Counterflow
Counterflow
A traffic and safety data platform for local and regional government. It ingests what you already hold, makes it comparable, fills the gaps with modelled estimates, and reports the numbers a business case needs.
What goes in
Counts and traffic
- Local authority counter networks
- Department for Transport manual and automatic counts
- Telraam citizen science sensors
- Vivacity and camera-based systems
- Combined authority models and extracts
- Client CSV, GeoJSON and shapefile exports
Safety and casualties
- Collision records (STATS19 format)
- Severity, mode and location attributes
- Exposure denominators for risk per billion kilometres
Network and place
- OpenStreetMap, including cycle infrastructure tags
- Ordnance Survey road and pavement networks (PSGA)
- Population and employment distributions
- Client design options and scheme geometries
What happens to it
- Network integration. Disparate network representations are joined into one routable graph, with attributes carried across. Our matching algorithm is about 100 times faster than established techniques, which is what makes regional-scale integration practical.
- Traffic estimation. Annual average daily traffic and mode split are estimated for every link from observed counts plus network and place characteristics, with validation on held-out counters.
- Demand and potential. Origin and destination flows, route assignment and modelled uptake are used to show where latent walking and cycling demand exists.
- Risk. Casualty risk is expressed per billion kilometres travelled, so a quiet street and a busy junction can be compared honestly.
- Scenarios. Proposed network changes are applied to the model to estimate mode shift, flow change and casualty change.
Validation, not assertion
Models are calibrated and reported against held-out observations. When a model explains little, we say so, and the platform reports confidence alongside every estimate.
The current Oxfordshire cycle model reaches a held-out R squared of 0.31 with a rank correlation of 0.64 across 339 counter sites; the walking model reaches 0.39 and 0.69.
What comes out
Map application
Baseline multi-modal traffic, infrastructure and casualties on one map, with filters by mode and aggregation level.
Gap analysis
Where provision is missing or poor relative to modelled demand, including low public transport provision.
Business case inputs
Before and after estimates, retrospective benefit to cost ratios, and flow and casualty change for a package of schemes.
Outputs are also available as data, not only as a web map: per-region database extracts, vector tiles and CSV exports for your own analysis and reporting.
How it is built and delivered
Components
The dashboard is assembled from parts that can be licensed or deployed separately:
- A DuckDB schema that accepts disparate count and collision sources
- Ingest adapters for each source, with idempotent daily aggregation
- The network integration and matching engine
- Traffic, demand and risk models with validation harnesses
- Vector tile and static data bundle generation for fast browsers
- The map and chart front end
Delivery
- Static bundle
- Runs entirely in the browser with precomputed data, so hosting costs are trivial and no personal data is processed.
- Hosted service
- We run the ingest pipeline and publish refreshed regions on a schedule.
- Own infrastructure
- Full pipeline and data delivered to your own environment, for clients who cannot use a hosted service.
- Data licensing
- Source licence terms are documented per deployment, including the Ordnance Survey position.
Live demonstrations
| Deployment | Coverage | Access |
|---|---|---|
| Counterflow dashboard | West Midlands, West Yorkshire, Greater Manchester, Oxfordshire | Open |
| Critical Safety Issues | West Yorkshire network, screened against Active Travel England criteria | Password |
| Oxfordshire counters | 832 automatic and citizen science counters, 20 km study area | Password |
Want a walkthrough on your own authority’s data? Book one here.