What Deliverables Signal Which Level

"He seems capable" is not an evaluation. Look only at the materials left behind and the actions taken while making them. From that visible evidence, place the creator somewhere between L1 and L4. This article builds that mapping.

Think in driver's-license stages

Driving skill cannot be measured by talk. So a license splits stages by "moves you actually performed" - hill starts, parking. A learner's permit, a full license, and someone who has driven years without an accident. All say "I can drive," yet the visible behavior behind it differs completely.

Making materials is the same. A self-report like "I check my sources properly" tells us nothing about whether the person is L1 or L4. The only material for judgment is the deliverables that actually appeared and the actions taken along the way. What this article builds is that "mapping of what was seen to which level" - an anchor. An anchor is a marker that pins a vague impression to one fixed point.

There are four stages, L1-L4. L1 = finishes only the one assigned case as told. L2 = has learned the pattern and can reproduce it. L3 = understands why and can apply it elsewhere. L4 = can design the system itself and set a workplace standard. We translate these four into "what must be visible at each level" for each skill.

The same job splits by visible behavior

Picture a kitchen apprentice. Making one dish by the recipe is L1. Producing the same dish reliably with memorized amounts is L2. Understanding why the salt goes in first and applying it to other ingredients is L3. Writing the kitchen's procedure so a newcomer hits the same taste is L4. Not just the plate, but the person's movements tell you where they stand.

In material creation, take the most important floor: "can you return to the source" (grounding). An L1 person's deliverable may miss the cited page number, or mix in second-hand citation (using someone's summary in place of the original). At L2 the source column is complete and the original can be produced on request. L3 reads the source's limits - target patients, observation period, sample size - and adjusts the wording accordingly. L4 turns the source-checking procedure itself into a checklist so the whole team stands on the same floor.

Flip a deviation case and the level boundary appears

Recall a health checkup. The way abnormal values show up reveals where the body is under strain. Deviation cases work the same way: flip what was done, and you see the level the person was stuck at.

In a reported case, the creator prepared no material for the primary endpoint (the result the drug most wanted to confirm) and explained only a secondary result that showed significance. This is the "learned the pattern but does not understand why the primary endpoint comes first" stage - an L2-level omission. The psychological driver of protecting oneself by not speaking overlaps here.

In another case, the creator said "a difference is seen" for a small Japanese subgroup where none existed, and when challenged, deflected to authority: "the professor says it's fine too." This combines weak application with externalizing responsibility. To claim L3, one must explain the weakness of the subgroup numbers in one's own words. The table below maps cases, levels, and the skill that should stop them.

The third, the y-axis manipulation, deserves caution. The design skill to make a graph stand out is there. But fidelity to fact is not. This is what this section calls most dangerous - "high design x low fidelity = persuasive misreading" - and far from being high-level, the floor is cracked. Skill in appeal does not fill a hole in grounding.

Levels may differ by skill

When several people proof a galley (a test print before printing), one is strong on typos, one on phrasing, one on checking numbers. Even within one person, strengths and weaknesses split. The eight skills of material creation are the same: they almost never line up at one level.

For example, someone may be L3 in appeal design (making things easy to grasp) but L2 in source grounding. Looking at the total score alone, this looks "so-so." But in this section's pass logic, a floor like grounding is a non-compensable gate. Non-compensable means high scores elsewhere cannot fill the hole. If the floor is cracked, no amount of appeal skill earns a pass.

Lay out the visible deliverables by skill, and first confirm the floor (grounding) is not missing. If the floor is missing, do not use the height of other skills to raise the ranking. Only after the floor is met do you evaluate overall height as excellence.

So the evaluation sheet is not one number but a row of levels per skill. Shown as "grounding L2 / balance L3 / appeal L3 / self-review L2," the person can see where to grow to reach the next stage. This is not a table for condemnation. It is a map for the person to self-monitor which of their own circuits is weak.

How to use the anchor table, and its pitfalls

At airport baggage screening, sample photos are posted - "stop if this shows up" - so judgment does not drift between officers. The anchor table plays the same role: a sample so judgment lines up even when the evaluator changes. But applying the sample mechanically can cause misjudgment instead.

There are three pitfalls. First, a good deliverable may just mean the person happened to get a good request and good material. So line up several jobs, not one, and check whether output is stable (the core of L2). Second, behavioral evidence may not remain. If a source check was done but unrecorded, a third party cannot confirm grounding. Keeping records is itself evidence of L3-L4. Third, an excellent device not on the table may be undervalued simply because it is not on the table. The anchor fixes the floor and the typical; it does not cap excellence.

The next article, the eighth, addresses how far we can trust the level we assigned this way - whether a different person seeing the same deliverable lands on the same level, the question of judgment reliability.

Level judgment uses as material only the deliverables left behind and the actions taken while making them - not intelligence or enthusiasm. The L1-L4 anchor table is a sample that aligns judgment even when the evaluator changes, and at the same time a map for the person to find their own weak circuit.

The order must not be forgotten. First confirm the grounding floor is not missing; if it is cracked, no amount of design skill earns a pass. Only after the floor is met do you lay out per-skill levels to measure excellence. The next article asks how far this very judgment can be trusted.