universal averages on used oil analysis (UOA)

[johnsmith]

Some, but not all testing labs provide information regarding "universal averages" or "unit/location averages" along with the rest of their analysis. I did a quick search, and I can't seem to find any information about how they develop their norms. I have a question for the lab guys or really anyone who knows: what do they control for? Oil brand, weight, model year, mileage on automobile, time, does the "location" refer to region, state, hemisphere... that sort of thing. Also, what what computational method is used? I figured this had been discussed at some point and I just wasn't looking in the right place, but if it's a secret, then forgive me for asking. Just curious

 

[Gary Allan]

Well, the location is where the report is for. If it's your house/home/car/remote pumping station ..that's what average it's for. You may be testing a generator ..out of 30 generators/stationary units. Hence "unit or location" averages. "North 40 Farmall" The universal averages are for that engine ..of all that they've tested, or so I reason. I saw one UOA for a member here that the tech commented that it was the first experience with that engine ..and hence they had no bearing on what to state what was normal or excessive in terms of results. I don't know how far they go for in terms of defining the universe. It may be year and model ..or just the specific engine (regardless of year). You can always email Blackstone to get their version of the story.

 

[johnsmith]

Are you sure? I always just assumed the unit/location averages referred to the compiled averages from all unit Xs within a designated area (e.g. these are the unit/location averages for all the 1.7L Honda Civics in the Midwestern United States).

 

[Gary Allan]

No ..YOUR averages. For example. I had only ONE Dyson UOA package from Blackstone. 11-11/ULA 6/6 3/3 etc etc etc. If you have three test that average 30 for that element ..then that is "your unit" average (or in the case of a stationary installation "location") Here's one with multiple tests: Now (squint) add up the 3 test numbers and you'll see that it equals the average figure. Then compare it to the figure on the far right.

 

[GMorg]

From Blackstone on "universal averages": "We do not separate out the various oil brands and grades when we calculate universal averages, so in effect, the universal averages are a mix of all different types of oil out there. Therefore, the additives that are present in your sample will not match those in the universal averages column."

 

[johnsmith]

Now I get it! Universal averages = everybody's averages for this car, truck, or whatever. Unit/location averages = your averages on that particular car, truck, or whatever, or your averages on a particular group/fleet of cars, trucks, or whatevers. (In other words, they're bundling up your UOA history for whatever item it is you're testing and showing it to you.) It all makes too much sense when you look at it that way. Ouch... my pride. Alrighty then: if they don't control for oil type or grade when doing the "universal" averages, then, like they said, the add-pack numbers are no good. And you're right about them controlling for mileage, because I remember reading a few reports where they said something like: "the universal averages for this car are shown after 4,658 miles..." But what about everything else? The universal averages column could be extremely useful, but only if it's not packed with confounding variables. So say I get my 5.3L GM tested, and the copper is high. I start to panic, but wait... no problem, because I look over and see the universal average for my engine shows high copper as well. I don't have a problem, my engine is just wearing like everyone else’s. Or... I do a UOA on my 2.4L Acura and the lead is 5ppm. No big deal, what's 5ppm right? But then, I look over and see that 1ppm is the average for this particular engine, meaning that something might be amiss. Probably wouldn't have even been concerned otherwise. Or... http://theoildrop.server101.com/ubb/ultimatebb.php?ubb=get_topic;f=3;t=003438 I do a UOA on my 5.4L F-150, and my iron is half of the universal average. Maybe I have a magic bullet routine? Everything else is looking good, now I can possibly extend my OCI, and/or stick with the oil I'm using. http://theoildrop.server101.com/ubb/ultimatebb.php?ubb=get_topic;f=3;t=003373 Granted, these results would need to be repeatable, but hopefully you see my point about their utility. Now, that said, the universal averages are also ONLY useful if they aren't confounded by a million different variables. How do I know that the universal averages aren't thrown off by new car owners constantly doing UOAs? Newer cars tend to show more wear metals; is the UA column giving me an accurate comparison basis for my car with 60,000 miles? (Need to control for mileage on automobile). How do I know that Uncle Ted isn't sending in a UOA every 6 months for his Acura that sheds 45ppm of lead? Is that affecting the universal average? (need to control for outliers). How do I know that half the people with my car aren't using the recommended grade? Bulk 10w30 dino instead of the high dollar 0w30 synthetic I used? The two oils in the same engine are going to show different wear patterns (need to control for oil viscosity/type). What about filters? And on and on and on... The smaller the number of confounding variables, the more useful this piece of information is. The fact that most of the people doing UOAs in the first place are of the BITOGer sort is probably the biggest factor they have on their side. But what if that changes? Add to that the fact that the labs themselves have a certain margin of error, as well do the auto manufactures, as well do the oil manufacturers, as does the sampling technique itself, and with people dumping all sorts of additives in their oil and gas, the “comparison” will be “comparing” apples to oranges after a while. Sorry for the rant, but I just wish the information was more accurate. UOAs as a rule are only good for establishing general trends in your own engine anyway, but it would be so very nice if there were an accurate norm that you could compare your baby with. My car shows XXX amount of this and that in a trend covering the last three OCIs! Great. You can probably tell even without a universal average that your car is wearing jut fine. But overall, is that a result of your anal-retentive maintenance, or the engine/oil design? This accurate table of universal averages I have shows that most people who run bulk oil for XXX miles, or mid-range XwXX tend to show this kind of wear… is your anal-retentive method cost effective? Instant BITOG in a spreadsheet, just add water. We’ll have a mini “real universal average” with a lot the above mentioned variables controlled for and available for viewing when TomJones76 et al finish their project, but it would be so much better if the UOA companies had done this to begin with. (Or that they would publish the data if they have.) Oh, and I still don’t know what all the variables they control for are, if anyone knows. johnsmith, over and out

 

[GMorg]

Those light bulb over-the-head things are kinda cool, aren't they?

