Also, corporate overhead doubled in 2020 despite then firing half their “director and global management” team and shrinking the other divisions. I wonder how much segment level costs were stuff in the corporate line item to make it seem like the business is profitable on a segment/country level.
1. Why can't Deliveroo or Uber Eats just go out and capture the marketplace restaurants? It seems that TKWY is relying on these marketplace restaurants in the UK, but what prevents Roo to do in marketplace what TKWY is doing in logistics? In fact, it costs much more money for TKWY to add a 1P restaurantwhere they are handling logistics vs. Roo adding a 3P restaurant where they just put them on the platform. If you are a marketplace restaurant, it costs you nothing to be added to another app and even if you capture 1-2% more sales, it is already a win.
2. I see your point on white label, but at least you are capturing some demand. For example, if there 100 customers interest in Nandos day 1 and Roo captures 100 of them. In day 2, Nando incentivizes them to go straight to Nandos and Roo captures 50 of them. So yes Roo lost 50 day 2, but guess what - TKWY lost all 100 day 1 when Nandos is no longer on their app. So overall, it seems this movement is negative for the entire space, but Roo is still better positioned v TKWY if they have Nandos than if they do not.
it is a lot harder to capture marketplace restaurants than it seems. There hasn’t been a single country where a logistics player have built a sizable marketplace offering against an incumbent marketplace vendor. Deliveroo has tried for many years to no avail. I am not going to do the work for you, but ask yourself why hasn’t Expedia or Carguru had any success in Europe? Cant they offer 0% take rate as well?
On white label, you are right. Except that Nando’s is not on the JET platform at all. And guess what, TKWY can easily win the Nando’s contract in the future and that’s it for ROO. ROO immediately loses 10% of sales. I bet this will happen in the future as JET is the only vendor capable of offering 100% coverage in the U.K. versus ROO at 50%.
Think of it this way. Why don’t ROO just shut down it’s business and offer white label to all its chains? Isn’t that great!! Nope, because JET can just win that white label contract in the future. There is no moat and the core ROO network is broken
Could you share your maths on adjusting deliveroo GP 30-40% lower under the DASH accounting approach? I agree that their GP is inflated. But those "admin expenses" also include the army of footsoldiers who go and pitch to restaurants to sign them to the platform. That process is very manual/janky in Yr1, but in theory should be somewhat non-recurring + it scales as orders per XYZ increases.
Also anecdotally the back-end of Deliveroo is a bit of a shitshow, plenty of human beings typing things on spreadsheets. So whilst I agree those costs are variable for now, in 5yrs time they should be able to automate and therefore "steady state" business model could gain more leverage.
Art not science and I agree some adjustment is needed. But 30-40% sounded v punitive to capitalise to perpetuity so just interested to understand working (not saying it's wrong)
Those costs are recurring. It is not customer acquisition cost. It is the cost of onboarding all those deliveroo couriers (100,000 of them), that churns every 6 months. So to keep the courier base stable, DROO needse to acquire 200,000 couriers every year. This process takes A LOT of manpower to process these couriers. DROO got to make sure they are not rapists, pedophiles, have the necessary paperwork, have a suitable vehicle, etc.
Then you have the management part, what happens when the couriers can't find the customers? What happens if the restaurant kept them waiting for too long and that affected them (ratings, earnings power, etc.)
That is what those 1,800 employers are for. Those are in ops and admin, not S&M.
There is also customer service. ALl that time where your food arrives late and your kids are starving, wrong items, missing items. There are people somewhere in the world helping with that process. Suppose 1/10 of orders go wrong, thats about 18mn ordesr a year that you have to make sure is made right. You need to hire hundreds of people to do that.
Then there is restaurant management (not onboarding). It is the manage them. Making sure they get paid on time, making sure their items are properly updated on the website.
All of these have to be accounted for in the unit economics. When you compare to DASH, DASH accounts for ALL of that in COGS, but DROO doesn't (it is part of staff and overhead cost).
You are right that those admin cost include growth expense, but I have already removed those from the estimate. So DROO gave us admin cost and also marketing cost % of GTV for 2020. DROO also gave us admin cost in the corporate HQ (I assume all of that belongs to staff and overhead).
So we have a good idea of staff and overhead cost on the country level. I take 75% of that cost as the cost to maintain the operation. Because #1: ops and admin is 90% of the ops and admin + S&m employee base. #2: the company fired so many restaurant acquisition people in 2020 so I bet the vast majority of the cost is recurring cost (effectively COGS).
