Insights & Learning
Scope:
All shops (combined)
Valley Automotive
EuroPro Automotive
⬇ Export training data (CSV)
Suggested tuning (you approve — Phase 2)
- 💡 Diagnostic conversion is 0% (17 jobs) — a sign contact-time follow-up on diagnostics needs attention.
—
Estimate accuracy
0 jobs · est —h vs actual —h
22%
Recommendation hit rate
73 assigned jobs
0%
Diagnostic conversion
0 of 17 converted
76%
Deadline met
39 of 51 completed
By complexity tier
| Tier | Jobs | Avg actual hours | Est→actual ratio | Avg GP % |
|---|---|---|---|---|
| High | 26 | — | — | — |
| Medium | 350 | 4.63 | — | -103.0% |
| Low | 31 | — | — | — |
Observed technician performance (from completed work)
| Technician | Jobs | Avg actual hours | Implied efficiency | Avg GP % |
|---|---|---|---|---|
| Spirka, Charlie | 61 | 5.95 | need estimates | 67.0% |
| Verhaeghe, Jered | 37 | 8.32 | need estimates | 63.0% |
| Ian Anderson | 35 | 7.02 | need estimates | 50.0% |
| Anthony Miller DNC Staff | 18 | 4.03 | need estimates | -1058.0% |
| Robles, Fabrizio | 18 | 6.4 | need estimates | 59.0% |
| Tawnee J Ralston DNC | 14 | 8.17 | need estimates | 52.0% |
| Meldrum, Zack | 12 | 0.07 | need estimates | 5.0% |
| Voss, Joey | 6 | 1.0 | need estimates | 29.0% |
| Franklin, Blake | 4 | 0.5 | need estimates | 47.0% |
| Sean Lake | 4 | 2.98 | need estimates | 66.0% |
| Traweek, Derrik | 4 | 5.1 | need estimates | 62.0% |
| Beadell, Josh | 3 | 0.27 | need estimates | -4992.0% |
| Darren Grey | 3 | 4.93 | need estimates | 41.0% |
| Guzman, Gabriel | 2 | 1.12 | need estimates | -4979.0% |
| Halstead, Keagan | 2 | 0.4 | need estimates | 10.0% |
| Kelley, Bryan | 2 | 0.0 | need estimates | 7.0% |
| Reed, Josh | 2 | 0.0 | need estimates | — |
Implied efficiency compares estimated vs actual hours — the data-driven version of the number you set per technician. It sharpens as more imports land.
The road to combined ML
Phase 1 (now): every import snapshots outcomes here — pooled across both shops. Phase 2: the suggestions above become approve-able tuning. Phase 3: once enough history accrues, the exported training data trains models (job duration, best technician, deadline risk) whose predictions feed back into the workflow. Combined across shops means a bigger, richer dataset than either shop alone.