What a practice health grade can and cannot tell you
We built a mock inspection that scores your own walk the way an inspector weights one — heavier deductions for critical food-safety items, and a cap that makes it impossible to score well with a critical item open. Here is exactly how the model works, including the four things it will refuse to do.
· 7 min
Most restaurants find out how they are doing on the day somebody from the county walks in. The gap between “we think we are fine” and a posted score is where the bad mornings live.
So SavorCrew carries a practice inspection: a manager walks the building, answers the lines, and gets a grade. This piece is the model written out, because a score you cannot interrogate is a number you should not act on.
The arithmetic
grade = 100 − Σ(points on every line you flagged)
Each line on the walk carries a point value set by how much it matters. A critical food-safety line — temperatures, cross-contamination, handwashing — is worth more than a line about a scuffed skirting board. That is the whole of the weighting: it is not a curve and it is not a secret.
The cap, which is the part that matters
Any critical item still open → the score is capped at 69
You cannot arithmetic your way to a good practice grade with a critical item open. Pass ninety-five lines and fail one critical, and the grade is 69 — not 95, not 88.
This is deliberate and it is the single most important decision in the model. A points-only score lets a kitchen average away the one thing that actually closes restaurants. Real inspections do not work that way and neither does this.
The bands:
- 90+ — Inspection-ready
- 80–89 — Minor fixes
- 70–79 — Address before service
- below 70 — Critical items, fix now
The four things it refuses to do
It will not grade a half-finished walk. Leave required lines open and it declines, and says why: a grade on a half-done walk isn’t worth having. A partial score is worse than none, because it reads like a result.
It will not grade twice. One grade per walk, and the deductions are written once. Want another number? Do another walk. A score you can re-roll until it flatters you is not a measurement, and an audit trail that can be edited afterwards is not a trail.
It will not let a manager grade a building they do not work in. The venue is checked on both routes into the tool — the platform membership route and the employee-role route — so the two cannot drift apart and quietly become one open door.
It will not pretend to be a real inspection. Every single response carries this, verbatim, and the app renders it with every grade:
Practice grade only. This mock inspection uses a generic scoring model based on the U.S. FDA Food Code and is a self-assessment tool to help you prepare. It is not affiliated with, endorsed by, or predictive of any health department’s actual inspection or score. Local health codes vary by state and county; your official results may differ.
That paragraph is not a legal afterthought bolted to the footer. It ships inside the grade payload itself, which means no surface can show a score without it.
What it cannot tell you
Plainly, because the tool being useful depends on this being clear:
- It cannot predict your county’s score. Local codes vary by state and county, and inspectors exercise judgement a checklist cannot model
- It only knows what you told it. It is a self-assessment. A line marked fine that was not fine scores as fine
- It cannot see the day you did not walk. A grade is a snapshot of one walk, which is why the history matters more than any single number
What it is genuinely good for
Finding the thing before the inspector does, and watching a pattern across walks. The grade history is the real product — one score is a mood, six scores is a trend, and the line that keeps costing you points is the one worth spending money on.
There is a free browser version with no login and no email gate. The app version adds temperature captures, photo evidence on failures, offline drafts for walk-ins with no signal, and the saved history.