Split vanity metrics from decision metrics
Give the Admin Agent a **classification rubric** and your metric inventory so it tags **vanity** (moves without forcing a decision) vs **decision** (moves budget, scope, or calendar). Export a one-page **exec decision deck** and one **vanity quarantine** list for reporting polish only.
Plus: three prompts — **classify every line on a dashboard** with a kill-test question, **upgrade one vanity metric** into a decision proxy with threshold plus owner, and **invert usage** so the agent argues why a favored metric should **not** decide spend.
| Audience | Everyone · Executives |
|---|---|
| Time | ~12 min |
| Prerequisites | You can attach context files and create agents ([Add your first context file](/tutorials/add-your-first-context-file), [Create an agent from scratch](/tutorials/create-an-agent-from-scratch)). Helpful when metrics tie to portfolio choices: [Track strategic decisions and when to revisit them](/tutorials/track-strategic-decisions-and-when-to-revisit-them), [Run a quarterly initiative audit from scattered notes](/tutorials/run-a-quarterly-initiative-audit-from-notes). |
| You'll end up with | A **Metric classifier** agent (or documented rubric in chat), a labeled table sorting your current metrics list into **vanity** vs **decision**, and at least one **decision proxy** row with **threshold**, **owner**, and **what we do if it crosses**. |
When a tutorial shows italic text in quotation marks, it usually mirrors a label or helper string inside Auxot. Product copy changes between releases — if something reads differently in your workspace, trust what you see on screen.
Callouts with a Worth knowing gold accent are meant as must-read context before you move on. Blockquotes that open with Tip are lighter, optional depth.
Why this matters
Dashboards seduce everyone the same way: big green deltas feel like leadership. Vanity metrics answer did we look busy? Decision metrics answer what do we fund, pause, or kill next month? Mixing them in one slide deck trains the room to applaud motion instead of choices.
Framing the split as recipes, not culture policing, keeps the spine safe for executives: you are not judging people. You are labeling numbers by whether they clear a decision bar. Auxot holds the rubric and re-classifies when you paste a new export. You still choose thresholds and owners.
Today you wire one classifier loop. The next time finance asks what actually drove the outcome, you point at a short decision-only list instead of re-litigating every chart.
The agent reads exports and definitions you attach. It does not pull numbers from live BI tools unless you paste them.
Quick start
- Paste your metric inventory into a context file or chat: names, definitions, refresh cadence, and where each number appears today (board deck, weekly email, all-hands). Redact customer names if needed.
- Open Chat with the Admin Agent — ask for a Metric classifier agent whose charter uses one decision test only: If this metric moved 10 percent in either direction next Friday, would we change budget, headcount, scope, or meeting cadence within two weeks without another study? Yes → decision candidate. No → vanity unless upgraded.
- Run the classify prompt from The agent can do that? → 1 below. Export the agent’s markdown table into a context file called something like
Metric labels. - Pick one vanity row you still love — run The agent can do that? → 2 to propose a decision proxy (tighter definition, segment, paired metric, or threshold).
- Share the decision-only short list — five bullets max for your next leadership read. Leave vanity charts in appendix pages labeled orientation, not choices.
Done? At least three metrics tagged decision with named owners in the table. If every row stayed vanity, tighten definitions until something clears the bar or admit no metric currently drives spend (that admission is useful).
The agent can do that?
1. Classify a whole dashboard in one pass
Attached: pasted metrics table or bullet list (metric name + definition). Rules: (1) DECISION if the decision test passes (would change budget, headcount, scope, or cadence within two weeks if it moved 10 percent Friday to Friday). (2) VANITY if it fails unless paired metric supplied. (3) UNKNOWN if definition too fuzzy to test — flag **needs definition owner**. Output markdown table: Metric | Tag | Why | Missing pair if vanity | First upgrade experiment (one sentence). No motivational language.
Why it’s non-obvious: Teams argue taste (“this matters to morale”). The single test forces the same bar for every line because you wrote the rule once.
2. Upgrade one vanity metric into a decision proxy
Metric labeled VANITY: [name]. Business outcome we actually want: [one sentence]. Propose exactly one upgraded metric or pair (Metric A + Metric B) that could pass the decision test. Include: threshold band, data owner, failure mode if crossed, and what we stop funding if it stays red two periods. If you cannot upgrade without new instrumentation, say **cannot upgrade yet** and name the smallest measurement increment to buy.
Why it’s non-obvious: Vanity survives because it correlates loosely with success. The upgrade step asks for operational consequences, not a stronger chart title.
3. Devil’s advocate on a favorite number
Inverted usage. Pick the metric leadership celebrates most:
Metric: [name]. Argue three concrete ways this metric could improve while the underlying business gets worse — real scenarios, no cynicism for its own sake — end with whether it still deserves DECISION tag after those scenarios — markdown bullets — max six lines total.
Why it’s non-obvious: Exec decks optimize for uplift narratives. Forcing good metric, bad business scenarios exposes proxies early before board prep repeats them.
Go deeper
Definitions beat tools
If BI renamed fields last quarter, stale definitions poison classification. Keep a definitions snippet beside the metric table in the same context file. Refresh dates belong in the file header.
Portfolio bridges
When metrics tie to strategic decisions or initiatives, pair this exercise with Track strategic decisions and when to revisit them or Run a quarterly initiative audit from scattered notes so numbers attach to choices, not gut feel.
Cadence
If you run a weekly exec rhythm, Set up your Monday morning briefing can surface decision metrics only so vanity charts stop stealing Monday oxygen.
Walkthrough
Step 1: Freeze the decision test
Copy the decision test from Quick start step 2 into the agent charter verbatim. Optional second sentence: UNKNOWN rows require a named definition owner before tagging.
Step 2: Paste messy reality
Export does not need to be pretty. Paste CSV fragments, slide bullets, or objectives lines. Run The agent can do that? → 1.
Step 3: Resolve UNKNOWN first
Anything needs definition owner blocks classification. Assign owners in your planning forum, update definitions, rerun only those rows.
Step 4: Upgrade or quarantine
For each persistent VANITY row leadership still wants visible, either run The agent can do that? → 2 or move it to a Reporting polish appendix with no decision claims in the title.
Step 5: Publish the short list
Decision metrics list caps at five for verbal reads. Everything else is appendix or deep link.
Next steps
- → Recover vanished meeting commitments with owners and drafts you send when the metric story needs row-level owners and sends — not only classification labels.
- → Track strategic decisions and when to revisit them when metrics should tie to ideas shelved or revived, not just quarterly variance.
- → Run a quarterly initiative audit from scattered notes when you need portfolio-level oxygen checks beside metric hygiene.
- → Take Auxot’s pulse in 10 seconds if you want the same honest snapshot habit at product level before you scale the executive rubric wider.