Supercharger congestion - a modest proposal

[sorka]

What if I live in a loft apartment that shares the top floors of a building that is also an office building? What if I work the night shift and my location looks reversed? How will the systems automatically determine this without it being a customer relationship liability to Tesla and an additional cost factor in maintaining a team that goes to verify this stuff?

None of those conditions apply to an MS that is parked in a driveway or a garage of a residential address. They could exclude the hard to tell cases when a house is next to or even near an office building. My car not only knows that I'm on my own property but it has me within 3 feet of where I am in my garage without fail. It's never wrong.

There are all kinds of corner cases that Tesla can exclude from automatic abuse tagging but 50% of them will be the guy that's supercharging and then parking at home on a full tank all night....night after night. The other 50% won't be so obvious or will be in multi unit dwellings, or will be the guy who lives in a residential loft downtown and is not parked in a driveway or garage of a residential address. Tesla can let those slide.

 

[ohmman]

What if I live in a loft apartment that shares the top floors of a building that is also an office building? What if I work the night shift and my location looks reversed? How will the systems automatically determine this without it being a customer relationship liability to Tesla and an additional cost factor in maintaining a team that goes to verify this stuff?

Those are all valid questions, but I don't think it's as hard as you're making it seem. Once you have a model, you can tweak your cutoffs to limit false positives. For example (and this is kind of the inverse, but you'll get the idea), I met a guy from Nigeria at NIPS. Infant asphyxia is a leading cause of neonatal morbidity and mortality there. Because they don't have monitoring equipment in most of the hospitals (not to mention very limited electricity), they haven't had a way to deal with this issue. The guy I met received a grant to do work taking the sound of infants' cries and correlating them with their risk of asphyxia. He trained his model on tens of thousands of cases out of Mexico and was looking to apply them to Nigeria. The system ran as a native Android app, which worked very well because cell phones are ubiquitous in Nigeria. They were able to achieve 100% precision in identifying infants at risk of imminent asphyxia. They did so by permitting alerts on false positives as well - because in this case, unless every cry is a positive signal, even a relatively high level of false positives is a worthwhile cost.

My point is that Tesla could build a model (which would be much less complex than determining potential death due to vocal cries, btw), and set it up only to target cases that have a very high confidence level. Sure, they'd miss some true positives, but minimizing false positives would keep customers from being unfairly targeted.

 

[FlasherZ]

None of those conditions apply to an MS that is parked in a driveway or a garage of a residential address. They could exclude the hard to tell cases when a house is next to or even near an office building. My car not only knows that I'm on my own property but it has me within 3 feet of where I am in my garage without fail. It's never wrong.

There are all kinds of corner cases that Tesla can exclude from automatic abuse tagging but 50% of them will be the guy that's supercharging and then parking at home on a full tank all night....night after night. The other 50% won't be so obvious or will be in multi unit dwellings, or will be the guy who lives in a residential loft downtown and is not parked in a driveway or garage of a residential address. Tesla can let those slide.

But you're missing my point...

I'm asking how you programatically determine that it's "parked in a driveway" ? Do Tesla's servers pull Google maps and look for a driveway? Does Tesla have a database that knows which addresses are residential or business?

When you say that "Tesla knows I'm on my own property" - how does it know that you own your property and that it's not owned by your parents, or a trust,? And how do they attach it to you as the driver of the car? What if the car is owned in trust, as mine is?

I'm asking how Tesla automatically determines this. My point is that it's extremely difficult to come up with a plan that will not piss off a bunch of people. I hear you saying that they should just make some assumptions that if a car spends more than x% of nights at a single location, just assume it's their home -- but that doesn't work!

And just by way of example, in Tesla's first communication, they didn't get it right and pissed off a bunch of people -- just check out the thread about letters going to supercharger users. I know several in St. Louis who were wrongly tagged -- and they do charge at their single-family residences every single night.

 

[FlasherZ]

They did so by permitting alerts on false positives as well - because in this case, unless every cry is a positive signal, even a relatively high level of false positives is a worthwhile cost.

