Completeness
Finds what is missing or extra compared with the reference.
OutputOmission and commission rates by area or feature type
Geospatial consulting and products · Greater Toronto Area
From data to decisions.
Mapping that holds up at any scale.
We help private and public organizations build better maps and datasets, strengthen QA/QC, set best practices, create custom metrics, validate spatial data and use AI safely, backed by over a decade of big tech map making.
Illustrative terrain
What we do
Our goal is simple: spatial data that your organization can trust and act on. We work with private and public teams of any size, from a single map to a large-scale product.
Our core service
Geospatial consulting
Experienced GIS support when you need a clear answer, a better workflow or more capacity on a project.
View geospatial consulting- Spatial analysis using repeatable, defensible methods
- Cartography and map production for reports, permits and the web
- GIS strategy and workflow reviews
- Tool selection, open source or commercial, and training
Data engineering and AI
Spatial pipelines, databases, automation and AI-assisted mapping.
View detailsMapping products
Strategy, data, design and build support for large-scale map products.
View detailsDatasets and QA/QC
Custom datasets, QA/QC databases, automated checks and quality review.
View detailsMetrics and best practices
Custom metric suites, reviews of your existing metrics and best-practice implementation.
View detailsValidation and ground truth
Independent testing of spatial data and AI output, plus ground truth datasets.
View detailsAI training and safety
Training, safe-use rules and best practices for any industry, with a focus on local government.
View detailsWhy work with us
Experience at scale
Over a decade of building maps in big tech, across base map data, design, imagery and quality.
AI you can check
We use AI where it helps, then test the results against reference data before you rely on them.
Any tooling
We can work with the tools you already use, commercial or open source, or recommend what fits your project and budget.
Work you can maintain
You receive the outputs, the method and the files, so you are not left depending on us.
Who we work with
We work with teams of any size and industry, and these are the groups we focus on.
Tech companies
Geospatial depth for product and engineering teams building with maps and location data.
For tech companiesEngineering and construction firms
Site and corridor mapping, terrain analysis and combined survey, CAD and GIS data.
For engineering and construction firmsConservation authorities
Watershed, wetland and land cover mapping, with data you can trace and maintain.
For conservation authoritiesOntario municipalities
Asset and land use data, open data publishing, public web maps and GIS reviews.
For Ontario municipalitiesAI speed. GIS accuracy.
AI processes imagery and messy data in a fraction of the time. Every result is then checked against reference data, so you get the speed without losing confidence in the map.
See how we work View our sample workWays to work with us
Start small or commit to something longer. We shape the work around what you need.
Fixed-scope project
An agreed deliverable, timeline and price. Best when you know what you need.
Ongoing retainer
A set amount of support each month. Best for continuing updates and maintenance.
Embedded with your team
We work inside your team for a period. Best when you need capacity and expertise.
Audit or review
A short, paid assessment with clear recommendations. A low-risk way to start.
Any project, any scope
From a single map to a national dataset, we shape the work around what you need. We can support your team for a few days, lead a project from start to finish, or build a large-scale mapping product with you.
Geospatial consulting
Spatial analysis, cartography, GIS strategy and hands-on support.
View detailsData engineering and AI
Spatial pipelines, databases, automation and AI-assisted mapping.
View detailsMapping products
Strategy, data, design and build support for large-scale map products.
View detailsDatasets and QA/QC
Custom datasets, QA/QC databases, automated checks and quality review.
View detailsMetrics and best practices
Custom metric suites, reviews of your existing metrics and best-practice implementation.
View detailsValidation and ground truth
Independent testing of spatial data and AI output, plus ground truth datasets.
View detailsBeyond geospatial
AI training, safety and best practices
Practical support for any industry, with a special focus on local governments and municipalities.
View AI training and safety- Staff and leadership training
- AI policies and guardrails
- Risk and readiness reviews
- Pilot design and evaluation
Ways to work with us
Fixed-scope project
An agreed deliverable, timeline and price. Best when you know what you need.
Ongoing retainer
A set amount of support each month. Best for continuing updates and maintenance.
Embedded with your team
We work inside your team for a period. Best when you need capacity and expertise.
Audit or review
A short, paid assessment with clear recommendations. A low-risk way to start.
