We built eight example companies, from a CNC shop to a Montessori school. We ran the same model on all of them. Some pay back. Some do not. We left those in.
The base case
| Company | Revenue basis ($M) | Levers at 100% | Year 3 net EBITDA change | Year 3 margin change | Payback (months) | Share needed to break even |
|---|---|---|---|---|---|---|
| Fenwright Logistics | 6.54 | $359,824 | $158,736 | plus 227 bps | 25 | 41% |
| Lindqvist Precision | 7.34 | $237,765 | $105,075 | plus 68 bps | 24 | 40% |
| The Wexmoor | 5.98 | $163,012 | $55,169 | plus 64 bps | Over 36 | 69% |
| Kettle Creek Veterinary | 10.45 | $151,519 | $50,303 | plus 24 bps | Over 36 | 70% |
| Mendell Property Group | 0.83 | $83,783 | $31,779 | plus 322 bps | 30 | 51% |
| Larkhaven Montessori | 3.12 | $70,301 | $15,956 | plus 49 bps | Over 36 | 94% |
| Quillback Marine Supply | 2.57 | $54,245 | $3,687 | plus 6 bps | Over 36 | 115% |
| Talus Yard Climbing | 2.02 | $49,639 | ($5,734) | minus 53 bps | Over 36 | 167% |
Revenue basis is net revenue for the broker and fee revenue for the property manager. Margin change is measured on that basis. Payback includes the build. No baseline growth is modeled.
Three cases
| Case | Share of modeled value that lands | Companies paying back in 36 | Year 3 margin change, range |
|---|---|---|---|
| Conservative | 35% | 0 of 8 | minus 104 to plus 100 bps |
| Base | 60% | 3 of 8 | minus 53 to plus 322 bps |
| Stretch | 85% | 5 of 8 | minus 3 to plus 537 bps |
The one formula
Lever value = volume x change x unit value x flow-through.
Annual net change = lever total x realization x ramp, minus annual program cost.
Ramp is 40%, 80%, then 100%. Saved hours count at 50%, because saved hours are not saved dollars. Nothing compounds.
Program costs are invented too. These are not our prices.
What we saw
- Repetitive work with money attached pays first. Fenwright's billing paperwork and margin per load carry that model.
- Fixed cost decides small companies. Quillback's levers are real and still barely cover what it costs to run.
- Demand has to be the constraint. Larkhaven has 300 to 400 inquiries for 40 seats, so more leads are worth almost nothing. We modeled 2 added students.
- Time is the soft number. About a third of the value here is freed time. It only counts if it turns into a skipped hire, less overtime or more work.
Every assumption is labeled
Each of the 160 inputs in the sheet carries one of four labels.
- Fixture: read from the company data we already built.
- Benchmark: from a public source we checked.
- Derived: calculated from fixtures.
- Invented: our judgment. Shaded yellow. Challenge these first.
Most inputs are invented. That is why this is a model and not proof. Change any blue cell and it recalculates.
The experiment
Now we try to recreate this with real clients.
- Register first. Fill in the model before go-live, then freeze it.
- Baseline from the books. Trailing 12 months, from the ledger.
- Measure quarterly. Same formula. Measured change over modeled change, lever by lever.
- Publish the misses. A lever that underdelivers stays in the table.
- Validated means something. 12 months of data, tied to the books, at least 3 clients, median attainment at 60% or better. Until then the word is modeled.
No client appears here, named or recognizable, without written approval.
Reminder: these companies are fictional. Download the sheet and break it.