Dr. Wallace's Favorite Mental Models

The Five Whys

A simple habit of asking "why" again after you think you already have the answer — until you reach the cause that actually fixes the problem.

1Fast overview

In one sentence: when something goes wrong, ask "why did that happen?" about the answer you just gave, and repeat roughly five times, until you land on a cause that, if changed, would actually prevent the problem from coming back.

Developed inside Toyota's production system in the mid‑20th century and later adopted across Lean and Six Sigma practice, it is one of the simplest root‑cause tools available: no software, no training class, no data set. Just discipline about not stopping too soon.

Origin: Toyota Production System Domain: Problem solving & operations Time to learn: 2 minutes Time to apply: 10–20 minutes

2Mental Model Compass

Where this model sits relative to other problem‑solving tools, so you know when to reach for it.

Quick fixDeep diagnosis

Goes well past the symptom, but stays faster than a full investigation.

IntuitiveStructured

A clear repeatable format, but light enough to run from memory.

Solo useTeam use

Works alone, but strongest in a short group conversation after an incident.

Single causeMany causes

Best suited to problems with one dominant causal chain, not tangled webs of causes.

3Try the chain yourself

A delivery truck broke down on the highway. At each step, pick the answer that actually digs deeper instead of just restating the symptom.

Why 1 of 5
0 good calls
Why did the truck break down?

4Diverse examples

Manufacturing — the original Toyota case

  1. The machine stopped.
  2. A fuse blew from an overload.
  3. The bearings weren't lubricated.
  4. The lubrication pump wasn't pumping enough oil.
  5. The pump's intake was clogged with metal shavings — there was no strainer on the pump.

Customer service — a recurring complaint

  1. A customer complained their order arrived late.
  2. The package left the warehouse a day behind schedule.
  3. The picking team didn't get the order until the next morning.
  4. Orders placed after 3 p.m. queue until the next shift.
  5. No cutoff‑time notice is shown to customers at checkout, so expectations and capacity never match.

Software — a repeating outage

  1. The checkout page crashed under load.
  2. A database connection pool ran out of connections.
  3. A background job was holding connections open too long.
  4. That job retries silently on failure without releasing its connection.
  5. No one owns alerting for that job, so the retry bug shipped three months ago and was never caught.

Team conflict — a missed handoff

  1. A deliverable was late for a client meeting.
  2. The design file wasn't finalized when engineering needed it.
  3. The designer was still waiting on brand approval.
  4. Brand approval requests go through a shared inbox nobody monitors daily.
  5. There is no named owner for brand‑approval turnaround, so every request waits on luck.

5Worked application, step by step

Here is the full reasoning for the truck example above, including what to do with the answer once you have it.

Why 1
Why did the truck break down?
The engine overheated and the truck had to pull over.
Why 2
Why did the engine overheat?
The coolant reservoir was nearly empty.
Why 3
Why was the coolant low?
There was a slow leak from a cracked hose.
Why 4
Why did the hose crack without anyone noticing?
That hose model isn't on the fleet's standard maintenance checklist.
Root cause
Why isn't it on the checklist?
The checklist was written for an older truck model and was never updated when the fleet changed vehicles.
Fix
Update the maintenance checklist template for the current fleet, and add a process step so checklists are reviewed whenever a new vehicle model is added. Replacing just the one hose would have fixed this truck, but not the fleet.

6Common mistakes

⚠
Stopping at the first comfortable answer. Two or three whys usually lands on a symptom that feels explanatory but still isn't something you can fix once and be done with.
⚠
Turning "why" into "who." If the chain starts naming people instead of conditions, you've slipped from root‑cause analysis into blame, and people will stop giving you honest answers next time.
⚠
Guessing instead of checking. Each "why" should be an answer you could point to evidence for — a log, a record, a direct observation — not the most plausible‑sounding story.
⚠
Following only one branch. Real problems sometimes fork ("the hose cracked AND no one was monitoring coolant levels"). Treating the chain as strictly linear can hide a second root cause.
⚠
Treating five as a magic number. Five is a habit, not a rule. Some chains find the fixable cause at why three; others need seven. Stop when the next fix would actually prevent recurrence, not when you've counted to five.

7When not to use it

✓
Problems with many interacting causes. If three different teams, two systems, and a timing coincidence all contributed, a single linear chain will oversimplify. Use a fishbone (Ishikawa) diagram or a fault‑tree instead, which can hold several branches at once.
✓
Statistical or variation‑driven problems. "Why is defect rate up 2%?" is usually a data question before it's a narrative question. Run the numbers first — control charts, correlation checks — then use Five Whys on what the data points to.
✓
High‑stakes, novel incidents. A major safety event or large‑scale outage usually deserves a cross‑functional formal review, not one person's five‑minute chain of reasoning.
✓
When the cause is already well understood. If last month's postmortem already found this exact root cause, you don't need to re‑derive it — you need to check why the fix didn't hold.

8Related models

Fishbone (Ishikawa) Diagram
First Principles Thinking
Second‑Order Thinking
Inversion
Chesterton's Fence
Hanlon's Razor

9Try it yourself

Describe a real, recurring problem and get a ready‑to‑use prompt for running a Five Whys session with an AI assistant or in a short team conversation.

Your generated prompt will appear here.
Copied ✓

10Save as skill

Turn this model into something your AI assistant can run on demand — preview it, copy it, or download it as a portable skill file.


        

          Copied ✓
          
        

      

      

        
        
        

          
        

        
Your generated prompt will appear here.

        

          Copied ✓