Tesla Reservations Model and Demand Generation

[SteveG3]

I think Morgan Stanley is backing into numbers that allow them to be bullish on the stock without sticking their necks out further than they need to.

In other words, if they have a buy on the stock, why speculate on 500,000 cars sold in 2025 when 150,000 backs into a target price today that warrants a buy recommendation (of course with 20% run past couple of weeks, they'll likely tweak their model after earnings call so they can raise target a few bucks and maintain buy).

Morgan Stanley stuck their neck out in the past. I'm looking at their 3/31/11 report where they had a $70 price target. In that report they had bear, base and bull scenarios modeled.

Here's total units sold (complete vehicles, excluding powertrains sold to other manufacturers) they had for 2025 in 3/31/11 report:

Bear: 106,000
Base: 454,000
Bull: 681,000

in another table in the 2011 report they show 370,000 of the base case units sold as Gen3.

I don't have the current report, but I strongly suspect that the current report calling for 150,000 Gen3 in 2025 Mulder has shared with us is driven by their wanting to back into a target price of about $50 rather than $100 (neighborhood their model would project with 370K gen3 assumption) , and does not reflect their genuinely believing the 150K 2025 sales figure, down over 50% from the 370K they modeled as a base case two years ago. I.E., though they've moved their price target down and back up in past two years, I strongly suspect price target they are comfortable publishing is the tail wagging the dog of what they publish as their sales model. Fwiw, IF Gen3 is executed more or less as well as Model S, I would think we see 150,000 units by 2020 (if not sooner).

Not total, but 150k annual deliveries by 2025. This is how their model is estimating the ramp-up for Gen 3:

2016 15,000
2017 22,500
2018 37,500
2019 67,500
2020 90,000
...
2025 150,000

 

[CyberDutchie]

I think Morgan Stanley is backing into numbers that allow them to be bullish on the stock without sticking their necks out further than they need to.

In other words, if they have a buy on the stock, why speculate on 500,000 cars sold in 2025 when 150,000 backs into a target price today that warrants a buy recommendation (of course with 20% run past couple of weeks, they'll likely tweak their model after earnings call so they can raise target a few bucks and maintain buy).

Morgan Stanley stuck their neck out in the past. I'm looking at their 3/31/11 report where they had a $70 price target. In that report they had bear, base and bull scenarios modeled.

Here's total units sold (complete vehicles, excluding powertrains sold to other manufacturers) they had for 2025 in 3/31/11 report:

Bear: 106,000
Base: 454,000
Bull: 681,000

in another table in the 2011 report they show 370,000 of the base case units sold as Gen3.

I don't have the current report, but I strongly suspect that the current report calling for 150,000 Gen3 in 2025 Mulder has shared with us is driven by their wanting to back into a target price of about $50 rather than $100 (neighborhood their model would project with 370K gen3 assumption) , and does not reflect their genuinely believing the 150K 2025 sales figure, down over 50% from the 370K they modeled as a base case two years ago. I.E., though they've moved their price target down and back up in past two years, I strongly suspect price target they are comfortable publishing is the tail wagging the dog of what they publish as their sales model. Fwiw, IF Gen3 is executed more or less as well as Model S, I would think we see 150,000 units by 2020 (if not sooner).

Actually, their total unit delivery estimate also includes the Tesla Roadster v2, albeit much lower unit numbers, but with higher revenue per unit ($131k-$154k). Here's what that model's estimated ramp looks like:

2017 800
2018 1,200
2019 1,320
2020 1,386
..
2025 1,381

 

[SteveG3]

Mulder, I'm not sure what the numbers in the chart in your last post represent.

 

[CapitalistOppressor]

Anecdotally, Cattledog recently spoke to an associate at one of the Texas stores who related that they were averaging 1-2 cars per day at that store, and that the number was far higher for stores in California.

My research has long pointed to an internal Tesla sales target of ~2 cars per store day for the new format stores. My model also assumes that new store openings will be the primary driver of sales growth going forward. The most simple model points to the existing 25 U.S. stores supporting ~18,000 sales per year, but many of those locations have only opened in the last couple of months, some are of the old format, and some are hampered by local regulations.