The other way you can do it is to back into it based on commentary from other people. You can triangulate from TKWY data. You can also compare DROO staff and admin cost to DASH and GRUB R&D +G&A % of GTV. They all end up at about 3% of GTV, which is really only 60c/order. But 60c/order x 180m order = >100mn in cost.
You can also think of things this way, the average courier does about 1,000 ordesr before churning. It probably costs $500-$1000 to recuirt, onboard, and maintain that courier. Hence, recruiting and maintaining that courier is already 50c - $1.
Another way to think about it is, the average customer service rep probably goes through about 6 issues per hour (I am throwing this number out, but I bet issues are seen once every ten orders). That customer service rep costs 12pounds easily, so its costs 2 pounds per issue, or 20c/order in customer service.
There are a lot of ways you can triangulate it. But it all comes down to over 100mn in cost that is not accurately reflected in COGS.
Oh. Also, I think I read somewhere, maybe it was from Just Eat or an expert network call that I did back in the day. It probably costs $1 in overhead in courier management and customer service (excluding growth). So i think my estimates might be too conservative actually.
Thanks a lot - very interesting! Defs agree the driver recruitment/churn and customer service is a fully variable cost currently. I think the bull case for all these names is that more of it can be automated over time (lots of cities are still in "venture" / ramp-up mode which I bet internally is just guys scrambling and taking each day as it comes, but should be able to systematise over time). But likely applies to everyone equally so agree DROO should adjust. You mention some great ways of approaching it.
Ties back to some back-of-envolope I was doing on TKWY. Seems once logistics gets bedded in (2022/23) they could hope to get back to 10% "dirty" gross profit per GMV - 50/50 mix of $3 per marketplace order and €1 per logistic if they charge €2 delivery fee (if lower can spend less marketing below the line). Knock off your 4-5% variable support costs gets you 5-6%, then another 1-2% on marketing (either via adverts or cheaper pricing above the line) and you get potentially quite a sustainable 4% "fully loaded" minimum which is v hard for competitors to attack and probably has some fat in it for growth.
Stick €500m of head office costs and pick your GMV growth. At €30bn GMV in a few years thats a €700m EBIT biz against current EV c€10bn ex Brazil.
Also, corporate overhead doubled in 2020 despite then firing half their “director and global management” team and shrinking the other divisions. I wonder how much segment level costs were stuff in the corporate line item to make it seem like the business is profitable on a segment/country level.
Really enjoy the blog. Couple key questions:
1. Why can't Deliveroo or Uber Eats just go out and capture the marketplace restaurants? It seems that TKWY is relying on these marketplace restaurants in the UK, but what prevents Roo to do in marketplace what TKWY is doing in logistics? In fact, it costs much more money for TKWY to add a 1P restaurantwhere they are handling logistics vs. Roo adding a 3P restaurant where they just put them on the platform. If you are a marketplace restaurant, it costs you nothing to be added to another app and even if you capture 1-2% more sales, it is already a win.
2. I see your point on white label, but at least you are capturing some demand. For example, if there 100 customers interest in Nandos day 1 and Roo captures 100 of them. In day 2, Nando incentivizes them to go straight to Nandos and Roo captures 50 of them. So yes Roo lost 50 day 2, but guess what - TKWY lost all 100 day 1 when Nandos is no longer on their app. So overall, it seems this movement is negative for the entire space, but Roo is still better positioned v TKWY if they have Nandos than if they do not.
it is a lot harder to capture marketplace restaurants than it seems. There hasn’t been a single country where a logistics player have built a sizable marketplace offering against an incumbent marketplace vendor. Deliveroo has tried for many years to no avail. I am not going to do the work for you, but ask yourself why hasn’t Expedia or Carguru had any success in Europe? Cant they offer 0% take rate as well?
On white label, you are right. Except that Nando’s is not on the JET platform at all. And guess what, TKWY can easily win the Nando’s contract in the future and that’s it for ROO. ROO immediately loses 10% of sales. I bet this will happen in the future as JET is the only vendor capable of offering 100% coverage in the U.K. versus ROO at 50%.
ROO is effectively burning its network effect, the only thing that builds long term valuable and is a moat, to be a commodity contractor.
Think of it this way. Why don’t ROO just shut down it’s business and offer white label to all its chains? Isn’t that great!! Nope, because JET can just win that white label contract in the future. There is no moat and the core ROO network is broken
Could you share your maths on adjusting deliveroo GP 30-40% lower under the DASH accounting approach? I agree that their GP is inflated. But those "admin expenses" also include the army of footsoldiers who go and pitch to restaurants to sign them to the platform. That process is very manual/janky in Yr1, but in theory should be somewhat non-recurring + it scales as orders per XYZ increases.