 

[Gary Allan]

quote:

---

The smaller the number of confounding variables, the more useful this piece of information is.

---

Or the broaden the number of samples. It would neutralize individual anomalies rather well. For all we know they do a normal statistical distribution and only allow so many standard deviations to the data. Outside those confines ..they reject them.

 

[GMorg]

Since you can't go below zero on these measurments, the distribution will be "normal" over a narrow range. I too would hope that the catastrophic failure numbers would not be included, but since most of the reports here seem to me to be below average wear, I'm guessing that the averages are just the averages.

 

[johnsmith]

The problem is that you would need at least 30 UOAs for EACH variable if you were going to force a normal distribution. Then you could make up your cutoff scores and whatnot to leave Uncle Ted's UOA out of the picture. ****, maybe I just answered my own question about controlling for outliers: how much business are these UOA companies getting anyway? If it's not a lot, maybe they just don't have enough samples for stuff like this. I have no idea. Add that to my list of questions. It would be really nice to see what the average wear values on a healthy engine were for say, a 3.1L GM. I don't need everyone's coolant failures contaminating the averages; it would be nice to just leave them out. It would just be so peachy if they could control for some of the other junk, but if you can't chop off the extreme values, then the "averages" for each condition wouldn't mean anything really. And how would you know what's an extreme value one way or the other if you had only run a handful of samples for 1.7L Honda Civics that used 5w20 Syntec along with a dose of LC and a bottle of Techron? Plus, you have to figure that changing oils in between runs will result in different values than if you had run that oil for three OCIs straight. "Mobil 1 was just cleaning up some of the stuff left behind, or your engine was adjusting" or whatever it is they say. Hmmm... this is no good. Here's what was in my mind as I started this thread... ***I send in a UOA for my 1999 X-ride with 5,000 miles on SM Havoline 5w30 dino. They would say: hey John, here's your wear values for this run, here's the averages for everyone with your same car, and here's the values for everyone with your same car who ran 5w30 Havoline dino, 5w30 Mobil 1, and 10w30 Chevron Supreme adjusted for a 5,000 mile OCI *** That would be awesome. Maybe that's asking too much for $20/30 though, even if they had the ability to do it. Is this kind of stuff part of what's included in a "Dyson Analysis"? I've never done one; I would pay extra for it. Also, how do they come up with a "universal average" or for that matter, a "unit/location average" if the time/mileage isn't exactly the same on all the samples? Wear levels through an OCI aren't linear are they? Do they have some sort of formula they use to convert the wear values from a 3,000 mile OCI up to a 5,000 mile one, or vice-versa? I think I need to leave this alone. UOAs on personal vehicles at present are good for establishing general trends in your engine for the purposes of comparing how your car does with different oils or OCIs, and for the purpose of a VERY VERY general comparison with the rest of the world to see if your engine is wearing about the same as cars owned by other people. Maybe that's good enough; I'm too unrealistic sometimes.

quote:

---

Those light bulb over-the-head things are kinda cool, aren't they?

---

Not if you can't do anything about it. Just makes you sad. "universal averages on used oil analysis (UOA)" = diary of a madman. Sorry folks.

 

[Gary Allan]

quote:

---

Also, how do they come up with a "universal average" or for that matter, a "unit/location average" if the time/mileage isn't exactly the same on all the samples? Wear levels through an OCI aren't linear are they? Do they have some sort of formula they use to convert the wear values from a 3,000 mile OCI up to a 5,000 mile one, or vice-versa?

---

This is not a science. It's a study. All the data is inferential. You need to be someone like Terry to really get a grip on the details. The only truly reliable data in the UOA is the TBN, Visc, flash point, water content, coolant (and their components) and insolubles. The rest is to warn you of "troubles". Virtually no one here even bumps the walls in terms of elemental indications to any substantial level. Most of the data is comparative. That's why everyone will tell you that trending is the only reliable way to interpret the results.

 

[Gary Allan]

quote:

---

Originally posted by GMorg: From Blackstone on "universal averages": "We do not separate out the various oil brands and grades when we calculate universal averages, so in effect, the universal averages are a mix of all different types of oil out there. Therefore, the additives that are present in your sample will not match those in the universal averages column."

---

Yes, so you get a mass composite of elemental and additive numbers from all samples tested. Mileage is also average. They only differentiate the engine. Then the average mileage is indexed with the average result. That's how they get: "Your wear is below averages from oil that had been in use xxxyyyzz miles. You did less mileage on this interval, so we recommend zzzyyywww miles on your next interval". So you get the average wear over the average mileage at that average additive level.