For stores that existed at the end of 2012 the model predicted they would generate ~11k U.S. sales for 2013, which was fairly close to the guidance that Tesla provided, and could have represented a conservative prediction based on then existing assets.

Its important to recognize that these are internal sales goals, and stores appear to get credit for all reservations within their sales area. In theory, new stores could just be cannibalizing "sales" that would have otherwise been assigned to another store. But I believe that the stores are the fundamental driver of sales, and I think Tesla believes that as well.

A key tell will be whether Tesla continues to expand stores. If demand stays level (or falls) they will cease expansion, because the sales/store will fall below their target. No need to invest in new stores if the ones you just built aren't generating additional sales.

Another tell will obviously be the production rate. There are rumors of Tesla moving to a production rate of ~650/week at the end of Q3. Whats interesting about that is that the current(?) 500/unit/week rate is extremely consistent with the simple U.S. model + a gimpier (more complex, less successful) sales model overseas + drawing down their order book. There is anecdotal evidence supporting good U.S. sales, and more uncertain foreign sales because Tesla is just beginning the process of expanding into the E.U., let alone the rest of the world. An expansion to 650/units/week would indicate success with the simple model in the E.U.

I've tried modeling reservations based on historical trends, or via other methods. But I've not been as confident in the predictive capacity of those models.

Predicting based on historical trends in the segment (ie, how well the ICE brands sell month to month) was always a failure, as its clear that the Model S has many buyers who would never consider traditional vehicles in this class (which if you think about it, is a failure on the part of traditional automakers to fully capture their potential market).

Projecting based on known reservation rates (before Tesla discontinued that system) from 2012 and early 2013 also seems fraught with difficulties. Do you project based on December rates, or January rates? Both are problematic for a variety of reasons, and predicting based on the overall trend isn't really better.

My best predictive results have come from the basic store based sales model, and an assumption of a large number of deferrals/cancellations at the end of 2012. I believe the new non-lease-lease was designed to recapture as many deferrals as possible, and there is anecdotal evidence of success on that front.

I question whether Tesla would have gone to a 500/unit flow rate if they were concerned about demand. That likely increased margins, and benefited the stock price, but if it was done at the cost of a major production slowdown in a couple of months that would be a terrible decision. I can't imagine that suppliers would be willing to invest to expand capacity for a short term production contract. If Tesla was forced to cut back to 300/units per month in July to compensate for an empty orderbook it seems likely to me that there would be penalties incurred by their suppliers.

Based on that supposition, I would hope that management would only act to increase capacity if they were seeing demand levels which necessitated it. The simple sales model and the large number of store openings in the last couple of months points to a possible reason for the seeming confidence of upper management.

 

[CapitalistOppressor]

Just to throw in some numbers, Tesla currently has -

25 U.S. stores
9 E.U. stores
3 Asian stores

Simplest of the simple models (ie I don't actually use or rely on this as my real world model, but its interesting) 2 cars sold per day per store = 27,010 cars sold per year.

If the current production rate is still ~500 cars per week, that works out to about 26,000 cars produced per year. 520 cars per week = 27,000 cars/year. The last reported production rate (from a factory tour in mid March) = 509/week.

Again, I don't rely on that for real world projections, but it tends to support my research that Tesla targets 2 cars sold per store per day. They are likely failing to do that right now for a variety of real world reasons, but I find the apparent correlation between factory capacity and retail sales goals to be interesting. Of course, there is a chicken and egg argument to be made that the sales goal is set based on production capacity.

But there are repeated anecdotal reports across a wide range of dates and from many stores which support the notion that Tesla is consistently coming close to the goal at most U.S. stores and heavily exceeding it in California. I don't read European, but fortunately we still have reservation rates which show sales appear to be in excess of 2/store/day.

Edit: In my previous post I mentioned that 500/units/week assumes "gimpier" sales because there is a major expansion of retail stores underway right now in Europe. If those stores (along with every other Tesla store) produce sales consistent with a simple sales model, production will need to be increased substantially.