Ahh, and that's the big difference!

No parent is going to argue with intervention activity for a false positive when it comes to the life of his/her child, especially when the upside is so high and the downside is minimal. In addition, the problem you state is a simpler one (although more life-critical) -- a single factor (vocal cry) comes into play.

In the Tesla case, it involves a much greater complexity -- and my point is simple: how can Tesla reliably determine what is a person's home? Today, I suppose that we can consider a greater majority of Tesla customers today are people who have a daily job on a business-day basis and sleep in a single-family home -- it's easy to make assumptions then: if a car shows up in the same place for 60% or more of nights, then consider it a home. But fast-forward to Model 3... do more live in apartments? do more work the night shift? are homes co-mingled with business so that it's not understood?

I'm not arguing that they couldn't make those assumptions and be right even 90% of the time... I think they could, just on a very simple algorithm. But I'm arguing the impact from the 10% - 5% - or even 1% who would be wrongly accused of being an abuser is too much downside and that your ability to identify these abusers must be near-perfect.

For Tesla, it's 95% downside if they have a false positive, with the only upside being that "arriving at a supercharger gives you a greater chance to have a stall available for your use". That's nothing compared to the downside: the customer gets pissed off if they're being accused of being a local abuser; or worse yet, they can't charge because the system locked them out - can you imagine the headlines published when I can't get to my dad's hospital bedside in time because the supercharger miscalculated where my home was located and it disallowed me charging or clamped me down to L2 charging rates? My ICE wouldn't have that problem!

So, how can Tesla determine with nearly perfect precision (99.9%) where someone's "home" is, considering all these factors?

 

[ohmman]

My point was that you can do the inverse - you can tweak confidence levels so that you're only alerting on the ones for which you're very confident. I can imagine a model that only identifies one guy who is obviously abusing the system. It wouldn't generalize well, but I can imagine it, and it would have 100% precision. So start backing that away and reach a model that generalizes.

Also, just to be clear - a vocal cry is not a single factor (though it is to us). To a model, it's a very complicated feature set. Without looking at it, I can't say, but I wouldn't be surprised if it were a larger feature set than the set of factors going into a Supercharger abuse model.

 

[sorka]

I'm asking how you programatically determine that it's "parked in a driveway" ? Do Tesla's servers pull Google maps and look for a driveway? Does Tesla have a database that knows which addresses are residential or business?

This data exists widely in both private and public form. It's part of the census records. It's part of the public property records that countless companies scrape to provide big data to companies like Zillow, real estate, insurance, utility, and marketing companies.

Heck, I can enter a GPS coordinate into google maps, pull an adddress from that and then enter it into Zillow and get the dwelling type, square footage, etc.

 

[FlasherZ]

This data exists widely in both private and public form. It's part of the census records. It's part of the public property records that countless companies scrape to provide big data to companies like Zillow, real estate, insurance, utility, and marketing companies.

Heck, I can enter a GPS coordinate into google maps, pull an adddress from that and then enter it into Zillow and get the dwelling type, square footage, etc.

So I tried that on a few addresses that I know of...

First my home: It appears to be pulling data from a database of county property tax record cards (like almost every other public/private database), because I recognized the same errors in their listing that I had to correct about 8 years ago with the county (so the data is woefully out of date). That said, it doesn't reflect that this is also the address of a business.

I then gave Google Maps GPS coordinates for my sister's carport as measured by my phone's GPS receiver (which shows up properly on the map, so there's no sense error). Google Maps identified the address as the school across the street - and, unfortunately, no record in Zillow.

It did successfully find my brother's home address and classified it properly as a home. It doesn't seem to know whether there's a business operating there, though.

It found my parents' home address but didn't know of the business operating from there.

It did not have a record for our hunting lodge, where I park overnight every other weekend or so.

Finally, I tried a case that would be fairly interesting - it's a professional services business that used to be a home but was converted about 10 years ago or so, and sits in a residential area. Zillow had no record of the address at all, much less the ability to tell me home/business differences.