Tools we work with
We are not tied to any one platform. We can work with the tools you already use, or recommend and build with what fits best, whether commercial or open source.
GIS and mapping
Data and databases
Code and automation
Cloud and AI
These are examples, not limits. If your team uses something else, we can work with it.
Common questions
- Do you take on small projects?
- Yes. We work on projects of any size, from a single map or review to a large-scale product.
- Can you work with the tools we already use?
- Yes. We can work with your existing tools and systems. We also deliver solutions using open source technologies and commercial GIS platforms.
- Do you provide legal land surveys?
- No. We do not provide legal surveys or stamped engineering. We work alongside your licensed surveyors and engineers.
- Who owns the finished work?
- You own the deliverables we create for you, as set out in our agreement.
- Is our data used to train AI models?
- Not unless you agree in writing. We confirm how your data will be handled before work begins.
Geospatial consulting
Experienced GIS support when you need a clear answer, a better workflow or more capacity on a project.
Stream density by watershed unit
Thematic maps turn tables into patterns people can act on. This one compares mapped stream length per square kilometre across the four units.
- Bolton Dam2.07 km/km²
- West Humber1.99 km/km²
- East Humber1.92 km/km²
- Black Creek1.30 km/km²
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
What we offer
- Spatial analysisSuitability, proximity, network, terrain and change analysis using repeatable methods.
- Cartography and map productionClear maps for reports, permits, public engagement, presentations and the web.
- GIS strategy and workflow reviewA practical look at your data, tools and processes, with priorities for improving them.
- Tools and trainingHelp choosing tools such as QGIS, and training shaped around how your team works.
When to call us
- You have a location question and need an answer you can defend.
- GIS work is scattered across files, people and tools.
- You need to choose between open source and commercial software.
- A deadline is close and your team needs more hands.
What you receive
- Finished maps in print and digital formats
- Analysis results with documented methods
- Clear recommendations and next steps
- Project files and training materials
Data engineering and AI
Reliable systems for moving, processing and improving spatial data, with AI used where it can be measured and checked.
From raw sources to trusted data
A repeatable pipeline keeps people and AI in the right places, and it can run again whenever the data changes.
- CodeCollectBring in imagery, open data and your own files.
- CodeClean and alignFix formats, schemas and coordinate systems.
- AIExtractUse AI where it speeds up feature creation.
- Code + peopleValidateTest against rules and reference data.
- People sign offPublishDeliver data, documentation and quality results.
What we offer
- Geospatial pipelinesRepeatable workflows to ingest, clean, conflate, transform and publish spatial data, built with tools such as Python and GDAL.
- Spatial databasesPostGIS and SQL databases designed around your data, queries and growth.
- AI-assisted mappingFeature extraction, change detection and data cleanup that speed up production.
- Training data and model evaluationPrepare training data and test model output against reference data.
When to call us
- Updates depend on slow manual steps.
- Your data sources do not line up with each other.
- Your data has outgrown desktop tools.
- An AI pilot looks promising, but nobody has measured its accuracy.
What you receive
- Documented, repeatable code and workflows
- Database schemas and data models
- Accuracy reports for AI output
- Handover notes your team can maintain
Mapping products
Support for teams building maps and location features into products, drawing on over a decade of large-scale map making.
One dataset, three map styles
The same part of the Humber River watershed is drawn for three different audiences.
TechnicalContours and streams for analysis
ThematicWater features for a focused story
PublicMain rivers for a general audience
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
What we offer
- Product strategyShape roadmaps and priorities around what your data can realistically support.
- Data sourcingCompare open, licensed and custom data on coverage, quality, cost and terms.
- Map designClear map styles and experiences for your users, across zoom levels and devices.
- Build supportWork with your engineers on the data models, pipelines and checks behind the map.
When to call us
- You are adding maps or location features and need a clear plan.
- You are not sure which data sources will meet your coverage and quality needs.
- Your map is hard to read or does not feel right to users.
- Quality varies by region and is hard to track between releases.
What you receive
- Roadmap input and data requirements
- Data source comparisons
- Map style specifications
- Quality metrics and release checks
Datasets and QA/QC
Spatial data built to your specification, and the systems to prove that it is right.
Try our QA/QC database sample or see our validation and ground truth service.