 

[Cattledog]

C.O. - Good analysis. One tweak is my story about a Tesla employee sharing their '1-2 cars per day sales average' is that it was from the Denver store.

 

[GSP]

Are Tesla's stores open 365 days per year? If not, I would only count the days they are open.

I would not use 52 weeks per year for factory operation. Shutdowns are usually needed for factory maintenance and upgrades. Tesla has already shut down the first week of 2013. I don't know their future plans, but I would start with a 50 week per year assumption.

GSP

 

[CyberDutchie]

Mulder, I'm not sure what the numbers in the chart in your last post represent.

Actually, I should have said next-gen Tesla Roadster, not v2. The numbers are the ramp according to Morgan Stanley for this car that is expected to go into production in 2017.

The plans and timing of the next-gen Roadster was recently confirmed in a George Blankenship interview, discussed in another thread here.

 

[luvb2b]

I've tried modeling reservations based on historical trends, or via other methods. But I've not been as confident in the predictive capacity of those models.

cappy, i've tried a bunch of ways to forecast reservations since the sequence number data stopped. some of the things i experimented with:

1. some kind of poisson process based estimation, using the reports of new reservations in the ordering threads. this is a fail because people aren't reporting reservations very often any more as delivery times have shortened. this makes sense as it takes a user a while to discover the tmc forums and become comfortable enough to post some information. shorter delivery times mean less time to discover and post to the reservation threads.

2. estimating us reservations of eu reservations pace. also a fail as the variability between eu and the usa is quite high. sometimes the usa is running 2x the eu, sometimes 5x.

3. looking for sequential data in rn numbers. failed, rn's are not sequential. much more complex coding of the rn would require many more data points to work out.

4. trying to follow twitter for people reporting ordering or purchasing model s. fail, "ordering model s" search leads you to an article about ordering the murder of a brazilian model. not enough people boast this way.

5. historical trends, using seasonal adjustment factors. there are a bunch of seasonal factors available for auto sales, which show that sales in march are much higher than jan/feb. but as you said, these methods are far from satisfactory.

so i am at a loss. i have no idea how to forecast reservations. and apparently now tesla won't tell me.

all i have is the model i posted to kick off the thread. it shows that under even relatively conservative assumptions they'll produce at full capacity through the year. so i have high confidence that q1-q3 of 2013 earnings should be good.

 

[CapitalistOppressor]

cappy, i've tried a bunch of ways to forecast reservations since the sequence number data stopped. some of the things i experimented with:

1. some kind of poisson process based estimation, using the reports of new reservations in the ordering threads. this is a fail because people aren't reporting reservations very often any more as delivery times have shortened. this makes sense as it takes a user a while to discover the tmc forums and become comfortable enough to post some information. shorter delivery times mean less time to discover and post to the reservation threads.

2. estimating us reservations of eu reservations pace. also a fail as the variability between eu and the usa is quite high. sometimes the usa is running 2x the eu, sometimes 5x.

3. looking for sequential data in rn numbers. failed, rn's are not sequential. much more complex coding of the rn would require many more data points to work out.

4. trying to follow twitter for people reporting ordering or purchasing model s. fail, "ordering model s" search leads you to an article about ordering the murder of a brazilian model. not enough people boast this way.

5. historical trends, using seasonal adjustment factors. there are a bunch of seasonal factors available for auto sales, which show that sales in march are much higher than jan/feb. but as you said, these methods are far from satisfactory.

so i am at a loss. i have no idea how to forecast reservations. and apparently now tesla won't tell me.

all i have is the model i posted to kick off the thread. it shows that under even relatively conservative assumptions they'll produce at full capacity through the year. so i have high confidence that q1-q3 of 2013 earnings should be good.

Back in July 2012 when I was making my decision to invest I tried projecting sales using every method you mention, and all were unsatisfactory. At the time we had great reservation data, but reservation rates didn't really support 20k units per year. And other methods of projection just sucked.

But George Blankenship kept emphasizing the new format stores and was citing foot traffic increases at every shareholders meeting. Substantial time was spent on those metrics, which made me elevate the importance I placed on analyzing the retail model. I have tried to keep my models as simple as possible because its not my full time job.

