One address, compared with Ontario.

The Ontario comparison

The dated register contains 5,038 mapped active pharmacy registrations at 4,968 distinct coordinates. The comparison uses 4,539 locations with complete inputs; 429 are excluded for incomplete local Census inputs. Each coordinate counts once. The report compares an address with the middle result, or median, across that group.

Experimental location percentile

The percentile combines four equally weighted comparisons: local population within 750 metres, the share of residents age 65 or older within 750 metres, public healthcare sites within 1 kilometre, and pharmacy registrations within 1 kilometre. Higher population, older-adult share and healthcare counts, and lower pharmacy counts rank higher. These directions and weights are assumptions, not learned effects on business performance.

Version 2 uses the same great-circle distances and supply rules in the website and report. All registrations within the radius count, including ones at the searched or reference coordinate. Closest-pharmacy distance is displayed separately and no longer affects the score. Whole Census areas are assigned by their representative point; these are not exact populations inside a circle.

The component percentiles are averaged and ranked against the same 4,539-location group. Ties receive half weight, using the midpoint of strictly-below and at-or-below fractions. A 93rd-percentile result describes a relative rank; it does not mean a 93% chance of success.

Weight sensitivity: nine settings are tested: equal weight, then half and double each category weight in turn, with the reference locations reranked every time. The displayed range is the minimum and maximum rank across those settings, not a confidence interval or evidence of future success. It does not cover all possible weights, missing locations, geocoding error or Census age.

The updated 888 Yonge comparison is the 91st percentile (Top 9%), with a 58th-to-98th-percentile weight-sensitivity range. The previous Top 7% result is superseded. The wide range means its rank is sensitive to the scoring assumptions.

Methods: SciPy percentile definitions and the OECD composite-indicator handbook. Download the calculation specification.

What is measured

The address review combines the OCP register, Statistics Canada Census data, Ontario public healthcare locations, municipal traffic records, zoning records, mapped buildings and parking records. Each figure names its distance, year and comparison where one is available.

Distances shown as walking time are estimates based on straight-line distance increased by 25%, using a walking speed of 4.8 kilometres per hour. They are not pedestrian routes. Maps are included whenever the individual locations matter more than a total.

Site facts are kept separate from area facts. A whole-building footprint is not presented as the available unit size. Public parking is not presented as dedicated customer parking. Zoning is not presented as permission to operate a pharmacy.

Asking-price study

The separate acquisition calculator uses 121 public pharmacy-business listings. It studies seller asking prices, not completed sale prices and not the value of a bare address.

The code, anonymous listing rows, formula and reproduction tests are available on the open-source page.

Retired prescription experiment

A previous experiment tried to estimate prescription workload from 15 named stores with reported counts. Its exact-count R² was 0.065. The counts were not independently verified and the sample was too small to support a product claim. That experiment has been withdrawn and is not used by the address check, report or asking-price calculator.

Costco and other factors

The 15-store experiment also tested hospitals, clinics, Costco, Walmart, other nearby businesses and local demographics one at a time. Adding hospital proximity improved the test R² by 0.069; adding Costco proximity reduced it by 0.296. These are small-sample test results, not evidence that either business causes better or worse pharmacy performance.

All 14 comparisons remain available in the factor results and the report appendix. They are not included in the location percentile.

AI-assisted model experiments

Three sets of AI-suggested operating and financial inputs were tested on 117 listings. The best within-dataset R² rose from 0.654 to 0.717, but every expanded model performed worse when an entire marketplace was held out. The prescription-volume baseline remains in use.

This tested suggested inputs in conventional models; it did not fine-tune an AI model. No verified owner-education data were available, so UofT or other education backgrounds have no measured effect and do not change the score.

What the report cannot know

Public records do not provide a pharmacy's prescriptions, sales, profit, rent, staffing, patient loyalty, clinic referrals or completed sale price. The location percentile does not fill those gaps or predict a commercial outcome.

Register status, addresses, traffic, buildings, parking and healthcare locations can change. The report is a starting point for research, not a substitute for a lease review, site visit, accountant, lawyer, appraiser or regulator.