# Model Card

## Model

`asking-price-business-log-rx-alpha-4.0.0`

## Intended use

The model provides experimental context for a seller-reported pharmacy
business asking price in Ontario. It is most useful for relative comparison
and for flagging an asking price that sits far from public listings with
similar annual prescription volume.

## Formula

```text
asking_price_cad = exp(5.5145764264758075
  + 0.8154307097087575 * ln(1 + annual_prescriptions))
```

The published display range multiplies the fitted estimate by the 10th and
90th percentiles of cross-marketplace log residuals. It is an observed error
band, not a statistical confidence or prediction interval.

## Training data

- Geography: public Ontario pharmacy listings
- Snapshot: August 23, 2026
- Input rows: 129
- Excluded below support floor or invalid: 1
- Exact annual-prescription and asking-price duplicates removed: 7
- Model rows: 121
- Source split: 113 / 8
- Input: annual prescription volume
- Target: seller-reported business-only asking price in CAD

## Validation

The primary result is leave-one-marketplace-out validation. Each marketplace
is predicted by a model trained only on the other marketplace.

| Metric | Result |
| --- | ---: |
| Dollar-scale R-squared | 0.400265 |
| Mean absolute error | CAD 672,506 |
| Median absolute error | CAD 305,831 |
| Median absolute percentage error | 38.46% |
| Within 25% | 33.88% |
| Within 50% | 72.73% |
| Spearman rank correlation | 0.834792 |

Leave-one-listing-out R-squared is 0.663427 with mean absolute error of CAD
492,870. This is secondary because both marketplaces remain represented in
most training folds.

## Limitations

- The target is an asking price, not a completed sale.
- The smaller marketplace has only 8 observations.
- Annual prescription volume is the only input.
- Profit, payer mix, clinic dependence, staffing, lease economics, prescription
  type, inventory, working capital, real estate, debt, and deal structure are
  not modeled.
- The model does not predict future prescriptions, survival, or location
  success.
- The software and analysis were developed with AI assistance and have not
  received independent valuation, accounting, or software review.