How to Run a Five Whys Analysis on Recurring Operations Logs

Operations teams in aviation collect a steady stream of small records: delayed departures, late paperwork, missing sign-offs, rejected forms, scheduling clashes. Most of these logs are read as a list of incidents. A more analytical use is to treat each repeated entry as a symptom and ask why five times until the cause is a step in the process. This how-to takes you through the method one stage at a time, while staying realistic about how far a simple chain can take you. It concerns administration and does not touch flight procedures or regulatory requirements.

Step 1: choose a repeating entry, not a single event

Open your log and look for a line that appears more than once in a similar form. “Crew briefing started late” three times in a month is a better candidate than a single odd day. Count the occurrences and note the period. If you cannot show that it recurs, you are analysing an anecdote, and the root you find may be unique to that one day.

Output of this step: a one-sentence symptom with a count and a time span, for example “briefing started later than scheduled on several mornings in a four-week period”.

Step 2: gather only the facts you can confirm

Before asking why, collect what is on record: dates, times, who was involved by role, which template or form was used. Leave out guesses. A common error in analytical work is to fill gaps with plausible stories; those stories later become the answers in your chain, and the chain looks neat while resting on nothing.

Output: a short fact sheet. Mark each item as confirmed or assumed.

Step 3: ask the first why and attach evidence

Write the first question, “why did the briefing start late?”, and answer it from the fact sheet. If the log shows that the room was occupied until just before the start, your answer is “the previous session ended late”. Attach the log entry or a note about where the evidence lives. Where you have no evidence, label the answer as a hypothesis and plan to check it.

Output: answer one, with an evidence note or a hypothesis flag.

Step 4: repeat, one layer at a time

Ask why about the answer you just wrote. Keep each layer to a single sentence and add a tag such as “confirmed”, “reported” or “assumed”. A worked sequence might look like this.

  1. Why was the previous session late? Its debrief took longer than the slot. Confirmed by the timetable and the end time in the log.
  2. Why did the debrief run long? It covered more topics than in earlier terms. Reported by the facilitator.
  3. Why more topics? New items were added to the agenda without removing any. Confirmed by comparing two agendas.
  4. Why were items only added? There is no owner who reviews the agenda against the time available. Assumed; to be checked.
  5. Why is there no owner? The agenda was inherited and has never been formally assigned. Reported.

You now have two confirmed layers, two reported and one assumed. That mix tells you how firmly to hold the conclusion before you alter anything.

Step 5: test the weakest link

Find the layer with the least evidence and check it. In the example, the assumption that nobody reviews the agenda can be tested by asking three people who is responsible for it. If they give three different names, the root is supported. If they all name the same person, your chain is wrong at that point, and you should revisit the fourth answer and follow a different branch.

Output: the chain with the weakest link confirmed, corrected or flagged.

Step 6: state the root cause as a process fact

Write the root as a sentence about the process: “The agenda has no named owner and no scheduled review against available time.” Avoid wording that blames a person or implies that a failure of attitude was the cause. A process fact points to an action, whereas an attitude does not.

Step 7: choose one change and define the measure first

Decide on the single change that addresses the root. Then, before putting it into effect, decide how you will measure whether it worked. You already have a baseline from step one: the count of late starts over a period. After the change, collect the same count over a comparable period. Do not compare a short period to a long one, and do not assume that one good fortnight proves a fix.

Output: one change, one owner, one measure, one review date.

Step 8: review honestly after the period ends

On the review date, compare the figures. If the symptom has faded, record the chain and the change in a shared place so the reasoning is available later. If it has not, do not conclude that the method failed. More likely the weakest link was wrong, or a second cause exists. Reopen the chain and branch.

As a parallel from the web, an orthopaedic medical tourism directory whose menu keeps its clinics listing, specialist profiles, news writing, community board and procedure pages apart shows how a site can keep different kinds of information apart. Whether you maintain a website or an operations log, the same discipline applies: keep facts, interpretations and decisions in separate, clearly labelled places.

Common snags while working through the steps

Teams tend to hit two snags. One is an answer that is really two answers joined by “and”. Split it, because each half may have a different cause and a different fix. The other is a layer that restates the layer before it in different words, such as “the room was not ready” followed by “preparation was incomplete”. When that happens you have not moved down the chain; reword the question so that it asks what in the process produced the situation, and answer that instead.

A third, quieter snag is time pressure. If the group is rushed, the chain collapses into the first three answers that come to mind. Book the session in advance, keep to a fixed length, and stop at the end even if the chain is unfinished; return to it next week rather than forcing a conclusion.

Limits of the method

An analytical reader should be aware of what a five whys chain cannot do. It follows a single line of reasoning, so it can miss causes that run in parallel. It depends on the honesty of the people supplying answers. It does not quantify how much each cause contributed. For problems with several interacting causes, treat the chain as a starting hypothesis and combine it with other tools.

It also should not be stretched beyond its remit. Matters governed by formal requirements follow formal procedures. The method here is a lightweight aid for the many ordinary administrative problems that do not need a full investigation but do keep coming back.

A checklist you can print

  • The symptom recurs, and the count and period are recorded.
  • Every answer has an evidence note or an explicit hypothesis flag.
  • No answer ends at a person’s name.
  • The weakest link has been tested.
  • The root cause is a process fact.
  • One change, one measure and one review date are written down.

Summary

Treat your log as data. Pick a repeating entry, collect confirmed facts, ask why in short layers, tag each answer by how well it is supported, test the weakest link, and change one thing with a measure agreed beforehand. The result is not certainty but a better-grounded guess about where a process leaks, and a way of finding out soon whether the guess was right.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top