Employees

Every ratio is a ratio of sums. Gross per retail unit is total gross divided by total retail units at the grain being reported — never an average of daily figures, per-person figures or store figures, which are different numbers and all of them wrong.

Each figure has its own sample, and its own floor verdict. Gross per unit is governed by retail units; contact rate by valid assigned leads; appointment-set rate by contacted leads, never by all valid ones; show rate by eligible appointments on the scheduled date; show-to-sale by appointments shown on the show date. One person can be comparison-eligible on one figure and not on another in the same period, and the page says so per figure.

Four absences, four statements. “Not applicable” means the measure does not belong to the role. “Insufficient sample” means it does and the denominator is below the floor. “No data” means it does and nothing was observed. A zero is a real observed value and is none of those three.

History keeps its own store and title. Every row’s role, store and tenure band are the values that were true when the activity happened, taken from the employee version the transaction points at. A transfer or a promotion later does not move earlier activity to the new store or relabel it with the new title.

Certified units are used units. The certified count shown beside a mix is a subset of the used count, not a third category, and adding it to used would double count.

Cash deals are inside the finance denominators. Reserve and back gross per retail unit divide by every retail delivery, including cash deals, which cannot generate reserve. A different cash mix moves both figures for reasons unrelated to the finance office, which is why the structure mix is drawn beside them on the row.

Appointments and leads are different populations. One lead can produce several appointments, so the lead-grain rates and the appointment-grain rates do not share a denominator, and the BDC row draws them as two separate bands for that reason. Show rate is on the scheduled date and excludes appointments cancelled in advance — an exclusion a store can game, which is why the cancellation count is on the row beside it. Show-to-sale is on the show date, so period-to-date conversion improves as the data matures.

A response time nobody answered is not a fast one. A never-responded lead has no response value at all and is excluded from the median rather than counted as zero seconds. The count of leads never answered is on the row beside the median, because the statistic is blind to them.

Lead-source mix is context, not a score. This project publishes no lead-quality ranking, difficulty index or source weighting, and none is derivable here. The mix is shown because comparing two people’s contact rates without it compares two different jobs.

Nothing here is causal. A figure is credited to a person, observed for them, or on transactions they handled. The model records associations between people and outcomes; it does not isolate an individual effect, and no figure on this page supports a statement about individual skill.

Every person on this page is invented, and so is every number. The codes identify fictional employees in a synthetic dataset. No name, contact detail, hire date, termination date, exact tenure, age, pay, commission or protected attribute exists anywhere in the governed export this page reads, and no figure here is comparable to any published market figure.