Back to my UOA, I have some additional info ...
I'm working with Blackstone to understand the variance in UAs (universal averages) and how they are shifted by standard devations (quality control statistical analysis is what I do here at work).
Of note, we need to first understand that the UAs in a UOA (lot's of acronyms, eh?) are not static; they move up/down a tiny bit over time. If you keep your UOA reports from Blackstone, don't just look at the unit results (your vehicle data) but also look how the UAs slowly shift a tiny bit from report to report. You'll not notice this unless you lay them out side-by-side. But you begin to see that UAs are fluid; they do not stay constant, although they only shift perhaps 1ppm or a bit more over a long time period.
Second, you have to look at the mass group data stream to understand how ALL our UOAs are viewed in respect to each other. This is the concept of macro analysis. Micro analysis can be applied to your one unit and it's performance over time; that is where consistent use and inputs are key (no hopping from brand/grade to brand/grade or you'll distrupt the continuity!). But MACRO analysis allows us to compare our info to other similar equipment. Here, standard deviation plays into account.
I'm working with Blackstone to put together data for some popular engines as examples. One that I'm doing is the Dmax (my personal interest, and because it's a very consistent performer in UOAs and has essentially been unchanged in design/structure since 2001 introduction). Here's the info in a nutshell ...
I have 525 UOAs from Blackstone for Dmax engines over the last 12 months. (Don't worry, they do not share, nor do I seek, any personal client info - I only get the raw data and nothing more). The UAs are only part of the picture; the rest is the amount of normal variance (std dev).
Also, note that of the entire data stream of 525 UOAs, only about 15-20% of them are synthetic users. That means that the data is heavily and predominantly biased towards conventional lubes. Some are obvious because I recognize signature addtive amounts; others I have to infer because of differing grades, etc. But generally, you can easily say that 80% or more of the Blackstone Dmax UOAs are probably dino oils. Keep that in mind for later ...
Please note that I ran the data down to two decimals, but rounded to the nearest whole number for ease and clarity here, except for Cr which is so low, I had to use one decimal so that it would show up.
Here is the list:
element.... UA .... Std Dev
Al..... 3 ..... 1.2
Cr..... (.3) ..... (.5)
Fe..... 16 ..... 10
Cu..... 16 ..... 53
Cu'..... 4 ...... 4
Pb..... 2 ...... 3
Note that Cu is listed twice (Cu and Cu' for copper and copper prime). I had to do that because some Amsoil/RL/RP UOAs spike the data considerably. There are 40 of the 525 UOAs that have GROSSLY exaggerated Cu readings (often in the multiples of 100ppm). That really skews the data. Just look at the UAs for Cu and Cu'. The Cu is the total of 525 UOAs; the Cu' is all the rest of us (mostly dino oil) folks without the Cu spikes! Look at how the UA drops from 16ppm to 4. Further, look how the std dev dropped from 53 to 4! IOW, if we look at the WHOLE group, the Cu culprits really skew the data. In fact, of the 40 UOAs with high Cu, here is how they stack up:
15 wer over 50ppm
7 were over 100ppm
9 were over 200ppm
3 were over 300ppm
1 was over 400pm
1 approached 500ppm (494)
Now, does not that seem obvious in how the synthetic "chemical reaction" can skew Cu data???? That is why I separated out the Ams/RP/RL from the others. Again, look at the Cu and Cu-prime data for UA and std dev. That tells a big story in just a few numbers. HOWEVER - it is VERY important to note that I did NOT pull out all Ams/RP/RL UOAs; I only pulled out those with high Cu. There are several still in there that do NOT exhibit this trait, and I left them in.
Now, with this info, go back and look at my UOA posted in this thread. My Dmax is so very "normal" it's astonishing.
What I often try to get folks to understand that is when your UOA (regardless of what equipment your testing) is a few PPM off of the UA, you MUST understand what the Std Dev is to know what variance is normal. Most of us have heard of Six-Sigma; that's what this is all about. Macro analysis allows us some lee-way into viewing data with std dev in mind. Most certainly if you are within one or two std dev's, you're very "normal".
However, the Cu culprit UOAs are massively skewing the data in a Dmax UOA. They really should almost be their own subset, because they are really shifting the UAs and the std dev for Cu.
I'll also say this; there seems to be very little correlation between Cu spikes and other metals (most predominantly Fe). Of the 40 UOAs with very high Cu, only 3 (less than 10%) also had high Fe. We cannot hypothesize on such a small group; that's not good science.
It is also important to note that while the UA mileage is 7.6k miles for my data stream (slightly higher than what the current reported UA is at 6.6k miles), the Std Dev is almost exactly 4k miles. So if you run up to a 10-11k mile UOA, you are well within "normal" on your Dmax, and your data is not likely to skew the UA results much.
This is EXACTLY why you cannot look at a singular UOA and say that any one brand/grade of oil is "better" or "worse" than another. There is zero data to substantiate this type conclusion. But what you CAN state is that if your UOA falls within one or two standard deviations of the UA, your combination of lube/equipment is performing on par with all the other folks out there. It is exactly why I say that when both a dino and syn lube both turn in nearly identical numbers (well within one std dev of the UA) that there is no advantage to the syn; it did not perform any "better" than the massive group if dino lubes. Conceptually, there is no "lower limit" deviation. We can desire (although perhaps not attain) zero wear. When the UA is 4 and the Std Dev is 2, your second sigma is zero and your third is -2; oil cannot "add" metal back to the engine. So we only really concentrate on the UL (upper limit) 3rd sigma or less. Anything outside the 3rd sigma is "non-normal". That does not mean your engine will soon explode; it means the data is telling you that something is out of control range. You'd have to then know your data trends and ranges for your specific unit to see how it may or may not be "normal" within it's own sub-set stream.
I am going to write an article about this concept for a future posting here, but thought I'd take this moment to show you how "normal" and "non-normal" data develops. I'm going to use the Dmax 6.6L diesel and the Ford 4.6L gasser as examples because they are very prominant in their respective market place, and have a long history with little structural change.
Stay tuned.
P.S. Blackstone is not prepared for all of you to start calling and asking about the Std Dev for every vehicle you own, so please don't innundate them with calls and emails. They are not set up to statisically correlate data or run this type of methodology; it's not what they do for a living. This is an excersise we are putting together to help you all understand how to properly use UOA data and make solid conclusions, while not making unfair ones.