My home - Address Found Properly? Y Identified Home? Y Identified Business? N
My brother's home - Y / Y / NA
My sister's home - N / N / N
My parents' home - Y / Y / N
Hunting lodge - N / N / N
Tax business - N / N / N

Unfortunately, it's not looking good for a 99.9% rate, or even a 90% rate.

 

[sorka]

First my home: It appears to be pulling data from a database of county property tax record cards (like almost every other public/private database), because I recognized the same errors in their listing that I had to correct about 8 years ago with the county (so the data is woefully out of date). That said, it doesn't reflect that this is also the address of a business.

If's it an actual residential address that you're parking your MS at overnight every night or most night, then it doesn't matter if there's a business also be run from there.

I then gave Google Maps GPS coordinates for my sister's carport as measured by my phone's GPS receiver (which shows up properly on the map, so there's no sense error). Google Maps identified the address as the school across the street - and, unfortunately, no record in Zillow.

I've used google maps hundreds of times to right click "what's here" on houses and business's and it's NEVER been wrong. Not once. Perhaps you're phone GPS is not as accurate as you think.

It did successfully find my brother's home address and classified it properly as a home. It doesn't seem to know whether there's a business operating there, though.

That's what I would expect.

It found my parents' home address but didn't know of the business operating from there.

Same answer as your home above. If you supercharge at superchargers and then park at home every night where you could charge but choose not to, do you think you should be excluded because you also run a business from there?

It did not have a record for our hunting lodge, where I park overnight every other weekend or so.

And what's the problem with this? You would be tagged as an abuser by Tesla even if this was your place of residence. It would be a false negative, not positive.

I checked over 3 dozen friends and families places of residences in California, Washington, Arizona, and Missouri. Google maps identified the correct address when I found it by by arial and clicked "what's here" and Zillow correctly identified the dwelling type for each and every one of them. So I see 100%. It's fine if the system could only exclude 50% of the abusers do to false negatives but excluding 50% of the abusers would go a long ways.

 

[ToddRLockwood]

Last fall, I spoke at length with a member of the supercharger team about this. The warning emails are not generated by an automated system. There are human beings involved, and as I mentioned previously, there is not a specific formula used. Apparently, frequency is a bigger factor than geography. In other words, if you charge now and again at a supercharger near your home, it's not likely to trigger an email. But if Tesla notices that your home charging has dropped to zero and you're using the same supercharger day in and day out, that's more of a problem.

Say you have a daily 120-mile commute, and you stop at the same supercharger every day along the route. As long as you are also plugging in at home in the evening, I don't believe Tesla would consider this a problem.

 

[FlasherZ]

Last fall, I spoke at length with a member of the supercharger team about this. The warning emails are not generated by an automated system. There are human beings involved, and as I mentioned previously, there is not a specific formula used. Apparently, frequency is a bigger factor than geography. In other words, if you charge now and again at a supercharger near your home, it's not likely to trigger an email. But if Tesla notices that your home charging has dropped to zero and you're using the same supercharger day in and day out, that's more of a problem.

Say you have a daily 120-mile commute, and you stop at the same supercharger every day along the route. As long as you are also plugging in at home in the evening, I don't believe Tesla would consider this a problem.

Then they were *really* bad at what they did. A local St. Louis couple got an e-mail after only THREE uses of the supercharger in their car over a few months - and they had been charging nightly at home.

 

[FlasherZ]

I pointed out some cases where your solution wasn't working, and you say that it's my fault somehow.

My best to you.

 

[TaoJones]

Sigh.

Once again, with emphasis upon the last paragraph: What wasn't said.

Further, note that Tesla has committed to DENSITY as well as to DISTANCE for almost two years now.

 

[Ben W]

I think that by far the most effective solution to all this would be to implement per-minute charging for the Model 3. (Say $0.25/min). This is still cheaper than gas for roadtrips, while encouraging proper use: not supercharging beyond what you need, particularly not trickle-charging above 90% unless you really need it, and certainly not staying plugged in after charging is finished.

