Tool 16 of 32 · about an hour
Leverage Layer
The shallowest level at which a fix stops the symptom coming back. That layer is the stop line, and it makes the cost question answerable.
Third check in the Depth working session, after validity and compounding, with the stop line written before stability is scored.
What it is for
To find the shallowest layer at which a fix stops the symptom from coming back, write that layer down as the stop line, and make how much to spend an answerable question.
The failure it catches
Fixing shallower than the stop line means the symptom returns. Fixing deeper than it means spending money on depth the problem does not require. Both are how companies either patch forever or repaint the whole car over a curbed rim.
Why it works
Faults that push complexity into compounding almost always live in the deeper layers, which is why fixing the top ones keeps failing. Naming the shallowest layer that actually stops the return, then stopping, is what separates leverage from ego wearing engineering clothes.
How to run it
- Trace the problem through five layers, top to bottom: symptoms, behaviour, workflows and business logic, data models and state machines, business truth, meaning what the domain actually is.
- Ask: what is the shallowest layer at which fixing this stops the symptom from coming back.
- Write the stop line in the room: we fix to here and no further. Also write what tells you to stop: complexity stops increasing, bug categories not just bugs start disappearing, predictability returns, cognitive load drops, edge-case fights end.
- Watch for stop-digging signals, any one of which means too deep: fixing deeper layers no longer simplifies upper ones; refactor scope grows every week; a week of work buys no predictability; bug categories are not shrinking.
- The chosen layer and the stop line go down in writing before the Stability Test runs, because that test scores the moves that implement the fix to this line.
How to read the result
That layer is the stop line. Fixing shallower means the symptom returns. Fixing deeper means spending money the problem does not require. The stop line is what makes how much should we spend on this an answerable question. Past the stop line, digging is ego or perfectionism in engineering clothes.
How to run it well
- Write the stop line before anyone scores moves. Stability without a stop line scores a plan that does not exist yet.
- Discomfort at the real layer is usually the signal, not a reason to stop shallow.
Where it breaks
- Avoiding the real layer because it feels like blame, fix the surface instead of the model. The discomfort is the signal, not a reason to stop shallow.
- Digging past the stop line because deeper feels more thorough. Once compounding stops, deeper is ego.
What would prove it wrong
A fix at the identified layer that does not stop the compounding means the layer was misidentified.
What we watch
Whether fixes at the written layer stop the symptom returning, or whether the next month's recurrences say the layer was wrong.
Ask yourself
- At which layer does a fix actually stop this symptom coming back: symptoms, behaviour, workflows, data models, or the business truth underneath?
- If you fix at that layer, what tells you to stop?
- Is a week of work at the current depth buying predictability, or only more scope?
- Are bug categories shrinking, or just bugs?
- Which layer would be uncomfortable to name out loud?
Fed by: Convergence Test, Patch Validity Test, Compounding Check
Feeds: Stability Test, Definiteness Scoring