In my last dialog, I took a trip down forecasting lane. The discussion revolved around the nebulosity of the metrics of CAGR, and forecasting, in general. I stopped short of saying this but, it seems to have some similarities to gambling. I also discussed about how necessary it is to use this tool and how it is an integrated part of business planning.
The next day, what shows up in my inbox but another forecast. So, I thought I would take this opportunity to unpack forecasting a bit, relative to this particular segment. Let us drill down a bit on how these numbers play out in CAGR.
The data started out with the title “This Is How Many Autonomous Cars Will Be on the Road in 2025.” 2025 seems to be a popular year for analysts and forecasters to hang their hats on lately. I cannot find a good reason for that except it is mid-century and enough years to juggle forecasts if mitigating circumstances arise.
For this autonomous vehicle forecast, the estimate from one of the well-known research firms puts that number at 8 million. Their qualifier is that these cars will have level 3, or higher, autonomous driving capabilities. However, there is a second sentence that states that will be the number sold in 2025. So, the title of the story and the description conflict. Is it how many are on the road in 2025 or how many will be sold in 2025?
I decided to take the track of how many cars will be on the road, collectively, by then. So like any valid data set criterion, I threw out that 2025 sales number since that is not a collective statistic. Then I went looking for other data.
Before we unpack this, I think it would be prudent to define, quickly, autonomous car levels. So here they are:
· Level 0: This one is not really autonomous at all. It is just a reference level. The driver controls it all – steering, brakes, throttle.
· Level 1: This is the basic driver-assistance. Most functions are still controlled by the driver, but a specific function (like limited steering assist or throttle control) can be done automatically by the car. If we wanted to get cheeky, we could say we have level 1 forever in the form of the old curb feelers.
· Level 2: This is a basic driver assistance system. Multiple, minimal functions are controlled by the car (steering and acceleration/deceleration, for example). This is acquired from sensors that “see” the road and control the car from their input. Functions like speed, lane control and parallel parking are automated and the driver can be somewhat “hands-off.” This is the level most vehicles are at, presently.
· Level 3: This level adds the “safety-critical functions” to the vehicle, but only under certain traffic or environmental conditions. The driver is still present and will intervene if necessary, but does not require the same diligence level 2. However, this level has some issues in relying on the driver to be able to intervene instantaneously in critical situations.
· Level 4: Level 4 is Level 3 on steroids. Technically, it removes nearly all of the vehicles safety-critical driving functions and monitor roadway conditions for an entire trip from driver control or intervention. The issue that makes this less than level 5 is that the autonomous part is limited to the “operational design domain (ODD)” of the vehicle – meaning it is not capable of covering every driving scenario and has no non-visual communication with the infrastructure or other vehicles.
· Level 5: This refers to a fully-autonomous system that expects the vehicle’s performance to equal that of a human driver, in every driving scenario – including extreme environments, like dirt roads, that are unlikely to be navigated by driverless vehicles in the near future.
Ok, now back to the forecasting. Doing my own research, I came up with the following numbers (in no particular order, and just picked off the Internet and some discussions). This includes all levels, however, there are not likely going to be any real level 5 retail vehicles by then (unless you believe Elon Musk, who thinks his vehicles are already level 5), according to my expert source at California Path.
There may be some commercial, constrained, fully autonomous vehicles in fleet or narrow verticals but they will be heavily bounded in implementation (like special roads, airports, city centers, and the like).
Now let us look at some numbers:
· 600,000
· 15 million
· 36 million
· 60 million
If one plugs these numbers into the CAGR formula the results are, respectively (assuming the five years, 2020-2025 as n and we assume 10,000 autonomous vehicles as the starting value. It makes no difference what the baseline is for this discussion since the variance in CAGR numbers is what is relative):
· 126 percent
· 331 percent
· 414 percent
· 469 percent
Interestingly, the values are closer than I expected, except for the first value (600,000). As well they do not change as much as the starting value varies. For example, if the beginning value is 1,000 instead of 10,000, for the first line value of 600,000 is 259 percent versus 126 percent. For the last line item it is 802 percent versus 469 percent. Considering that the starting value has one order of magnitude difference that is an interesting observation. The change in CAGR is roughly 100 percent, versus an order of magnitude.
I am not a statistician but, I am a pretty good mathematician. My observation of CAGR remains somewhat skeptical. It does seem that the formula retains a modicum of linearity even with wide swings in input and different bounds for defining the data (autonomous car levels vary by data, for example). The output swings are much narrower than the input swings.
The question begs: is this a good metric for betting on the future? Observation of usage suggests it can be. However, in talking to some people about this, my favorite reply was “it is better than nothing.” It seems that it is a great tool for number crunchers and reports – less so for actual company planning.
OK. Obviously, there is much more to business forecasting than just CAGR. Another one of my executive business contacts told me that, after all the research and crystal-balling is in, CAGR is a high-level look at what type of growth might be expected if the data is deemed reliable. However, rarely is it taken as a primary forecasting metric.
So, in the end, CAGR is just another tool for business prognostication. I did a lot of legwork to dissect that and it turns out it is much less special than I expected – bummer.