Strategy of hub / nest and service area placement


#1

Over the past months, I’ve analyzed the location patterns of Hive and Tier scooters here in Vienna. Also, there will be rules dividing the 1500 scooters allowed in Vienna per company into three zones (downtown, inner and outer disctricts) with 500 scooters per company allowed in each zone. Vienna’s deputy mayor, on presenting the new rules, said that those rules are being put in place because there’s too many scooters downtown while the outer districts are “undersupplied”. However, I don’t think the new regulation will do much to better serve those outer districts. Thus I’ve come up with a strategy how to better develop those outer areas…

I’ve seen some companies put out scooters in one spot in large quantities (sometimes up to 12), and in areas with lower demand they still put out the scooters in groups, but with the groups being spaced further apart, or they don’t deploy any scooters in some locations at all because it doesn’t pay off to put a group of scooters there. I think this is a slightly wrong strategy…

I think an area is best supplied with scooters if, for as big a part of the area as possible, one can find a scooter in a reasonable distance… one scooter, not 4 or 5. Therefore the scooters should be spread across the area as best as possible so there’s only a group of scooters where most likely all of them will be taken within a day. The number of locations where scooters are brought out, due to my calculation, should be about half the fleet size, so that a fleet of 350 scooters would be spread across 175 locations. And about half of those locations (those with lower demand) should contain only one scooter. I know this is more work for juicers and chargers (and also for employees doing the same thing), but I think this would result in more rides being taken.

Now where to place those? Some companies say in places with highest demand, which is measured as the number of rides started there. But this number is skewed if in some areas there are next to no scooters, so no scooters can be used because people can’t find any. Therefore I’d look at each scooter individually and measure how long it stayed in place before it was taken for a ride again. This is not so easy because sometimes scooters are being collected before they get the next ride, so the collection of course doesn’t count as a ride. Now each time a scooter gets ridden again (or for the first time after being deployed) after less than 24 hours, this would be a potential place for a scooter placement, unless it’s proved that this was just a fluke. For example, if a rider has placed a scooter at a place, and it is ridden again after 2 hours (and there’s no history for this place in the database), this place becomes a 1-spot hub the next day, and if no scooter is deployed, it will continue being a 1-spot hub until another scooter gets there. For the next scooter placed there, it’s again measured how long it stayed before being ridden, and the times are then being added up. For instance, a scooter gets deployed, and it’s taken after 12 hours this time. This gives a total of 14 hours of scooter coverage and 2 scooters ridden, which is an average of one scooter every 7 hours. Since now there have been 2 scooters ridden, the place is promoted to a 2-spot hub, and the next day, 2 scooters become deployed there. Unfortunately, this time none of them gets ridden over 3 days, so now the place has had a scooter supply for 3 days and 12 hours, but only 2 scooters have been ridden, thus the average stay length now is 42 hours, which is longer than 24 hours, so now the place ceases to be a hub entirely. As soon as a place has had at least 24 hours of scooter coverage, it’s “proved”, that is, you can relatively reliably tell how much demand there is at this place. But of course there may still be flukes from time to time which then are being corrected by placing one scooter there and seeing how long it lasts.

This strategy should basically be followed for all places where scooters are being ridden to. For downtown areas, I think this would prevent places from being overserved by placing too many scooters there. For uptown areas, this way over time different places would be tried, and there are always SOME places which are currently being tried, so you can at least find some scooters on the rider map, which is better than none even if it doesn’t supply all areas at once.

In examining the data, I’ve found that the demand seems to be highest at major train and public transit stations, as well as along main roads and especially at the crossing between two main roads. Thus it might be a good idea to look where traffic lights are because they often are at intersections of two main roads. Not all of those intersections will turn out having high demand, however, but the chances are higher than at other places.

The second part of the strategy is how aggressively to move scooter that haven’t been ridden for a time. Some companies basically clear the outer areas of scooters every night and then only place them at the hubs. I think this is a bad strategy (unless the hubs are dense enough) at least for uptown areas because it regularly starves the area of scooters so that in the morning, riders can’t find any scooters to ride. I think Lime once had a strategy where only scooters not having been ridden for more than 24 hours were collected, which I think is better, and maybe you even have to go higher, but that depends on the market. With lowered bounties, some randomization comes in as well because some scooters now are being left on the map which would have been available for collecting. Unfortunately, this also affects the low battery scooters which should be collected because they can’t serve a rider in their current state.

This part can also be affected by the first part of the strategy. To continue with our example, on the day the two scooters have been deployed, at “pull time” (9 p.m.) the total coverage has gone up to 28 hours, so we have 2 scooters ridden in 28 hours which is 1.71 scooters per day. This is being rounded town to 1, and the spot turns back from a 2-spot to a 1-spot hub. However, 1.71 is still closer to 2 than it is to 1, so I would actually round up the numbers and only pull the 2nd scooter when the number drops below 1.51, which happens on the next day when we have 2 scooters ridden in 52 hours, which is 0.92 scooters per day. The 2nd scooter gets pulled when that number drops below 0.5, which would happen after 2 more days when we have 2 scooters ridden in 100 hours (over 4 days). Then the 2nd scooter gets pulled unless it has now been ridden. That way, at any “proved” places where demand is between 0.5 and 1 scooters per day, scooters typically wouldn’t get pulled if left there by a rider.

A third part of the strategy is where to place the service area. I’d follow a rather generous approach here and say that everywhere scooters are brought out, at least in groups of more than one, there should be about a mile of service area surrounding all of these places so that people have somewhere to go to on those scooters. The service area can still be softly defined by where the scooters are brought out and how aggressively they get collected in some areas because I’ve found that most of the scooters stay in an area of at most a mile around there they initially were being deployed. But that way you give riders the freedom to go where they want, and maybe the places where they go could become new hubs over time, or at least be included in the fabric of scooter supply. For areas which haven’t been tried yet, I would actually restrict the service area only when several places there have already been “proved”, and it has been shown that the average demand there (if you divide the total scooters ridden by the total coverage days in that area) is below 0.5 scooters a day, or at most below half of the average demand of the whole service area.

I know this is probably the wrong place to post this to because it’s rather advice for scooter companies than for chargers and juicers (who can’t choose where hubs and nests are), but I don’t know where a right place should be since I really haven’t seen many places on the Internet where topics like this are being discussed in public. But I know that Victor, the leader of this site, also operates Spring, a support company for scooter sharing companies, and maybe he will deem this topic important enough to do something with it. I think it is important because this strategy is a vital part of the business model of a scooter company and can decide if the company thrives or fails.