I actually think data is the problem. People were making their purchases online but the in store experience is what was convincing them to make the purchase in the first place. Hard to measure that kind of conversion.
There's this article about a Google engineer who reduced the size of the player from over 1MB to less than 100KB. But the data showed that page load times increased to 2 minutes instead of reducing.
They couldn't figure out why. They almost rolled back the change until someone figured out the reason. Millions of users with very slow internet couldn't even open YouTube before the change. Now, they could.
Tesla probably weren't tracking store visits, and test drives in relation to online purchases. And this missing info misled them.
Took over an hour of search (before ur request) - both Google and Bing were useless. Just crossed my mind that I may have upvoted it here on HN and voila!
Also, thanks for updating your original comment. Your numbers were stuck in my head. After reading the article - I was like - It's not making sense. So, I reread your comment and saw the updated numbers :-)
It's not super hard to get estimates like this, e.g. look at sales in comparable cities with or without stores, or how online sales changed after stores were opened. You won't get exact numbers but you should be able to get within a factor of 2, which is enough for most of these decisions. Companies like Warby Parker, etc. did this successfully at much smaller scales.
I disagree about it being within a factor of 2, at least for this year.
Consider the following:
* In January, the $7500 Federal Tax Credit became a $3750 Tax Credit for Tesla. The $3750 tax credit will be halved again some time this year.
* The price of the M3 has been in flux: the $35,000 car has been released, there were $2000 price drops (in reaction to the tax credit), and other issues.
* The M3 was a preorder vehicle: many people put down $1000 reservations, but its difficult to measure exactly which vehicle they wanted (a lot of them seemed to be waiting for the $35k version). Attributing the sales to preorders vs sales conversions is clearly a difficult problem.
* The M3 availability fluxuated grossly, based on Tesla's ability to produce it. You may be waiting 6+ weeks for a M3 if you ordered in 2018, but only 2 weeks if you ordered more recently. Simple knowledge of the decreasing wait time would spur many to purchase in of itself.
* Consumer Reports, and other major review magazines, released reviews which almost assuredly changed sales numbers.
The sales are just not stable enough on a month-to-month basis to do what you're suggesting.
They couldn't figure out why. They almost rolled back the change until someone figured out the reason. Millions of users with very slow internet couldn't even open YouTube before the change. Now, they could.
Tesla probably weren't tracking store visits, and test drives in relation to online purchases. And this missing info misled them.
http://blog.chriszacharias.com/page-weight-matters
Edit: Changed 20 minutes to 2 minutes. After rereading the article.