Teaching & Leadership · Working Reference
Six portable diagnostic tools — a structured way of thinking you can carry across very different situations, each one leaving you with a defensible analytical artifact at the end.
Each instrument below is a portable diagnostic — a structure you run your thinking through so you don't skip the steps you'd otherwise route around. None of them is novel in a textbook sense. What's unusual is using them consistently, together, across every decision context, for years. That's how the substance accumulates.
Read them as a working reference, not a reading assignment. Each is laid out the same way: what it is, when to reach for it, how to actually run it, the failure mode that quietly defeats it, and — in the spirit of citing what came before — its lineage in the established literature.
How these were found
These instruments weren't drawn up on a whiteboard. They were found. Working with Claude, I synthesized roughly 1,500 pages of my own working notebooks — fourteen years of them — and asked a question no one can answer from memory alone: which structures do I actually reach for, again and again, across completely different situations?
Read one notebook at a time, you see a to-do list. Read the whole corpus at once — which only becomes possible when an AI can hold all of it in view simultaneously — and the repeated shapes stand out: the same two-column sort before a hard prioritization, the same paired-list register at every year-end, the same decomposition reflex whenever a number moved. Six recurred with enough consistency to name.
That's the part worth being precise about. The instruments themselves are not original — most map onto established methods in accounting, selection science, and management, cited with each one below. What the synthesis surfaced isn't a new toolkit but a true one: the specific six I run on, and the finding that the leverage isn't in any single tool — it's in using them consistently, together, for years. That's a claim you can't make honestly without the full record to back it, and reading the full record is the part that used to be impossible.
A two-column decision frame separating actions that consume capital, headcount, or political weight from actions that don't. The first column is what you can do right now without spending anything. The second is what would take real resources.
Strategic planning. Turnaround situations. Anytime you face more good ideas than capacity to execute them. Anytime someone says "we just need to do everything." Any review where the question is what do we push first.
Draw two columns. List every plausible action. Sort each into Costless (cadence changes, meeting restructures, redirected time, process tightening, conversations you haven't had) or Cost (hires, capital spend, tech investments, organizational changes). The Costless column gets executed in the next 30 days. The Cost column gets prioritized, scoped, and brought back for a real decision.
A good Costless/Cost analysis followed by no decision. The matrix produces a defensible analysis; it does not produce the decision. Set explicit decision deadlines or the matrix becomes a substitute for the action.
A two-list register: what was accomplished in the period, paired with what didn't get done or didn't work. Used for self-evaluation and for status reporting to a senior counterparty.
Annual reviews to your boss. Quarterly team reviews. After-action reviews of major initiatives. Personal year-end self-assessments. Any time someone asks "how did it go?" and you want to answer honestly rather than performatively.
Write the (+) list first, with sub-items beneath each accomplishment naming what specifically was done. Then write the (−) list with the same discipline — what was missed, what's still open, what didn't work. The (−) list is the harder list and the more credible one. A review with only accomplishments reads as a performance; a review with both reads as a report.
The form is honest but it is not strategic. It tells you what happened; it does not tell you what should happen next year. Pair it with an explicit forward-priority list — three to five things you commit to for the next period — or it becomes a backward-looking artifact.
Decomposing a top-line number across three to five simultaneous axes — typically volume × rate × mix × cost structure × counterparty share. Used to find where a result is actually coming from rather than just whether the number moved.
Any aggregate financial or operational number where the headline alone is uninterpretable. Revenue. Margin. Unit count. Utilization. EBITDA. Customer concentration. Any metric where two months in a row could mean five different things.
Take the top number. Ask: what are the 3–5 things that compose this? Lay them out in a table or in a row of math. Watch each axis move over time, not just the aggregate. The pattern you care about is rarely in the headline; it's almost always in which axis is moving and which isn't.
Decomposition without judgment. Five axes produces five trends; deciding which trend matters requires judgment the decomposition doesn't supply. The instrument tells you where to look; it does not tell you what to do. Pair every decomposition with a forward question: given which axis is moving, what changes?
A fixed, pre-authored question set deployed across multiple counterparties asking the same questions. Used to triangulate among interviewees, not to learn from any one of them.