If Tesla wanted to further incentivize long distance vs local supercharging, they could waive the fee if the car were driven 100mi+ within a 24h window of supercharging. (So if I first drive 100mi and then plug in at a SC at my destination to recharge, that's also ok.) Staying plugged in after charging would of course always be billed. This system would make most "legitimate" uses of supercharging free, while penalizing abuse in a firm but not onerous way, and it properly handles nearly all the cases I've thought of. Can you think of a counterexample? (Commercial uses such as taxis exempt of course.)

If selective adjustment/waiving of fees were considered, the other advantage of basing it purely on mileage rather than location is that it is much less invasive: Tesla doesn't need to know where you are or where you live for the system to work, just how many times the wheels have spun.

 

[WarpedOne]

Complications to fix issues because of complications.
Yes, that surely works.
There is one simple and right solution: build more superchargers.
To costly? Charge bigger upfront payment.
Is upfront payment getting to high? Change it into annual package.

Just don't go the beancounters way.

 

[Ben W]

Getting billed at home for the number of gallons of water you use? That's too complicated. It should be an upfront flat fee for as much water as you can suck out of the pipes. Upfront payment too high? Make it an annual package.

See the problem? Yeah, me too.

Not just trying to be snarky here. The overuse of the resource will have to be baked into the cost of each and every SC-enabled Model 3, which in a price-conscious marketplace could wind up hurting Tesla quite a lot. For the Model 3 in particular, Tesla has every reason to keep supercharging access as economical as possible, and all-you-can-eat access is quite possibly the worst way to do that.

 

[EVie'sDad]

Put your mind at ease...

"FREE FOR LIFE IS SUSTAINABLE"
Submitted by SamO on April 14, 2016 (source "FREE FOR LIFE IS SUSTAINABLE" | Tesla Motors)

Tesla Vice-President of Business Development Diarmuid O’Connell was in Amsterdam yesterday for the AVERE E-mobility Conference. The long-time Tesla executive, and one of the company’s first 50 employees, gave a short presentation before doing a Q&A with the audience.

A few nuggets:

Supercharging - Free Forever is sustainable.
Tesla wants to provide off the shelf solutions to other OEMs for batteries and other EV parts.
Tesla has a 3-3.5 year product cycle, about half the industry average.
Evolution of Net Metering going to make storage more viable and interesting.
V2G depends on development at the utilities. Currently at "smart charging" but eventually to arbitrage.
Tesla product line increase is coming beyond Model 3 including a truck.
Great story is shaping up in the China Destination Charging program.
Approaching 400,000 people have put down their reservation.

Video link in excellent article by Electrek.co

Tesla Vice President says Model 3 reservations are ‘approaching 400,000’, real success will be delivery

 

[Ben W]

Put your mind at ease...

Tesla Vice-President of Business Development Diarmuid O’Connell was in Amsterdam yesterday for the AVERE E-mobility Conference. The long-time Tesla executive, and one of the company’s first 50 employees, gave a short presentation before doing a Q&A with the audience.

A few nuggets:

Supercharging - Free Forever is sustainable.

Well, let's take a look at what Diarmuid O'Connell actually said (and thank you for the link btw, fascinating video):

Q: Tesla understood very well that the success of the electric car lies in building the charging infrastructure. How can your supercharger network be a sustainable business model for the long run?

A: As someone once said, in the long run we’re all dead. [laughter]

The supercharging network was set up to address a very practical consideration, which is how do we make the Model S a complete solution, a complete replacement for the other vehicles in the segment: the Mercedes S-Class, the Audi A-8, for which there is already an installed base of gasoline and petrol filling stations. It started with that proposition, and at the scale we introduced it, the scale we’re operating right now, the free-for-forever proposition is sustainable.

We’re pragmatic and as we look to the future, I could imagine our charging network and other charging networks evolving in a number of different ways. The other reason why we put the supercharging network out there is that no one else was doing anything like that, especially in the US, maybe less in the Netherlands where you had some early investments in charging infrastructure, no one was doing anything that made sense. There were a lot of government investments for charging stations that went in front of town halls because that was convenient and graphic for a ribbon cutting ceremony but it had nothing to do with how people would actually want to use a charger, and it had nothing to do with the kind of charging people wanted, which is point-to-point, long-distance charging and quick charging. So we’ll see what the future holds, but for right now we’re holding to our model.

