What a Competency Checklist Should Actually Contain

Why this matters

Most shop competency checklists are lists of topics with a box next to each one. "Electrical basics." "Customer communication." "Safety." Someone ticks the boxes at the end of a training period, everybody signs, and the document goes in a folder. Then a tech signed off on "electrical basics" makes a call that shows they were never actually verified on anything, and the checklist turns out to be evidence of nothing except that a meeting happened.

A competency checklist is worth building only if a line on it means something specific enough that two different leads would sign it the same way. That property comes from the anatomy of each line, not from the topics you choose. This card is about the line, not the list. Which competencies to track, and how to lay them out as a grid, have their own articles. Get the line right first, because a hundred badly written lines is not a training system, it is paperwork.

The four parts of a verifiable item

Every line needs all four. Missing any one of them and the item cannot be signed honestly.

1. An observable action. Something a person visibly does, expressed as a verb you could film. "Isolates and verifies the circuit is dead before working on it" is observable. "Understands electrical safety" is not, because nobody can watch understanding.

2. The condition it is performed under. Where, with what, and under what constraints. The same action is a different competency in a clean shop bay than on a wet slab in a customer's basement with the customer watching. If the condition is not stated, everyone assumes the easiest one.

3. The standard that counts as pass. The specific reading, sequence, tolerance, timeframe, or outcome that separates done from not done. This is the part most often missing and it is the reason two leads sign the same item differently.

4. The evidence the signer looked at. What the verifier actually observed or read. Direct observation, a completed ticket, a photo of the finished work, a teach-back. Without this, sign-off drifts to "I think they can do it," which is a reputation, not a verification.

Bad item against good item

Read the rewrites for the pattern rather than the wording. Each one names an action, a condition, a pass standard, and a countable evidence base. Each one also takes longer to sign, which is the point: a line that can be signed in two seconds was not verifying anything.

The verbs to avoid

Any item beginning with these cannot be verified: understands, knows, is familiar with, is aware of, appreciates, has been trained on, has completed. The last two are especially seductive because they are true and feel like evidence. A tech who sat through the walkthrough has been trained on the procedure and may still be unable to perform it.

Replace them with verbs that describe a performance: performs, isolates, verifies, measures, selects, states, diagnoses, corrects, records, escalates, teaches.

What the levels have to mean

A binary tick loses too much information. A tech who has watched a procedure and a tech who can teach it both get a checkmark and the grid cannot tell them apart. Use levels, and define them so tightly that a lead cannot slide one.

Level 3 to level 4 is the only transition requiring a formal act, and it is the one shops blur. At level 3 someone qualified was standing there. Sign-off at level 4 means the shop is willing for this person to do this work with nobody watching, and that is a decision with liability attached, so it gets a name and a date on it.

Level 5 is what lets a checklist scale past one lead. A shop where only the owner can sign anything has a training system that stops the week the owner is on a roof.

Evidence: what counts, and who is allowed to sign

Acceptable evidence, in descending order of strength:

• Direct observation on a live job, by someone at level 5 for that item. Strongest, and the only acceptable evidence for anything with a hazard attached.
• The completed work record, checked against the standard. Good for documentation, catalog selection, and readings, where the artifact carries the proof.
• Teach-back. The tech explains the procedure and the reasoning to someone else and answers questions. Catches the memorized-sequence case that observation on one easy job can miss.
• A simulated run in the shop. Acceptable for procedures that are rare in the field, weaker than the others, and marked as such on the record, because shop conditions remove most of what makes the field version hard.

Not evidence: attendance, elapsed time in the role, a certificate for a course, the absence of complaints, or another tech's general opinion.

Restrict signing to level 5 for that specific item. A lead who is a strong installer signing off a diagnostic competency they have never been assessed on is how a checklist quietly becomes fiction. And never let a tech sign their own line: self-assessment on a skill you are still learning is unreliable in both directions, the careful ones underrate themselves and the confident ones do the opposite, and neither error is visible on the form.

Recency, expiry, and the items that decay

Some competencies decay and some do not. Mark which is which on the item itself.

