Shared Scooters Don't Last Long (substack.com)

Alison Griswold, writes on her newsletter Oversharing: I took a look at data on scooter rides in Louisville, Kentucky, shared online as part of the city's open data policy. The latest data is available here. The data set I used was older and included monthly data on scooter trips from August through December. It also included a unique "ID" for each scooter, a detail that was key to my analysis and has been stripped out of subsequent data sets published by Louisville. The data doesn't differentiate between Bird and Lime, but as Bird started operations in August 2018 and Lime that November, you can assume it skews toward Bird.

With that preamble, here are some things I found: The average lifespan of a scooter in Louisville from August to December was 28 days. Median lifespan was 23 days. If you stripped out scooter IDs that first appeared in December, to focus on older vehicles, the average lifespan increased slightly to 32 days and the median lifespan to 28 days. Still stripping out scooter IDs that started in December, the median scooter took 70 trips over 85 miles.

Scooter lifespan is a key factor in scooter unit economics, as you may recall. The more trips and miles a single scooter can cover, the better for shared scooter companies, which have to recoup the cost of each vehicle before they can start making any money. In October, The Information reported that Bird was spending $551 per scooter with a goal of reducing that cost to $360. At the time, I said that meant Bird needed five rides a day on a $551 scooter for 5.25 months just to recoup the initial cost. The picture painted by the Louisville data is even worse.

[...] So, our scooter company walks away with $2.32 in revenue per day from the average scooter in Louisville. As we said at the beginning, Louisville data indicates that the average scooter was around for between 28 and 32 days. That means the typical scooter generated something like $65 to $75 in revenue for the company after most operating costs over its lifetime.

Comments

Deep 6

[PuddleBoy]

They did a trial of the scooters in Portland (which has a river running thru downtown) and there is speculation about just how many of the scooters ended up at the bottom of the river. They were able to trace a few of them to that watery grave.

The biggest concern here was riders without experience and who did not use helmets. There was an uptick in ER visits for scooter accidents.

Re:Deep 6

[Anonymous Coward]

Each day in Austin, 20+ people on scooters wind up in the ER.

As for underwater visits, if you passed Waller Creek near Red River in Austin, you would see at least several Lime or Bird scooters a day in the drink, especially around UT, where people just yo-heave-ho them as a way to silence the constant noise they make when tipped over.

Scooters have a bad rep. Mainly because people use them in buildings, park them behind cars. The people who charge them in the morning sling them on private property, on handicapped ramps, around people's cars. In Austin, it was told to local businesses that if there are scooters, the best option is to pick them up and sling them into the nearest dumpster, rather than waste time impounding them.

A lesson on socialism

[Anonymous Coward]

Ownership matters.

Have trouble believing it's really that short

[SuperKendall]

the average lifespan increased slightly to 32 days and the median lifespan to 28 days.

I can believe the scooters would last that long before being pulled for servicing.

But I can't believe scooters after a month are so trashed you cannot repair them and get them back out in the field. Even with rough use and vandalism, you should be able to have the units in service for at least half a year...

That doesn't account for outright theft but I don't think so many are taken outright is affects the overall stats.

Re:i bet landfills will be filled

[jbmartin6]

Yes, just like the piles of abandoned rental bikes in China. Well it will be a convenient source of metals for future generations, unless they get classified as historical artifacts.

Re:A lifespan of only 23 to 32 days?!

[Fly Swatter]

That is what happens when you use a consumer product for commercial purposes. If you want something that lasts, only buy products that retain their warranty under commercial use.

Longtime database designer here.

[hey!]

These scooters likely get stripped down to parts, which are completely interchangeable, so there's no point in tracking individual parts that "go together". This leaves you with the Sacred Galley of Athens question -- is it the same boat after you've replaced every single part? If you completely disassemble a dozen schooters and reassemble a dozen scooters from randomly selected parts, what happens to the "identity" of the scooters that were taken apart? The answer is you don't need it anymore.

If it were firearms, we associate the identity of the firearm with the receiver -- the metal housing into which the barrel and moving parts of the firearm are assembled. But that's purely conventional; you could just as reasonably define the identity of the firearm by the barrel. But why even have a concept for the "identity" of a firearm? Really one only: to track ownership and custody of a firearm, you have to have some kind of database. Databases require identifiers. Seventeenth century gunsmiths didn't stamp serial numbers on their guns because nobody was tracking them.

You could take the same approach as firearms to scooters by declaring that the identity of a scooter sticks to, say, the scooter's deck. But what *function* would that serve? The function of a rental scooter's id is to track user custody of company property and determine when a scooter needs to be serviced. Once the scooter is brought in for repair the need to track that ID disappears. If you insisted on having an id that persists through the rebuild process it would do something that only bad database designs do: constrain physical operations to serve the record keeping system.

In my experience every database design can be invalidated by expanding the universe of questions it must answer (or equivalently, processes it must support). This is the problem with identity in the relational model; it's *implicitly* tied to the questions the designer anticipates. That's why UUIDs are such a robust solution to many identifying tasks: their uniqueness is not tied to any particular set of questions you might want to answer, or to any context (i.e., they are unique *between* databases).