The vast majority of people underestimate the cost of delivery, more specifically two things: the duration of delivery or rather how many deliveries per hour of work at peak time, and the utilisation rate.
It’s hard, in the best circumstances (good weather, constant demand, professional restaurant, dense habitat) to have a driver handle two deliveries per hour. If you take a £2 free from the customer and a £5 commission over a £17 order, you can pay a rider £14 per hour. At that rate, you typically would get the left wing press to call you exploitative every other day. That’s excluding everything else: marketing, customer service, tech. You can’t improve anything without capital that has to come from investors.
All that ignore the key problem availability: if you have 100 riders on schedule, about 60 would show up. They are not employees after all, and you can’t fire half of your contractors every day when you are growing like weed. Probably fewer than 30 will show up if it rains, but 80 might if the weather looks nice and there’s nothing on TV. If you have reasons to believe that you’ll get 360 deliveries during the three hour dinner shift, you should be good but… if you get 350 orders and 70 riders show up (nothing unusual exceptional) they’ll all get one fewer order than expected and they will complain they are are not making enough money. More likely, half will do six, a third will do five and a dozen not enough to justify them showing up. That’s excluding any exceptional case: hail, Premier League finals, etc. 20 riders for 4,000 orders wasn’t out of the question with bad enough weather. Your marketing team is quite likely to tell you that means 3950 unhappy customers because most riders will switch off the app after the second of third customer insults them because the website isn’t responsive.
There are many ideas about how to increase the number of delivery per hour of work:
* telling people to pedal faster isn’t a good idea;
* asking more than 30% from restaurants will get chef’s knife thrown your way (not a metaphor);
* paying people any less than £14/h gets you call a slave trader by the international press;
* asking customers for more than £2 is a crime against humanity and will tank your retention faster than spilling broth and bringing cold pizza;
* setting an ordering minimum to something as high as a meal for two people, i.e. £20, is an effective way to become a trending topic on twitter for all the wrong reasons; reasons that my grandmother would wash out of your mouth with soap;
* handling over two deliveries from the same restaurant to a single rider can work at times, but it’s hard to find the rare good cases and you get very angry customers ver fast; handling two deliveries from different restaurants is… ::shivers:: Let’s not talk about it.
So, you are right: razor thin margins at best in most cases.
What works is looking at the above and seeing in as a stats game:
1. 360 orders ± 30, 60 riders ± 20 isn’t a great combination. 3,600±100 and 500±50 are better: your margin can be improved with just the large number theorem. That means, in any places without the population density of central Paris, to feed a significant portion of the population. It sounds absurd but having 20 riders out in a hail storm means you will have none within an hour, even if the sun goes back (thank you non-sensical May weather); having 200 means they see their peers riding and wait it out.
2. Another thing where scale really helps is opportunities: if you have more than 3,000 deliveries in three hours per square mile, that’s 50 deliveries in the last three minutes. One is bound to be nearby, less than two minutes away. So rather than pay riders to do delivery, then shlep for the next one to a restaurant half a mile away and only then to start carrying food again, you get to pay them to deliver food, then turn the corner and do it again. That way, you can hope riders to handle more than two deliveries per hour. That considerably helps your business models.
3. What can help a little bit more is that with ten times more order, the rare cases when one rider can pick up two orders and not make it awkward increases as O(n^2) so they go from rare to worth the time to implement an assignment logic for those.
That should give you a hint why you’ve heard the words “billion raised in a new round of financing” used around that business model: there is path to profitability but it has to go through extreme growth.
I’ll let you do the math on rider pay, restaurant, fees, possible subscription models, population density, etc. but if you do, remember that a service like that requires, at maturity, thousands of engineers hoping Silicon-Valley level pay.