Also anecdotally the back-end of Deliveroo is a bit of a shitshow, plenty of human beings typing things on spreadsheets. So whilst I agree those costs are variable for now, in 5yrs time they should be able to automate and therefore "steady state" business model could gain more leverage.
Art not science and I agree some adjustment is needed. But 30-40% sounded v punitive to capitalise to perpetuity so just interested to understand working (not saying it's wrong)
Hey there. Thanks for reading.
Those costs are recurring. It is not customer acquisition cost. It is the cost of onboarding all those deliveroo couriers (100,000 of them), that churns every 6 months. So to keep the courier base stable, DROO needse to acquire 200,000 couriers every year. This process takes A LOT of manpower to process these couriers. DROO got to make sure they are not rapists, pedophiles, have the necessary paperwork, have a suitable vehicle, etc.
Then you have the management part, what happens when the couriers can't find the customers? What happens if the restaurant kept them waiting for too long and that affected them (ratings, earnings power, etc.)
That is what those 1,800 employers are for. Those are in ops and admin, not S&M.
There is also customer service. ALl that time where your food arrives late and your kids are starving, wrong items, missing items. There are people somewhere in the world helping with that process. Suppose 1/10 of orders go wrong, thats about 18mn ordesr a year that you have to make sure is made right. You need to hire hundreds of people to do that.
Then there is restaurant management (not onboarding). It is the manage them. Making sure they get paid on time, making sure their items are properly updated on the website.
All of these have to be accounted for in the unit economics. When you compare to DASH, DASH accounts for ALL of that in COGS, but DROO doesn't (it is part of staff and overhead cost).
You are right that those admin cost include growth expense, but I have already removed those from the estimate. So DROO gave us admin cost and also marketing cost % of GTV for 2020. DROO also gave us admin cost in the corporate HQ (I assume all of that belongs to staff and overhead).
So we have a good idea of staff and overhead cost on the country level. I take 75% of that cost as the cost to maintain the operation. Because #1: ops and admin is 90% of the ops and admin + S&m employee base. #2: the company fired so many restaurant acquisition people in 2020 so I bet the vast majority of the cost is recurring cost (effectively COGS).
The other way you can do it is to back into it based on commentary from other people. You can triangulate from TKWY data. You can also compare DROO staff and admin cost to DASH and GRUB R&D +G&A % of GTV. They all end up at about 3% of GTV, which is really only 60c/order. But 60c/order x 180m order = >100mn in cost.
You can also think of things this way, the average courier does about 1,000 ordesr before churning. It probably costs $500-$1000 to recuirt, onboard, and maintain that courier. Hence, recruiting and maintaining that courier is already 50c - $1.
Another way to think about it is, the average customer service rep probably goes through about 6 issues per hour (I am throwing this number out, but I bet issues are seen once every ten orders). That customer service rep costs 12pounds easily, so its costs 2 pounds per issue, or 20c/order in customer service.
There are a lot of ways you can triangulate it. But it all comes down to over 100mn in cost that is not accurately reflected in COGS.
Hope it helps!
P.S. for the typo errors. Too much wine by now. Cheers mate.
Oh. Also, I think I read somewhere, maybe it was from Just Eat or an expert network call that I did back in the day. It probably costs $1 in overhead in courier management and customer service (excluding growth). So i think my estimates might be too conservative actually.
Thanks a lot - very interesting! Defs agree the driver recruitment/churn and customer service is a fully variable cost currently. I think the bull case for all these names is that more of it can be automated over time (lots of cities are still in "venture" / ramp-up mode which I bet internally is just guys scrambling and taking each day as it comes, but should be able to systematise over time). But likely applies to everyone equally so agree DROO should adjust. You mention some great ways of approaching it.
Ties back to some back-of-envolope I was doing on TKWY. Seems once logistics gets bedded in (2022/23) they could hope to get back to 10% "dirty" gross profit per GMV - 50/50 mix of $3 per marketplace order and €1 per logistic if they charge €2 delivery fee (if lower can spend less marketing below the line). Knock off your 4-5% variable support costs gets you 5-6%, then another 1-2% on marketing (either via adverts or cheaper pricing above the line) and you get potentially quite a sustainable 4% "fully loaded" minimum which is v hard for competitors to attack and probably has some fat in it for growth.
Stick €500m of head office costs and pick your GMV growth. At €30bn GMV in a few years thats a €700m EBIT biz against current EV c€10bn ex Brazil.
Sound reasonable?
i won’t comment on $TKWY now but i think your numbers are far too conservative.
nw appreciate the discussion. kudos for a great blog