The late Sir Winston Churchill once said that “Statistics are like a drunk with a lamp post: used more [for] support than [for] illumination”. While this is not always true for all statistics, one form of statistics, to wit, Google Diagnostics, is oftentimes remarkably un-illuminating. In this article, we will take a closer look at the phenomenon of automotive diagnostics via Google in terms of why it exists, why it is not always reliable, and how or why it is so often the bane of the professional technician’s existence. Let us start with this question-
Google Diagnostics is also known as “statistical diagnostics” and while the term means different things to different people, the basic idea involves different organizations collecting all known instances of trouble codes, parts failures, system malfunctions, and/or programming errors that occur on all vehicles. Car manufacturers also do this by various means during the pre-production test phase of new models to prevent serious faults and defects from passing into production models, which it must be said, with varying degrees of success.
So what’s the problem, you may ask? You may very well ask, because while consulting with workshop owners/managers/chief mechanics on tricky diagnostic issues, this writer has noticed a marked increase in the number of technicians that use Google, and especially YouTube as primary sources of technical information to diagnose some types of customer complaints.
We understand that it can be difficult to source reliable technical/service information, and while there are many excellent online resources available that offer OEM-level service information, many, if not most of these resources require that (prospective) users purchase expensive subscriptions to access the information they need. We also understand that while some franchise chains maintain internal diagnostic databases, or subscribe to relatively expensive services like Mitchell’s ProDemand or similar services like nastf.org, many small workshops, and especially those in economically depressed areas simply do not have the wherewithal to afford access to expensive service information. Therefore, against this background, many technicians resort to Google, since they can hardly be expected to pay subscriptions to specialized sources of information out of their own pockets-
Image source: Google (search result)
The search term in this partial screenshot from a Google search result page may be pretty generic, but even specific search terms produce similar results. In this instance, the query returned nearly 70 million results, and in almost all cases, the first page of search results will include several YouTube videos (including the one shown in the first place) that offer diagnoses and fixes for almost any imaginable car problem.
Note though that this is not the same as saying that everything on YouTube and other video channels is bad, wrong, inappropriate, or misguided. Far from it, because there are many excellent how-to guides available that offer real diagnoses and proven fixes for many car problems, but the problem is that if your search returns 70 million or so results, it becomes very difficult to distinguish between good and bad advice and/or information, which begs this question-
At the risk of overstating the case, it is worth repeating that it is not the intention of this article to “bash” YouTube. However, as far as automotive diagnostics is concerned, it must be stated that in almost all cases, the position of a particular video offering in search results has far less to do with the relevancy or accuracy of the information it contains than with that particular video's SEO score, marketing effectiveness, and/or its number of views. We need not delve into the reasons why so many people produce and place videos on car problems on this site, but the high number of views some videos enjoy, and the possibility of monetizing those views are powerful motivators, even though Google cannot distinguish between good and bad information.
Nonetheless, many people (including many professional technicians) simply type in a search query that contains perhaps a DTC and some symptoms plus the make, year, and model of the vehicle, and wait to see what turns up. This writer has witnessed professional technicians do just that, and then follow the advice offered by the first video that pops up. If the proffered solution does not produce the desired results, they simply go on to the next, then the next, and so on until they stumble across a solution to their problem. While DIY or non-professional mechanics might derive some benefit from Google Diagnostics, this is no way to perform modern diagnostics in a professional environment.
So are all forms of Google Diagnostics bad? Not necessarily: as a matter of fact, there is one form of internet diagnostics that can be extremely helpful, and you may be using it already without realizing it, so let us look at-
Although this process of performing diagnostics it is far from perfect, and almost invariably requires a subscription, it largely eliminates the “shotgun” approach and offers relatively focused results that are derived from the real-world experience of thousands of technicians, engineers, and other car repair professionals the world over. Note thought while these platforms avoid the shotgun, or “throw-parts-at-it-until something-sticks” approach, no single platform or resource will guarantee always to have the exact answers to your particular problem or set of symptoms, although this is becoming less common as databases are expanded.
Nonetheless, how do statistical diagnostics platforms work? It’s simple really; most operate as repositories of repair and service information that covers the entire spectrum of car repair. Typical information includes known fixes, waveform libraries, wiring diagrams, workarounds for tricky problems, and other types of repair information that is accessible to all members in good standing. You may already know this, but what you may not know is how organisations such as Mitchell 1, Haynes, Identifix, and others collect the information they offer for sale to members.
While some platforms provide service, repair, and other information drawn from the manuals they publish, others rely solely on members/users to post known/confirmed fixes to all manner of car problems. However, the publishers of manuals augment their databases with information supplied by users (technicians) working under real-world conditions, and in one case, technicians from around the world now have access to 50 million or so records in a single database, but-
Put simply, it means that there is a very high statistical likelihood that your preferred source of technical, service, and repair information will have the information you need to diagnose, repair, or resolve a particular customer concern. Here is how it works-
If you are a veteran of the car repair trade, you will be familiar with the term "pattern failures". If however, you are new to the car repair industry you should know that "pattern failures" is a generic or collective name for certain faults and/or failures that occur more frequently on some cars than on others. If a sufficient number of the same failures occur on a certain model or number of models within a model range, a pattern is established, and as a result, all failures that fall into that pattern are referred to as pattern failures, which form the basis upon which statistical diagnostics is built.
