Root Cause Analysis

Vagueness wastes everyone's time. "The copier isn't working" tells you nothing useful. Telling a colleague to check the paper is likely a wasted errand when the toner, scan light, power, or feed mechanism could equally be at fault. A strong problem statement answers what happened, where (geographically, in the process, on the product), who is affected, when it began, and how much (absolute numbers plus percentages).

Exclude any hint of the cause. Okes warns that smuggling assumptions into the statement derails diagnosis. One company labeled dangerous equipment failures "accidental" and chased solutions for months, until it discovered the acts were intentional. Naming a suspected person or cause prematurely blinds you. Two hospitals comparing surgery times got nowhere until they realized each defined "operating room time" differently, a missing operational definition.

Everything is a process. Before brainstorming causes, flowchart the steps between where the problem appeared and a sensible starting boundary, ideally four to eight boxes using action verbs. Okes uses SIPOC (Supplier, Input, Process, Output, Customer) to show any organization can be mapped this way. When output disappoints, something inside the process broke, usually because there were no standards, the standard was wrong, or the standard wasn't followed.

The dyslexic clerk reframes blame. A fitness club keeps misspelling members' names on cards. The owner scolds the desk clerk, who turns out to have dyslexia. Firing the clerk fixes only the physical cause; the real gap is a hiring or screening process that placed the wrong person in a spelling-dependent role. Quoting Sydney Dekker, Okes urges finding how people's actions made sense given their circumstances.

Pareto runs through every step. The 80/20 rule, that a minority of causes drives the majority of impact, is Okes's philosophical thread. Faced with multiple independent causes (an order error could come from data entry, a software algorithm, or a packer), don't fix all simultaneously and disperse your energy. Attack the cause with the highest frequency, cost, or risk first, then decide whether the next one is worth pursuing.

Filters protect overloaded staff. Okes introduces "corrective action density": divide the number of corrective actions opened per period by headcount. A 250-person firm logging 400 actions a year (eight a week) has overwhelmed the few people who actually do diagnosis. The answer is a filter that screens which problems warrant full investigation versus which get logged and watched for recurrence, judged by frequency, cost, risk, and current workload.

Two complementary maps. A flowchart gives the horizontal, time-ordered view; a logic tree (or why-why diagram) gives the vertical, structural view. The logic tree is a simplified fault-tree that breaks a problem into incremental cause-and-effect branches with infinite depth. The copier that ejects blank paper: since a blank sheet was picked up and ejected, those steps work, so only the image-scan and image-transfer steps remain as suspects. Each branch eliminated kills all the micro-causes beneath it.

Forbid magical jumps. When a student blamed damaged product on "operator attitudes," Okes pushed back: an attitude cannot physically dent metal. Damage obeys physics and chemistry. Good branches read "inadequate training," not just "training." When stuck on one diagram, switch to the other, exploiting that the brain processes patterns both sequentially and spatially.

Data means evidence, not just numbers. Step 4 hunts for the relationship between X (a suspected cause) and Y (the problem outcome). Run a thought experiment first: if this cause were the real one, where would evidence show up and in what form? Then collect via interviews, observation, records, scientific testing, pictograms, or check sheets. A clever swap test isolates causes: move step two between two identical lines and see if the failure follows it.

The bagging-machine case shows patience. Bags occasionally came out overweight. Full inspection revealed the heavy bags always appeared right after breaks. When idle but powered on, the machine's preload container overfilled. They couldn't redesign the machine, so they changed the procedure to shut it off when not running continuously. A child's logic captures the spirit: before yanking the fridge from the wall, just open the door and see if the light comes on.

Solution strength is a hierarchy. Drawing on the U.S. Department of Veterans Affairs, Okes ranks fixes: stronger ones physically change the process or environment, simplify, or mistake-proof; intermediate ones use job aids and reduce confusing similarities; weaker ones rely on training, warnings, and adding another check. The reflex to bolt on an inspection catches problems sooner but rarely prevents them, missing the whole point of corrective action.

Training is the comfort blanket. People reach for retraining because it shows due diligence in a courtroom, but if someone was already trained, retraining changes nothing unless you find why the first training failed. To generate stronger options, deliberately think differently: scale the problem up or down, use analogies, ask "what would X do," or suspend all limits. One firm that suspended constraints realized it could eliminate the mislabeled box entirely, since electronic data transfer had made the label obsolete. Favor Occam's razor: the simplest solution is usually best.

Verify both the fix and the outcome. After implementing, don't just confirm the problem (Y) improved; confirm the change (X) was actually carried out. Okes's solution-outcome matrix yields four cells, and two are traps. Y can improve even though X wasn't really implemented, often the Hawthorne effect, where performance rises simply because people know they're being watched. Conversely, X can be implemented perfectly yet Y stays bad, meaning you found the wrong cause or misunderstand the system.

Then institutionalize, or it reverts. Standardize documents, FMEAs, and training, spread the lesson to similar processes, and sustain the gain. The strongest tactic is making the old way impossible, for example removing the obsolete equipment. Until individuals fully internalize the change, the organization risks sliding back, so group and structural reinforcement matter as much as individual buy-in.

Emotion hijacks logic. Okes catalogs cognitive biases that sabotage root cause work. Confirmation bias makes you hunt only for data supporting your pet theory. Anchoring makes you call a result "good" merely because it beat the last one. Recency makes you assume a recurring problem has the same cause as before. Availability bias, what Herbert Simon called "satisficing," makes you grab the easiest data rather than the most definitive. Recall error corrupts interview testimony.

The antidote is slowing down. Effective diagnosis requires getting the rational neocortex in charge instead of the reptilian fight-or-flight brain that arbitrary deadlines and a blame culture provoke. Okes borrows from critical thinking (state your assumptions explicitly), Buddhism (see things as they really are, don't criticize the people involved), and Stoicism (accept what is and look objectively at what you can change). When told "there's a bird in my house, what does it mean?" he replied, "it means there's a bird in your house."