But I am convinced that Tesla sees the new format stores as the core of their sales strategy, and there is plenty of data there that is amenable to rational analysis.

Are Tesla's stores open 365 days per year? If not, I would only count the days they are open.

I would not use 52 weeks per year for factory operation. Shutdowns are usually needed for factory maintenance and upgrades. Tesla has already shut down the first week of 2013. I don't know their future plans, but I would start with a 50 week per year assumption.

GSP

Re: factory closures, and stores being open affecting the simplest of the simple models (which I don't actually use). Those factors would tend to balance each other, but are dominated by cancellations regardless. But if you assume that the sales per store day is based on a 363 day calendar (Christmas and Thanksgiving off, which is typical in retail) you are still talking almost 27,000 sales.

Taking 2 weeks off at the factory would drop production to ~25,000. That would imply an expected cancellation rate of ~7.5%. Not exactly unreasonable when new customers are finalizing immediately.

There are other layers you need to add as well, such as many stores are old format in low foot traffic areas, and likely underperforming the new format stores based on the effort Tesla is making to shut them down and open new format stores nearby. There is lag after opening before a store reaches goal. Some stores are just better because of higher foot traffic or better demographics. Etc.

You can add as many layers as you like. But Tesla is going to mine as many customers as possible and there is a two way relationship between the retail side and the production side, and you need to take into account the investment that Tesla is putting into new format stores, and new production capacity.

 

[kcveins]

Re: factory closures, and stores being open affecting the simplest of the simple models (which I don't actually use). Those factors would tend to balance each other, but are dominated by cancellations regardless. But if you assume that the sales per store day is based on a 363 day calendar (Christmas and Thanksgiving off, which is typical in retail) you are still talking almost 27,000 sales..

Here in IL, they are not allowed to discuss sales / pricing / features, etc on Sundays. So that would knock off 52 more days as well - it would be reasonable to use a 300 day calendar.

 

[Doug_G]

Just to throw in some numbers, Tesla currently has -

25 U.S. stores
9 E.U. stores
3 Asian stores.

Canadian stores, of course, simply don't exist.

 

[blakegallagher]

I thought Canada had one store? is that sarcasm Doug or is Canada really not represented with a store?

 

[aronth5]

I thought Canada had one store? is that sarcasm Doug or is Canada really not represented with a store?

From Tesla's web site. One existing (Toronto) and two "coming soon" Montreal and Vancouver
http://www.teslamotors.com/yorkdale

http://www.teslamotors.com/findus/stores

 

[Doug_G]

I thought Canada had one store? is that sarcasm Doug or is Canada really not represented with a store?

Yorkdale | Tesla Store Locations | Tesla Motors

 

[bonnie]

I think Doug_G was simply pointing out that CO had left Canada off his list.

 

[Cattledog]

You mean $247 in 2023...that would be a nice return on investment in ten years, I'd go for that!

Yes 2023, bad typing.

 

[vfx]

Here in IL, they are not allowed to discuss sales / pricing / features, etc on Sundays. So that would knock off 52 more days as well - it would be reasonable to use a 300 day calendar.

As I understand it Tesla use postal codes to denote store location sales. So if an order comes from a buyer's home computer, the zip code entered determines which store gets the credit for that sale. Regardless of day.

 

[Zzzz...]

As I understand it Tesla use postal codes to denote store location sales. So if an order comes from a buyer's home computer, the zip code entered determines which store gets the credit for that sale. Regardless of day.

Hmm... With 27 or so stores in North America and above mentioned target sales per store of one or two cars a day, Tesla should average around 50 cars a day. Which is not bad.

 

[c041v]

As I understand it Tesla use postal codes to denote store location sales. So if an order comes from a buyer's home computer, the zip code entered determines which store gets the credit for that sale. Regardless of day.

Would this not mean that the Toronto store would receive a disproportionate amount of the Canadian sales simply because it's the only store that can be assigned a sale?

The patriot in me would like to think that we Canadians boost sales projection numbers by at least a few hundred annually. Our one store could count for quite a lot!