warren and all the NGOs). Poor people need more money. Funnel it to them. Simple as that.
This is one of those cases where the answer as derived from the statistics doesn't tell a very compelling story, and it's hard to make it make sense.
Why would poor millennials use Doordash more than other groups? I really have no idea.
I’ve long heard that millennials are the one group who most continued with grocery delivery after the pandemic. I thought of this as older generations having less internet native habits so they gave up when the pandemic ended, and Gen Z having lower household incomes, but I’m not sure. Especially since in this case household income goes the opposite way.
As some other threads suggested, it would be interesting to know household structure here - is this parents who are barely managing with their kids, or is it singles in underpaid work?
My belief is that it's because they're hourly workers without long enough lunch breaks.
Maybe wealthier people are spending the money instead at more expensive sit-down restaurants?
Poor people have poor impulse control. DoorDash is a vice analog, like cigarettes, candy, etc.
Okay but then why is age a bigger factor than income? People 30-44 spend the most out of the group making <$50K, but they also spend the most out of the group making $50K-$100K, and they also spend the most out of the group making $100+K.
This age group didn't grow up with Doordash like the people younger than themselves did; they had to change their behavior as adults. And yet the people even older than them and in the same income bands didn't also change their behavior.
Cohort effects, you see similar things with smoking, alcohol, drug use. Consumer habits tend to solidify when people are 18-25ish. When people in this age bracket were 18-25, everything was subsidized by VC and food delivery was less expensive.
I might’ve overlooked this somewhere, but you’re only looking at the data for *delivery* not *takeout* (which can also be ordered through companies like DoorDash), right?
Our dataset covers any purchase from DoorDash, Grubhub, Postmates, and Uber Eats, so that would include both delivery and takeout orders; however, takeout is likely a small share of the business these companies do.
Aren't the prices higher regardless? It seems silly to use doordash for takeout (unless you have a gift card and don't want to wait an hour for your food.)
"it took ... 30 minutes to review Claude’s work. If we had done this whole project without the use of AI, it probably would have taken us closer to 10 hours. (Rest assured that we carefully reviewed all of the code and analysis that Claude produced.)"
30 minutes of review is not plausibly enough to carefully review an analysis which would take you around 10 hours. Especially given ai code + methods is often extremely verbose.
Well, duh. The rich people are all on GLP-1s.