Hiring panels where multiple people interview the same candidate. Diagnostic situations where no single source is trustworthy. Strategic-articulation work where you need to test how different parts of the organization see the same question. After-action reviews with multiple participants giving independent accounts.
Write the questions before you talk to anyone. Use the same questions in the same order for every interviewee. Take notes by question, not by person — so when you're done you have every answer to question 3 in one place, not everything Sarah said in one place. The comparison across interviewees is the data.
The instrument can substitute for direct evaluation if you let it. Twenty answers to five questions don't decide anything by themselves. Treat it as a triangulation tool, not a decision tool. The decision is downstream of the triangulation; the triangulation clears the ground.
A management-conversation form (bidirectional, four-question). The meeting opens with a fixed agenda of substantive questions; the counterparty is explicitly invited to ask the same kind of question back. The form runs from descriptive ("what is happening") to evaluative ("what is working and what isn't") to forward-looking ("what should change"). Used to operate from substance rather than from authority.
One-on-ones. Performance discussions. Difficult conversations. Hiring and termination conversations. Annual reviews. Customer-issue conversations. Anywhere two people sit down to align on something with operational consequences.
Open with the four questions in some variant of: What can I do to help you? · What do you need from me to be successful? · What is in your way that I should know about? · What should I be doing differently? The questions go both directions — you ask them of the other person; you explicitly invite them to ask the same of you. The bidirectional invitation is what makes it work. One-directional, this is paternalism in question form. Bidirectional, it's operational partnership. Then run the three phases — describe, evaluate, commit — and close with a forward action both parties own.
The bidirectional invitation must be real. If you ask "what can I do to help you?" and don't act on the answer, you'll get fewer answers next time and the methodology silently deteriorates. The form is load-bearing on the trust it builds. The bidirectional question is a commitment, not a conversational move.
Return on equity decomposed through its constituent ratios — typically operating margin × asset turnover × leverage, expanded to four, five, or six factors depending on the variant. The textbook financial-analysis tool, used as an operational diagnostic rather than a financial-reporting one.
Board-level financial discussions where you need to explain why ROE or ROIC moved. Operational-strategy discussions where the question is which lever to pull. Investor or owner conversations about capital allocation. Cross-business comparisons where two units look similar at the top line but operate very differently underneath.
Take the return metric. Write out the multiplicative decomposition. Plug in actuals for each factor over multiple periods. Watch which factor is doing the work and which is dragging. When you talk about the result, talk about the factors — margin held, turnover fell, leverage masked the margin decline — not the headline number.
Sophistication without audience. The full six-factor DuPont is wrong for most operational audiences; they need three factors at most. Match the depth to the listener. A frontline manager needs margin and turnover; a CFO can handle the full decomposition; the owner cares about whichever factor is changing.
Pushed far enough, DuPont stops being a diagnostic and becomes a management tool. Keep decomposing past the financial factors: margin and turnover give way to operational inputs — in equipment rental, to utilization — and those give way to things a named person actually does, like closing a preventive-maintenance work order on time or returning a quote within a day.
The stopping rule is simple: keep asking "what feeds it?" until the answer is no longer a number but a behavior someone owns. Above that line is math; below it is management. It's also the accountability line — an input you can only move by doing the real work is one no one can game. Manage those leading inputs daily and the lagging outputs — ROI, net income — follow. You can't manage a number no one has a dial for.
Adapted from "Managing Inputs, Not Just Outcomes," presented at IMA's 2026 Annual Conference.
Not six tools you pick from — one practice
The discipline isn't using each instrument in isolation. It's having them all available, and choosing which one fits the situation in front of you. The instruments don't think for you; they keep you from skipping steps you'd otherwise skip. That's why they compound over years.
In the spirit of citing what came before: none of these six instruments is claimed as an invention. Each has clear antecedents in accounting, selection science, coaching, or military practice — noted with each instrument above and listed below. Where I can no longer trace whether I built a habit myself or absorbed it from something I read years ago, I've credited the established method rather than claim it.
What is original is narrower, and I think more useful: the finding — surfaced by synthesizing a fourteen-year documentary record with AI — that these particular six form a coherent operating set, and that the compounding comes from consistent, combined, multi-year use rather than from any single tool. The tools are borrowed; the practice, and the evidence that it holds, is the contribution.