Where QA/QC issues were found
Showing issues on a map reveals patterns that a list can hide. These 20 candidates still need review by a person.
- High severity2
- Medium severity15
- Low severity3
Review them in our QA/QC sample
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
What we offer
- Dataset productionCustom spatial datasets from imagery, public records and other sources, ready to use in your systems.
- QA/QC databasesDatabases that track errors, corrections and sign-offs, so quality can be measured and audited.
- Automated validationRules that check geometry, topology, attributes and schema before data is released.
- Independent quality reviewA focused review of an existing dataset, covering completeness, accuracy and consistency.
When to call us
- The dataset you need does not exist, or is not good enough.
- You do not know how accurate or complete your data is.
- Errors are found late, after they have reached users or reports.
- There is no record of what was checked, changed or approved.
What you receive
- Datasets in your required format, with metadata
- A working QA/QC database
- Validation rules you can run again
- Quality reports with clear findings
Metrics and best practices
A suite of metrics built around your needs, or a clear review of the ones you already use, so you can see data quality and act on it.
Stream name coverage by watershed unit
A useful metric shows where to act. East Humber has the highest share of stream length with an official name, and Bolton Dam has the lowest. See more in our Humber River dashboard.
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
What we offer
- Custom metric suitesMetrics designed around your goals, your data and the decisions you need to make.
- Existing metric reviewsCheck that your current metrics measure the right things, then improve how they are defined and calculated.
- Best-practice implementationPut data standards, workflows, documentation and QA/QC processes in place across your teams.
- Baselines and trackingMeasure where you are today and how your data changes over time.
Metrics we can build
Every suite is tailored to your work. Common starting points include the measures below, and you can see real examples in our Humber River sample.
Positional accuracy
Measures how far features sit from their true location.
OutputRMSE, offset direction and error maps
Logical consistency
Checks geometry, topology and schema against the rules.
OutputRule pass rates, gaps, overlaps and invalid records
Thematic quality
Tests whether classes, names and attributes are right.
OutputConfusion matrix and accuracy by class
Temporal quality
Shows how current data is, and how quickly changes arrive.
OutputAge by region, update lag and freshness trend
Coverage
Shows where quality or detail changes across an area.
OutputGrid scores, regional rankings and heat maps
Model agreement
Compares AI or vendor output with ground truth.
OutputPrecision, recall, F1 score and error examples
Error resolution
Tracks whether quality work is making progress.
OutputOpen issues, time to fix and sign-off rate
These are illustrations of how each metric works, not measured results. We tailor every suite to your data and goals, and can align it with quality elements from ISO 19157-1.
When to call us
- You cannot tell whether your data is getting better or worse.
- You have metrics, but people do not trust or use them.
- Teams use different standards for the same data.
- Leaders need clear measures to support a decision.
What you receive
- Clear metric definitions and calculation methods
- A baseline of your current data quality
- Recommendations to improve existing metrics
- Documented standards and best practices
AI training, safety and best practices
Practical support to help your organization use AI safely and well. We work across industries, with a special focus on local governments and municipalities.
Why it matters for public bodies. In January 2026, Ontario's Information and Privacy Commissioner and Human Rights Commission released six joint principles for responsible AI in the public sector. They are recommendations, not legal requirements, but they offer a useful starting point. Some municipalities, such as Norfolk County, have already adopted AI policies.
A practical path to safe AI use
Start where you are. Each step can stand alone or build on the last.
- Step 1AssessLearn how AI is used today, and where the risks and opportunities are.
- Step 2Set the rulesDefine approved tools, data limits and human review.
- Step 3TrainGive staff and leaders skills suited to their roles.
- Step 4PilotTest a focused use case and measure the results.
- Step 5MonitorTrack quality, incidents and lessons, then update the rules.
How we help
- Staff and leadership trainingHands-on sessions on what AI can do, where it fails and how to check its output.
- AI policies and guardrailsDraft practical rules for approved tools, sensitive information, disclosure and review, for your legal team to check.
- Risk and readiness reviewsIdentify where AI is already in use, the risks involved and the gaps to close first.
- Pilot design and evaluationChoose a safe first use case, set success measures and test the results.
- Vendor and tool assessmentAsk the right questions about accuracy, privacy, security and data use before you buy.
