Audi e-tron GT

[Cujet]

My question is, why is their initial offering so bad in terms of range?
The mighty Porsche Audi team, with all it's storied history, comes out flat in the #1 EV measure.
They were promising 400 mile range and fast charging. Oops.
I don't get it.

Edmunds Tested: Electric Car Range and Consumption | Edmunds

Which EVs have the best electric car range? The EPA provides range estimates for every electric vehicle, but they don't test EV range in the real world -- and Edmunds does. Read all the latest real-world EV test results from the Edmunds experts.

www.edmunds.com

Some interesting test results. Every Tesla they tested fell short of projected range in the 60/40 driving cycle.
Interestingly, the Porsche Taycan turbo S beat the projected 203 mile range by 59%, achieving 323 miles in the test.
Even the Mustang Mach E beat it's advertised range of 270 and went 304 miles.

Maybe even more telling is the actual consumption to go 100 miles. The Porsche and the Tesla S performance both use 32.3 and 32.6 kwh respectively. The Porsche is a touch more efficient.

Mr. Musk is quite a salesman, that's for sure. I love his stuff, but there is clearly a difference between how manufacturers rate their EV range.

It's also good to note that, (I think) all the manufacturers recommend not fully charging the batteries to maximize battery lifespan.

 

[JeffKeryk]

Edmunds Tested: Electric Car Range and Consumption | Edmunds

Which EVs have the best electric car range? The EPA provides range estimates for every electric vehicle, but they don't test EV range in the real world -- and Edmunds does. Read all the latest real-world EV test results from the Edmunds experts.

www.edmunds.com

Some interesting test results. Every Tesla they tested fell short of projected range in the 60/40 driving cycle.
Interestingly, the Porsche Taycan turbo S beat the projected 203 mile range by 59%, achieving 323 miles in the test.
Even the Mustang Mach E beat it's advertised range of 270 and went 304 miles.

Maybe even more telling is the actual consumption to go 100 miles. The Porsche and the Tesla S performance both use 32.3 and 32.6 kwh respectively. The Porsche is a touch more efficient.

Mr. Musk is quite a salesman, that's for sure. I love his stuff, but there is clearly a difference between how manufacturers rate their EV range.

It's also good to note that, (I think) all the manufacturers recommend not fully charging the batteries to maximize battery lifespan.

Looking at the tests and results, one issue is ambient temperature. Better range results seem to favor higher termeratures.
One would guess the tests are not apples to apples. They certainly do not adhere to Scientific Method standards.

Edmunds summary: Which number is more accurate? "The short answer is neither."
The reason I have to agree is, the test methodology is all over the map. It's pretty much bogus due to the lack of strict standards.
At least the EPA uses a standardized methodology.

My point is, Porsche promised 400 mile range for their gorgeous Taycan in the early advertisements.
If Taycan range really is 59% better than the EPA range and close to their 400 mile forecast, I would think Porsche (and Audi) would have something to say about it. Range is a huge differentaitor in the EV world.

Just my 2 cents. All good, Cujet.

 

[Cujet]

Looking at the tests and results, one issue is ambient temperature. Better range results seem to favor higher temperature.
One would guess the tests are not apples to apples. They certainly do not adhere to Scientific Method standards.

I agree, non scientific testing is not generally all that valuable. However, a statistical model can be built over time, and 'regression towards the mean' (the more results we have, the more the results mirror an accurate middle, average or mean number) will provide the expected accuracy. In much the same way as a pool of prius drivers who submit their MPG to "fuely" or "fuel economy.gov" generate a good number. Eventually, the MPG number accurately reflects what a typical owner can expect.

We already see a little of that in the Edmunds data, with every Tesla falling short of the advertised range. I have to wonder if the cars they tested had 10 miles on them or 10,000. As battery performance follows a curve of slight improvement, followed by a slight decline.

 

[JeffKeryk]

I agree, non scientific testing is not generally all that valuable. However, a statistical model can be built over time, and 'regression towards the mean' (the more results we have, the more the results mirror an accurate middle, average or mean number) will provide the expected accuracy. In much the same way as a pool of prius drivers who submit their MPG to "fuely" or "fuel economy.gov" generate a good number. Eventually, the MPG number accurately reflects what a typical owner can expect.

We already see a little of that in the Edmunds data, with every Tesla falling short of the advertised range. I have to wonder if the cars they tested had 10 miles on them or 10,000. As battery performance follows a curve of slight improvement, followed by a slight decline.

Linear Regression is the most widely used statistical function in forecast models. I have found it is far more useful (accurate prediction) when data is broken down into meaningful groups, such as geographical regions, relevant time buckets and similar business units.
Also, regression models are improved over time as we observe what they tell us. Models have to be adjusted beyond the regression values as we better understand the data.
This is where the Edmunds results fall short; the data is skewed and does not even mention the number of observations.

Fun fact: I developed a regression model using Microsoft SQL Server against tens of millions of rows of sales data versus sales team tenure. I modified some college sql programming and coded a Linear Regression function in Transact SQL. Sometimes the forecast predictions were thrown out by executive staff as "Can't be right" only to be (unfortunately) accurate as the (poor) fiscal quarter sales numbers materialized. Then the exec's questions became far harder. Over about 2 years, the model evolved and was improved based on what we observed and learned.
The project was both scary, difficult and fun. Answering their questions with their steely eyes made me stutter my words... Luckily the model and assumptions used were clear and explainable. Of course the number one problem was, as always, the data. ERP data over time is garbage, but it's all we have. Hence, Data Science becomes necessary.