Istari · Deep Dive v4 · a working GIS view
Your block, on the real map. Priced, scored, ranked.
This is the half a spreadsheet can’t show: the city’s own administrative data — assessment, permits, transit, census — projected onto the real map, by neighbourhood. Colour it by cost, safety, walkability or who lives there. Then tell it what you care about, and it ranks the blocks that fit. Built to show municipal teams what their data can already do.
01 · the neighbourhood GIS explorer
Pick a city and a lens — or switch to house-hunting mode.
Real neighbourhoods, in their real positions, clipped to the real municipal shape. Hover anywhere for the numbers; orange rings mark where new homes are actually starting. Illustrative — borders between neighbourhoods are drawn by nearest-centroid (Thiessen) tessellation around real centroids inside real, simplified municipal outlines; metric values are Istari-modelled from sourceable inputs (method below). Production drops in each city’s open-data boundary file and live assessment/permit feeds.
◯ new homes starting. Hover the map for detail; click to pin. Green is the better end. Borders = nearest-centroid approximation.
Downtown Core
Inputs (sourceable): centrality 97 · transit 98 · affluence 72 · new-build 92. Illustrative
National-scale view: the same affordability lens, all four cities. Click one to jump the map there.
How it’s built — and the honest limits. Each neighbourhood carries four inputs a city can pull from its own systems: centrality, transit access, affluence, new-build intensity. Every colour is derived from those by the formulas in §03 — consistent, not hand-painted. What it is not: live MLS, police, or census microdata, and the inter-neighbourhood borders are nearest-centroid approximations, not surveyed lot lines (we tried to fetch the real boundary files; the open-data exports exceeded the tooling’s transfer limit, so we project real centroids instead and clip to the real outline). Hand a municipal team its own administrative data and this is real within an afternoon — that’s the point of the demo.
02 · who this is for
Same map. Five different questions.
One dataset, tuned to the frequency of whoever’s reading. Tap a card to set the map up for that question.
Real-estate agent
Show a client why one block prices above the next — and three comparable areas they haven't considered.
Developer
Find where the cost surface and new-build intensity say the next project pencils.
City planner
See where amenities and transit lag the density — the gaps a plan should close.
Mayor
One screen: where building is starting, where it isn't, and what residents feel as cost.
New family / newcomer
Say what matters — affordable, safe, walkable, room to grow — and rank the blocks that fit.
Renter / first buyer
Your budget against the real map: where it actually reaches, and what you trade for it.
03 · the algorithms, published
No black box — here’s the engine.
The simulator’s four functions and the explorer’s metric formulas, in full. A model you can’t audit is just an opinion with a chart.
canFinance = ( income×39%/12 × (1−(1+r)⁻³⁰⁰)/r ) ÷ 0.80 r = max(5.25%, rate+2%)/12
// builder pro-forma → feasibility F
mustSell = (land + hard×sf + soft + fees×(1−cut) − 2,900×monthsSaved + interest) × 1.15 F = canPay ÷ mustSell
starts = normal × 1/(1+e−10(F−0.92)) × gate grent = −7% + 20%/(1+e−5(x−0.05))
// explorer — neighbourhood metrics from inputs c,t,a,d
cost/sqft = base·(0.55 + 0.95·a + 0.45·c) walk = 28 + 62·t + 14·c safety = 74 + 18·(1−0.6·c) + 6·a
share<40 = 30 + 45·c income = base·120·(0.6+a)
// house-hunting fit = weighted blend of good-direction, 0..1-normalised metrics
fit = Σ wᵢ·norm(metricᵢ) ⁄ Σ wᵢ weights you set: affordability, safety, amenities, space
Full worked example, the S-curve and rent-response plots, and every limitation carry over from the v2 method note. Validation: Toronto’s observed 2-bed math (≈81%) puts apartment starts near a tenth of normal — inside the measured 88–94% collapse. The model wasn’t tuned to that; the prices imply it.
04 · how we got here — fees vs the approval leaders
Fees went one way. The math went the other.
Toronto’s development charge on a two-bedroom apartment rose from about $17,000 in 2015 to $80,690 in 2025 — roughly 17% compounding a year, a 4.7× climb, while incomes barely moved. A development charge is a fixed up-front cost, so it doesn’t fall back when the market softens. Verified endpoints; path Illustrative.
Meanwhile the countries that stayed affordable made building faster, not pricier — Japan approves by checklist in ~1.5 months, Auckland by-right in ~3, while a Vancouver approval averages 15.2 months and a Toronto high-rise 2–4.5 years. Reported Canada raised the price and the time to build for a decade; the leaders held both down. That’s why the simulator’s two cheapest levers are a fee cut and faster approvals.
05 · the G7 / OECD scoreboard
Three dials say the same thing.
Homes per 1,000 people — last in the G7 (Canada 424 vs 471 average; ≈1.8M-home gap). Verified Price-to-income — the sharpest rise of 23 OECD countries over two decades, ~5× the US pace since 2000. Verified · OECD The non-market buffer — nearly absent: social/limited-profit rental is 20%+ in Austria, Denmark and the Netherlands; Canada sits at ~4%. Verified · OECD
The pattern across every winner: they turned two dials — legalized market supply (Japan, New Zealand, Houston) and built a non-market buffer (Austria, Denmark, Netherlands). Canada has barely turned either. Each dial maps to a lever in the simulator, and each lever has a working precedent.
You’ve seen the whole picture
Now go build the policy you’d vote for.
Head back to the model, pull your levers, and share what you’d change — or bring us a stuck system of your own.
Sources & verification
- VerifiedCity of Toronto DC schedules 2015–2025 ($17k→$80,690, 2-bed). toronto.ca
- VerifiedVaughan −88–92% (2024), Mississauga −50% (2025) DC cuts. millerthomson.com
- ReportedApproval times — Vancouver 15.2-mo avg, Toronto high-rise 2–4.5 yr; delay $2,900/home (BILD/Altus); Japan/Auckland by-right. bildgta.ca
- VerifiedScotiabank — Canada 424 homes/1,000, last in G7 (avg 471). scotiabank.com
- VerifiedOECD price-to-income & Affordable Housing Database PH4.2 (social-rental shares). oecd.org
- ReportedNeighbourhood names & centroids — City of Toronto, City of Vancouver (22 local areas), City of Burnaby town centres, Town of Oakville communities (open data). Production embeds the boundary GeoJSON.
- IllustrativeIstari model — explorer metric & fit formulas, structural-input estimates, Oakville/Burnaby pro-formas, all simulator effect sizes. This page + v2 method note document it.
Explorer geometry: real neighbourhood centroids + simplified real municipal outlines, nearest-centroid choropleth (Thiessen) clipped to the outline; metrics Istari-modelled. Oakville/Burnaby pro-formas are Istari estimates within sourced ranges. Full archive: DATA-AND-SOURCES.md in this module folder.