4 ms·

I'm sure HN has no shortage of ex-Uber engineers given how many people from their tech divisions they have let go of recently. I am curious to hear what they t

• jacobkg 6y ago
  Unfortunately in exchange for severance, these folks likely had to sign some sort of agreement limiting what they can say about the company
• bertil 6y ago
  Not Uber but Deliveroo — and it was long enough that I can talk freely. Mind you: that was a while ago. 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.
  • jacquesm 6y ago
    You are approaching this from the wrong end of the ledger. If someone orders food to go 'cheap' then they are not your customer if that squeezes out more than your delivery service needs because it will have to come out of your suppliers pockets and they are vulnerable already. The simplest way is to charge more to the buyer, and leave the restaurant their margins. Delivery is added value, not less value, and should command a higher price. And if your business model can't work that way then don't go into the delivery business in the first place.
    • bertil 6y ago
      > * asking customers for more than £2 is a crime against humanity and will tank your retention faster than spilling broth and bringing cold pizza; This was not my riffing around a hot-and-cold theme. This was the actual comparison that I made when I reported on the impact on retention of an AB-test on raising the order fee above £2: I used an established internal standard: unhappy meal (i.e. spilled liquid container or food is declared cold; either way the order is deemed inedible by the customer). Deliveroo (myself included) had a ton of ex-Facebook: opinions were not welcome, hard data was. Key levers like the fee was tested in countless ways (including a subscription model) and the company is well-known in London for their experimentation practice. I’d know: that’s what I’ve done full time since, including a workshop at last PyData London. You _can_ ask for more. Supper is a competitor to Deliveroo that offered a more premium experience in central London. Given how few restaurants they list and how expensive those restaurants are, I’m not sure they take a smaller commission: maybe the rate is lower, but not the amount. Fact is: no one has heard of them; no restaurant in this conversation recommends to use them; if you ask for options, their name don’t pop up. Last month, when all restaurants in London had to close, they were not mentioned as a potential option. Maybe they have the worst sales team in the world; I don’t know. But that’s when they barely take the fee to £3.25.
  • rdtwo 6y ago
    * 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. I think this is the area of most opportunity. You can’t really do this with a guy on a bike but in the us there is no reason you couldn’t establish delivery windows for certain neighbors and pickup windows to restaurants then have a guy with a meals on wheels like truck come, pick up items from adjacent restaurants and deliver them to customer. It messes up the delivery driver as disposable and replaceable mentality as timing is critical but if you want to have efficiency sometimes you need to pay a bit more.