Future work
What we’d add next, organised by what kind of work it is. Roughly in priority order within each section. None of these are committed; they’re the next obvious places where the model would gain ground.
data
Real pedestrian activity
Sidewalk counters or anonymised cell-network mobility traces would let the model see who actually shows up, not just whether the urban form theoretically supports use. The single biggest shift in confidence we could make.
Equity context layer
Toronto Neighbourhood Profiles are loaded but not yet operationalised in scoring. Adding income, density, and dependency indices would let narratives say 'this park serves an under-served neighbourhood' — a meaningful piece of urban-policy interpretation.
Event programming density
City permit data + Park People programming records would surface programmed parks separately from form-strong parks. Currently we conflate the two.
model
Street View / computer-vision facade reading
Active vs blank ground floors. The 3D Massing layer tells us building height; it doesn't tell us whether the ground-floor windows are real or just glass mounted on a parking podium.
Temporal comparisons
The schema versioning + historical snapshot array are in place. The pipeline needs to actually capture year-over-year diffs so we can show park changes over time.
Seasonal modelling
Different parks shine in different seasons. The current scores are season-agnostic; a winter score would weight ice rinks, fire pits, and shelter differently.
Public validation loop
Use feedback signal to reweight per-metric confidence, surface 'model-vs-public disagreement' patterns, and prioritise data improvements.
research
MLS / real-estate proxy data
A check on whether high-vitality parks correlate with neighbourhood vibrancy on a separate axis. Not a scoring input — a sanity comparison.
Resident-perception surveys
Jane's Walk-style observations + structured surveys to ground-truth our public feedback. The contested-parks page is the first step toward this.
ui
Borough / ward roll-ups
Each park has a neighbourhood; we don't yet roll those up to the legacy boroughs (Old Toronto / North York / Scarborough / Etobicoke / York / East York). Easy follow-up.
PNG export of charts
Scatter and radar can already be screenshot, but a one-click 'save chart as image' button would help researchers cite figures.
Interactive cluster map
Each cluster overlaid on the city map would make geographic patterns visible.
Email-verified feedback + park-update notifications
Move the structured feedback / expert-review / 9-dim validation flows behind an email magic-link (or basic OAuth) so anonymous abuse has friction. Once we have an address per submitter, opt-in notifications can tell users when 'their' parks gain new ratings, get flagged for review, or change typology after the next cache rebuild — turning feedback into an ongoing relationship rather than a one-shot drop.
Add park as favourite
Per-user favourite list (browser-local first, then server-backed once auth lands) so frequent visitors can build their own atlas of parks they care about. Foundation for personalised digests once notifications are in place.