So to me, this reads that for the time being, within the luxury high-profit Model S/X segment, the free-for-life model is sustainable (which it is). He is certainly not promising or even hinting that this will remain the case for the Model 3; in fact my take is that he's actually strongly hinting the opposite. So I stand by my assessment that Tesla is overwhelmingly likely to change the SC pricing structure for the Model 3 to pay-per-use. They may initially try to keep it unlimited for the loaded high-profit configurations (as they originally did with the Model S 85 vs 60), but I would be quite surprised if they offer it unlimited with the base configuration, even as a pay-upfront option.

 

[ToddRLockwood]

Then they were *really* bad at what they did. A local St. Louis couple got an e-mail after only THREE uses of the supercharger in their car over a few months - and they had been charging nightly at home.

Sounds like somebody messed up, or there's a glitch in the system that identifies possible abusers.

As an early Model S owner, I can attest that Tesla understated the Supercharger usage rules when the charging stations first appeared. Perhaps they hadn't recognized the problem yet. Their "superchargers are for long distance travel" wording came later.

 

[ToddRLockwood]

I think that by far the most effective solution to all this would be to implement per-minute charging for the Model 3. (Say $0.25/min). This is still cheaper than gas for roadtrips, while encouraging proper use: not supercharging beyond what you need, particularly not trickle-charging above 90% unless you really need it, and certainly not staying plugged in after charging is finished.

If Tesla wanted to further incentivize long distance vs local supercharging, they could waive the fee if the car were driven 100mi+ within a 24h window of supercharging. (So if I first drive 100mi and then plug in at a SC at my destination to recharge, that's also ok.) Staying plugged in after charging would of course always be billed. This system would make most "legitimate" uses of supercharging free, while penalizing abuse in a firm but not onerous way, and it properly handles nearly all the cases I've thought of. Can you think of a counterexample? (Commercial uses such as taxis exempt of course.)

If selective adjustment/waiving of fees were considered, the other advantage of basing it purely on mileage rather than location is that it is much less invasive: Tesla doesn't need to know where you are or where you live for the system to work, just how many times the wheels have spun.

On a technical note, it would make more sense to charge based on energy delivered, not the charging time, since the charging rate is not always the same. This is how ChargePoint operates.

Tesla spent a good deal of time studying driving habits of both electric and gasoline car owners before deciding on the pay-upfront-free-charging approach they currently utilize. They continue to gather data from the current vehicle fleet. What they have found is that the average Model S owner does the vast majority of their charging at home at night—somewhere north of 95%, I suspect. The Model S has a roughly $2,000 surcharge built into the cost of the vehicle which is used to pay for the superchargers, both the infrastructure and the energy. For a while, this fee was an option on the entry level Model S. Now it's just part of the vehicle price.

No question that the Model 3 will present some new challenges in terms of volume of vehicles, and the increased likelihood that the owner resides in rental housing where charging is not readily available. But I don't believe that charging for energy at the superchargers is the answer. Instead, I would prefer a more specific policy on supercharger usage.

 

[FlasherZ]

Sounds like somebody messed up, or there's a glitch in the system that identifies possible abusers.

As an early Model S owner, I can attest that Tesla understated the Supercharger usage rules when the charging stations first appeared. Perhaps they hadn't recognized the problem yet. Their "superchargers are for long distance travel" wording came later.

I'm one of the Sig folks too... I always understood Supercharging to be for long distance travel, but I also agree with you that they didn't quite expect the power of "free". They didn't expect the amount of local freeloading that they ended up getting -- and that's understandable from an economic point of view. Someone making $250/hour in the bay area going 5-10 minutes out of their way to drive to the Supercharger and sit for 30 minutes, all to avoid $15 in electricity costs seems insane; but it's well documented that people will distort reality for the word "free".