Your Job Descriptions Are Rubrics in Disguise. Almost Nobody Grades Against Them.

Go pull the job description for your intake or customer service position. Somewhere in it, you wrote things like "builds rapport quickly," "explains our process clearly," "follows all required disclosures," "moves the caller to a concrete next step." Now ask one uncomfortable question: when was the last time anyone was actually graded against that document?

For most businesses, the honest answer is the day they were hired, and never again. The job description gets a starring role in recruiting, then retires to a folder. Which is strange, because you already did the hard part: you wrote down, in plain language, what good sounds like.

The standard exists. The measurement doesn't.

Here's the bottom line: most companies don't have a performance-standards problem, they have a measurement gap. Leadership can describe a great call in detail. Supervisors know one when they hear one. The job posting literally lists the criteria. What's missing is the machinery that compares the eight hundred calls your team took this month against the standard everyone already agrees on.

Without that machinery, evaluation defaults to impressions. The rep who's confident in meetings gets graded up. The quiet one who quietly follows every step gets overlooked. And the criteria drift: what gets praised this quarter depends on what the manager happened to overhear.

Turning a job description into a rubric takes about an hour

The conversion is more mechanical than people expect. Take an example from a consumer-facing intake role:

• "Follows all required disclosures" becomes an auto-fail rule: miss the disclosure, fail the call, no matter how pleasant the rest was. Regulators grade this way, so your rubric should too.
• "Explains our process clearly" becomes a weighted category, say 20 points, scored on whether the caller's questions got answered without jargon.
• "Moves callers to a concrete next step" becomes 15 points: was an appointment set, a document sent, a callback scheduled? Yes or no is audible on the tape.
• "Builds rapport" becomes a smaller weight, maybe 10 points, because you care about it, but not as much as the disclosure.

Add a pass mark and a rule for borderline calls (ours route to "needs review" so a human makes the judgment call), and the document you wrote for recruiting is now an operating standard. The standard is yours. The grading just finally matches it.

The one-hour build, step by step

Here's the actual working session, because "we build a rubric together" shouldn't be mysterious. Bring the job description and, if you have them, your script and your required disclosures. First pass, fifteen minutes: go line by line and sort every stated expectation into three buckets: must-happen-every-call (candidates for auto-fail), should-happen-and-varies (weighted categories), and nice-to-have (small weights or drop entirely). Second pass, fifteen minutes: for each weighted category, write the one-sentence definition of what earns full marks, phrased so that two different people listening to the same call would score it the same way. "Explains the process clearly" becomes "caller's questions were answered without jargon and the caller was not asked to repeat information they'd already given." Vague criteria produce arguable scores; arguable scores produce arguments.

Third pass, twenty minutes: set the weights by asking one question per category: "if a rep did everything else perfectly but missed this, how upset would I be?" Your gut answer, in dollars or in blood pressure, is the weight. Set the pass mark last, and set it slightly lower than your instinct suggests, because the first weeks of real scores will recalibrate everyone's sense of what normal looks like. Total elapsed time: about an hour, and most of it is decisions you've already made implicitly a hundred times. The session just writes them down.

What the first scored month typically surfaces

Take an example: a consumer firm scores its intake position for thirty days against a rubric built exactly this way. Three findings show up so reliably we'd almost promise them.

First, the criteria don't weigh what leadership assumed. The rep everyone considered the closer scores middling on "concrete next step" because half her charm-heavy calls end warmly and vaguely. The quiet rep nobody discusses turns out to run the cleanest disclosure record on the floor. Impressions and measurements disagree, and now you know by how much.

Second, one criterion is failing everywhere, which means it's not a people problem. When eight of ten reps score low on "explains the fee structure clearly," the script is the problem, or the training is, and no amount of individual coaching would have fixed it. A census finds systemic issues that per-rep sampling structurally can't, because sampling frames every finding as being about a person.

Third, the rubric itself needs a revision. Some criterion turns out to be unmeasurable as written, or two criteria overlap, or the pass mark was set on optimism. Good. The rubric is an operating document now, and operating documents get revised. Expect to tune weights in the first month and rarely after.

Why this beats a generic "call quality score"

Plenty of tools will hand you a canned quality score built on somebody else's idea of a good call: talk-to-listen ratio, sentiment, politeness. Those numbers aren't useless, but think about defending one in a coaching conversation, or worse, in a dispute. "The algorithm says 62" convinces nobody. "You skipped the fee disclosure on Tuesday's 2:14 call, here's the clip, and it's an auto-fail because it's in your job description" ends the argument before it starts.

Position-based rubrics also mean sales gets graded as sales and service gets graded as service. A collections call and a new-client consult should not be scored on the same sheet, and with rubrics per position, they aren't.

Where a rubric is the wrong tool

Two honest exclusions before you try this. If a role's real output isn't conversational (a records clerk, a billing specialist), scoring their rare phone calls tells you almost nothing about their job. Measure the work product instead. And if a position is genuinely new, with no settled process, resist the urge to grade it. A rubric enforces a standard; it can't invent one. Run the position for a quarter, learn what good looks like, write it down, then grade. Scoring an undefined job just makes the confusion numerical.

The part your team will actually like

Reps are rightly suspicious of surveillance. What changes their mind is fairness: the criteria are public, they're the same ones in the job posting, the weights are known, and the score comes with the clip attached. Nobody gets dinged on a manager's mood. Your best people tend to become the system's biggest fans, because for the first time the numbers prove what they've been doing all along. And your AI employees, if you have them, sit on the same roster with the same rubric, which keeps everyone honest, including the software.

Rollout order matters more than most owners expect. Publish the rubric to the team before the first call is scored, not after, so nobody experiences it as a trap. Run the first two weeks as calibration: scores accumulate, nobody's comp or standing moves, and reps can challenge a score and watch the clip with their manager. Then turn it on for real. Firms that skip the calibration window spend the next quarter arguing about the instrument. Firms that run it spend the next quarter coaching, which was the point.

Try it on one position, free

This is exactly what our free people QA preview does: you pick one position, we turn its real job description into a rubric together, and we score real calls so you can see your own standard enforced for the first time. Request a working session, or read more about how the Voice AI Dashboard works.