The Method

How to read where a person stands, to the standard of a PIP, by hand and for any role. This is what /pip and /performance-review do, written out so a human can run it without AI — and check the AI when they use it.

The read is the hard part: turning a quarter into a defensible here is where they stand, here is why, here is the move. The target, the instrument, and the write-up all follow from getting it right. Eight steps. But first, the two things you score against — the parts a philosophy leaves out and a PIP never does: what each level means, and what to look at.

The Performance Management page is the why and the map. This is the how.

What each level means

Level is a unit of leverage, not a title or years served. It answers one question: how much can you hand them before they need structure? That makes it role-agnostic — it reads the same for an engineer, a PM, or a marketer.

As the level rises, you provide less. The whole read is one comparison: is their demonstrated leverage below, at, or above the bar for their title? (Adapt the rungs to your own rubric; the shape holds — task → problem → direction → team → system.)

What to look at

You do not score a feeling; you score named competencies against the level bar. The dimensions change with the seat; the discipline does not. Use the role’s published rubric.

• Engineer — Product Engineer: Product Sense · Critical Thinking · Engineering Craft · Delivery & Operations · Communication · Ownership & Drive · Leverage.
• PM — Product Manager: Product Sense · User Insight · Strategy & Prioritization · Delivery · Data & Impact · Communication · Influence.
• The shared core (any role) — does it move the right outcome (judgment), do they get it done (delivery), are they understood (communication), do they drive without being pushed (ownership), do they make others better (leverage).
• No rubric yet (marketing, ops, design) — write the five to seven competencies that matter for the seat before you review anyone. Without a named bar, every read is a vibe. Name it first, in writing, and hold everyone in the seat to it.

For each dimension, “good” is what the rubric says at their level. That sentence is the bar you score against — nothing looser.

The eight steps

1. Gather the evidence

Pull the period’s real, dated artifacts before you form a view: shipped work, incidents, decisions, documents, what others said in the moment. Dates and specifics, not impressions. If you are reconstructing from memory of last week, stop — that is recency bias, and it is how a review goes wrong before it starts. Log as you go through the quarter and this step is already done.

2. Score each competency against the bar

For every load-bearing dimension, mark below / at / above the level bar, and attach one dated example. No example, no score — a claim without a date is a bias tell, not a finding.

3. Roll up to a state

The state is the floor of the load-bearing dimensions — not the average, not the peak. A lagging dimension that matters holds them below even with a spike elsewhere; a spike above the bar does not raise the level, it earns scope. One word comes out: under, meeting, or over their bar.

4. Diagnose

If they are below the bar, name why — the four gaps:

• Clarity — does not know what good looks like.
• Skill — knows, cannot hit it yet.
• Blocker — knows and can, something outside them is in the way.
• Will — knows, can, nothing blocks them, still not driving.

And separate two things a title hides: someone performing fine at a lower level than their title is mis-leveled, not failing — a different fix (reset the title, not a plan to exit).

5. Convert to targets — outcomes, not activity

Each gap becomes a target with four parts: a deliverable, a number, a date, and a reason. And the deliverable is an outcome, never an activity — the trap that makes a target meaningless:

• Engineer: “reviews three PRs a week” is activity → “no work they approve is bounced by QA” is the outcome.
• PM: “does more customer calls” is activity → “names the top three user problems, with evidence” is the outcome.

If a trait is not measurable (“thinks about the why,” “sounds like excuses”), convert it to a gradable proxy (“states the problem and the tradeoff before merge”; “on a slip, names the controllable cause and a recovery step”). If you cannot grade it weekly, the final score is contestable.

6. Pick the instrument

• Mentor — clarity, skill, or blocker. Coach it, raise the cadence; not performance management.
• Re-level — mis-leveling. Reset the title with a development plan.
• PIP — will, or genuine perform-or-exit. Only when the honest question is “perform at this bar or leave.” Name the honest floor: clearing a lower bar re-levels, it does not by itself mean exit.
• Promote / scope-up — over the bar. Build the case on dated evidence mapped to the next level’s bar.

7. Pressure-test it

Before it is final, argue against yourself. This is the step that separates a defensible read from a plausible one:

• Every claim carries a dated example. (“They tend to…” fails.)
• Read the whole period, not the last two weeks.
• Each dimension scored on its own evidence — no halo from one spike, no horns from one miss.
• Strengths named, not erased. If it reads “not capable” while they clear real competencies, it is wrong — narrow it.
• No team-wide gap held against one person. If most of the team misses it, it is not personal.
• Consistent with peers on the same profile, or you can say exactly why not.
• If below the bar: is this genuinely perform-or-exit, or mis-leveling dressed as a PIP? Honest floor named.
• A second reader, given the same evidence, would land on the same state.

8. Write it up

The read in one sentence. The standing table. The diagnosis, if under. The strengths. The instrument and cadence. The target(s). Direction in two sentences they can repeat without notes. Every finding with an owner and a date. Lean for a healthy person; the full weighted appendix only for a formal case.

Worked — three roles

The same method, three seats, so it is plainly not an engineering trick.

Engineer — a Mid, meeting the bar

Roll up: floor is at title, trend climbing. Meeting. Diagnosis: healthy; the one lagging dimension, Leverage, is the next rung. Instrument + target: direction, no formal instrument. The next hire ships to prod independently by day 30 and no PR they approve is bounced by QA this quarter — Leverage as the team moving faster, not a review count.

PM — a Senior, operating below

Roll up: floor is below title, flat. Under. Diagnosis: not comprehension — the strategy is strong. A delivery and measurement gap (skill, edging into will on surfacing risk). Instrument + target: mentor, weekly cadence. Every committed launch lands on its date or is re-scoped three or more days early; each ship reports its metric within a week. Reason: strategy that does not ship, measured, is a document.

Marketing — no rubric yet, a growth marketer

No published ladder for the seat, so the bar was written first: Audience & Positioning · Channel Execution · Data & Attribution · Creative · Experimentation.

Roll up: floor is below the written bar. Under (against the bar you set). Diagnosis: skill on measurement and experimentation; execution itself is fine. Instrument + target: mentor. Every channel reports CAC and attribution weekly; two concluded experiments a month, each ending in a decision. Reason: without attribution, spend is a guess.

Related: Performance Management (the why and the map) · /pip · How I Run