For instance, let us say that Platform A has 14 000 records for a particular vehicle- let's say a 2013/14 Toyota Corolla. Those 14 000 records will be divided into categories that typically include Components, DTC's, Systems, Common Symptoms, etc, and all searches will be based on the model designation, engine size, trim level, and in some cases, the vehicle's VIN number. Thus, if you are presented with such a Corolla that is exhibiting say, a severe misfire, you simply add the relevant DTC(s) to the required search parameters, and with some luck, you might retrieve a known fix for that particular issue. However, a word of caution is in order here-
While you may save some time by immediately implementing your retrieved fix for your misfiring Corolla, you are at this point not sure what had caused the misfire in the first place. Therefore, most statistical diagnostic platforms offer a series of additional resources such as guided tests to help you either confirm or eliminate the most common causes of misfires on your particular Corolla. If you are an experienced technician, you will likely forego these tests since you might recognise the misfire as part of a pattern failure, but if you are not so experienced, these guided tests can save you a lot of diagnostic time.
If we take another example, your source of technical information may have several hundred thousand records for a particularly troublesome vehicle/engine version, the most notorious of which is without a doubt the dreaded "No crank / no start" problem on the 6.0L Ford Powerstroke engine, which is an excellent example of a pattern failure. Almost all 6.0L Powerstroke engines from the early to mid-2000s suffered from it at some point in their lives, but the point is that in statistical terms, most, if not all of the best-known causes and fixes of the "No start" condition on this particular engine will be included in the database.
As with the Corolla example you will of course, have to test systems and components to arrive at a definitive diagnosis for the Powerstroke engine’s no-start condition, but the upside is that a precise search string will return a relatively focused set of possible causes and fixes that have been verified by experienced technicians many times before. As a practical matter then, statistical diagnostics does not depend on clever marketing or SEO rankings in any way to be relevant, and therefore, it is unquestionably more accurate and reliable than any other method of Google diagnostics.
However, for all its advantages, statistical diagnostics also has significant disadvantages, the most important of which is arguably the-
It must be understood that a), fault detection is very different from failure diagnosis, and b), that statistics-based diagnostic tools cannot distinguish between the two aspects of automotive control systems. This is an important point because while fault detection is concerned solely with determining that a fault had occurred somewhere in a system, failure diagnosis is concerned with determining what had caused the fault.
From our perspective as technicians, failure diagnosis is the more important scheme because it is designed not only to reveal the kind of fault that had occurred but also for the integration of more robust mechanisms to mitigate the effects of failures on both affected and related systems. This is quite a mouthful, but in translation, it simply means that failure diagnosis makes it easier to find the actual site of a failure, whether the fault was caused by a parts failure, a programming error, or a simple wiring issue.
However, while failure diagnosis makes it easier to trace faults even with generic scan tools, the complexity of modern automotive control systems has the corollary that in some cases, the cause of the fault cannot be determined. This circumstance is commonly expressed as No Fault Found, No DTS’s Present, Trouble not Identified, or Trouble Cannot Be Duplicated- depending on the scan tool or laboratory test method used.
In practice though, the cost of the No Fault Found phenomenon to car manufacturers can often be counted in billions of dollars over the life of a model or model variant, since components that are returned to manufacturers for testing are often found to be free of defects, errors, or other shortcomings. Moreover, the suspected failure that had caused the removal of the component can often not be replicated under laboratory conditions, which means that many parts and/or components are sometimes replaced needlessly in a "shotgun" repair because the primary fault cannot be identified, neither by a technician in a workshop nor by engineers in a laboratory setting, which leaves us with this-
Given the above, it could happen that replacing suspected failed parts or components purely on the strength of the high statistical probability that the suspected parts or components are indeed defective may not resolve a customer concern or complaint. Therefore, while statistical diagnostic tools and platforms might be more effective in resolving customer complaints than a random selection of YouTube videos, these tools and platforms must be used with some circumspection and with due regard to the No Fault Found phenomenon.
All major car manufacturers are currently engaged in developing and/or refining failure diagnosis strategies not only to reduce No Fault Found incidents but also to improve the accuracy of scan tools to allow for more informed part replacements by constraining the number of possible causes of even simple faults under real-world operating conditions.
Nonetheless, despite the ongoing developments in failure diagnosis and its application to statistical diagnostics we should resist the temptation to develop an over-reliance on statistical diagnostics since it is neither designed, nor intended to replace conventional and well-structured diagnostic procedures. Using a statistical diagnostic platform or tool to point one in the right direction might save some time, but as a practical matter, it is more profitable to reserve YouTube and similar resources as a source of entertainment between tasks-not as a primary source of service information to help us complete those tasks.