• Skills used weekly need no expiry. The work is the verification.
• Skills used a few times a year need a recency note. An item last verified 14 months ago on a job type the tech has not touched since is not a current level 4, and dispatching on it is how you get a bad outcome with a signed form saying it should have been fine.
• Safety and code items carry a re-verification interval set by the standard governing that item, not by a blanket shop rule. Many are annual, but not all: powered industrial truck operator evaluation runs every three years (29 CFR 1910.178(l)(4)(iii)), and some retraining is event-triggered rather than clock-triggered. Adopting "annual" across the board over-trains some items and leaves you non-compliant on others. Any item tied to a certification expires with the certification.

That is one extra column: last verified date. The grid then answers "who can do this" and "is that still true," which are different questions and only the second one prevents an incident.

What does not belong on the list

Scope discipline keeps the list alive. A checklist with 90 lines gets abandoned within a year; one with 20 well-written lines gets used.

Leave off: anything you would never dispatch on, personality traits, attitude, and anything the shop cannot actually verify. If nobody in the shop is competent to assess an item, it belongs on the shop's own gap list, not on the checklist. Leave off what everyone can already do: an item where every tech is at level 4 costs upkeep and answers nothing.

A worked example: rewriting a legacy checklist

Take a real-shaped legacy list of 6 topic lines: safety, tools, diagnostics, repairs, paperwork, customer service. Six ticks, one signature, one date, applied to every tech.

What it can answer. Whether a form was completed. Nothing else. Dispatch cannot use it, because "diagnostics: tick" does not say which faults on which job types. Training cannot use it, because there is no gap smaller than an entire topic.

The rewrite. Each topic breaks into the specific performances the shop actually dispatches on, and the 6 topics become 19 items. The distribution comes out roughly: 4 safety items, 2 tool items, 6 diagnostic items split by job type, 3 repair items, 2 documentation items, 2 customer-facing items.

Note where the expansion concentrated. Diagnostics went from 1 line to 6 because it is where the dispatch decisions live, and safety went from 1 to 4 because each stop-list category needs its own verification. Paperwork stayed small because two well-written items cover it. The expansion is not uniform and should not be, and a list that expanded evenly across topics was probably built by splitting words rather than by asking what the shop actually needs to know.

What it costs to run. Nineteen items across a crew of 5 is 95 cells. Most fill in from what the leads already know, in one sitting of an hour or two. The ones that do not are the interesting ones, and there are fewer than you expect: in a shop this size, typically 10 to 20 cells where nobody can honestly say. Those become the verification backlog, and at 2 to 3 verifications a month riding on jobs already scheduled, it clears in about half a year.

What it returns before any verification happens. Read the finished grid both ways. By row, any item where only one tech is at level 4 is a single point of failure, and in a 5-tech shop with 19 items it is normal to find 3 or 4. Those are your cross-training priorities, now named rather than suspected. By column, any tech at level 4 on fewer than half the items is either new or plateaued, and which items tells you which.

The failure version. The same 19 items, signed off in a single afternoon meeting with no evidence behind any of them. The form now says every tech is level 4 on everything, dispatch trusts it, and the first time it is wrong the shop discovers that a signed checklist with no evidence behind it is worse than no checklist, because it caused a decision that would not otherwise have been made. The tell is the dates: if a whole column signed off on the same day, no verification happened.

How to verify your checklist is doing its job

• The two-signer test. Hand one item and one tech to two different qualified signers, independently. If they reach different levels, the item's standard is not specific enough. Fix the item, not the signers.
• The dispatch test. Ask whoever assigns work whether they consult it. If dispatch runs off memory and the grid lives in a folder, the items are either wrong-grained or not trusted, and both are fixable but neither fixes itself.
• The stranger test. Show one item to someone who does not work at your shop and ask what they would have to see to sign it. If they cannot say, neither can your leads, they are just filling in what they already believe.
• Check the dates, not the ticks. A grid where most items were last verified over a year ago is a snapshot of a shop that no longer exists. The ticks are the easy part to keep; the dates are what make them mean anything.

References

• See related: Build a Skills Matrix: Who Can Do What, Building a Skills Checklist for the First 90 Days
• See related: The New Skill Introduction SOP, Checklist vs SOP vs Training Decision Matrix
• U.S. Department of Labor, competency-based apprenticeship standards
• OSHA requirements for documented training and demonstrated proficiency on hazardous tasks