- Data readinessImprove the quality and governance of the data that AI tools rely on.
Principles put into practice
These are the six principles from the Ontario guidance, with what each can mean for your organization.
Valid and reliable
Test tools before use and keep checking that they work as intended.
Safe
Monitor for harm, protect against security threats and retire unsafe systems.
Privacy protective
Limit the data used and build privacy in from the start.
Human rights affirming
Look for bias and correct discriminatory results.
Transparent
Document where AI is used and tell people when they interact with it.
Accountable
Name who is responsible, keep people in charge and make it possible to pause a system.
For local governments
- Council and senior staff briefings
- Staff guidance for everyday tools
- Rules for personal and confidential information
- Clear disclosure when AI contributes to public work
- Evaluation of AI in services such as permits and asset data
For any industry
- Role-based staff training
- Internal use policies and review steps
- Testing AI output for accuracy and bias
- Vendor and model evaluation
- Measures that show whether AI is helping
Where AI meets location data
AI support is not limited to geospatial work, but this is where our experience runs deepest. We can help you evaluate AI-assisted mapping, prepare spatial training data and test results against ground truth.
Good first projects
These are examples of work we can take on, not past client projects.
AI readiness review
Map current use, risks and quick wins across your organization.
Staff workshop
Run a practical session built around your real tasks and rules.
Draft AI use policy
Create a clear starting policy for your legal and privacy teams to review.
We do not provide legal advice. Please confirm legal and privacy requirements with your own counsel.
Independent validation and ground truth
We test spatial data and AI output against trusted reference data, and build ground truth datasets others can measure against.
How a validation works
Every result can be traced to a specification, a sample and a documented method.
- Step 1DefineAgree on the requirements, quality measures and pass thresholds.
- Step 2SampleDesign a sample that represents the regions and feature types that matter.
- Step 3CompareMatch the data under test to independent reference data.
- Step 4ScoreCalculate metrics and confidence, then review the errors.
- Step 5ReportDeliver a scorecard, error examples and clear recommendations.
Validation of your data
A second set of eyes on data you build, buy or depend on.
- Vendor data acceptance testing
- Independent review of another company’s dataset
- AI and machine learning output evaluation
- Comparison of competing data providers
- Scheduled checks between releases
Ground truth datasets
Reference data with known quality that others can test against, built from open datasets, high-accuracy sources and field verification when necessary.
- Benchmarks for AI model testing
- Checkpoint and reference feature sets
- Sample design across regions and feature types
- Annotation rules and quality control
- Versioning, metadata and documented uncertainty
What we measure
We choose measures to match your data and how it will be used.
Model agreement
Compares AI or vendor output with ground truth.
OutputPrecision, recall, F1 score and error examples
Positional accuracy
Measures how far features sit from their true location.
OutputRMSE, offset direction and error maps
Completeness
Finds what is missing or extra compared with the reference.
OutputOmission and commission rates by area or feature type
These are illustrations of how each metric works, not measured results.
Where our ground truth comes from
We match each source to the accuracy your project needs, and go to the field only when necessary.
- Start hereOpen source datasetsUse open data where its licence, coverage and quality fit the test.
- Add precisionExisting high-accuracy sourcesBring in authoritative data with documented accuracy for the features that matter most.
- When necessaryField verificationCheck on the ground when existing sources disagree, fall short or are out of date.
Checked before use
Open or authoritative does not automatically mean accurate enough, so we assess each source for the test at hand.
Traceable records
Reference records keep their source, date, method and known accuracy.
Licences respected
We confirm how each source may be used and shared, and credit it as required.
How we keep results credible
A validation is only useful if people can trust how it was done.
Independent sources
Checkpoints should come from an independent source of higher accuracy than the data being tested.
Separate from training
Data used to train a model is kept out of the data used to test it.
Conflicts disclosed
We disclose any involvement with the data, and do not validate our own work as independent.
Honest uncertainty
No reference is perfect, so we document how ground truth was made and its known limits.
When to call us
- You are buying data and need proof that it meets the specification.
- A customer or regulator is asking how accurate your data is.
- Your AI team needs a trustworthy benchmark.
- You need to choose between data providers.
- Quality may be changing between releases.
What you receive
- A validation plan and sampling design
- A scorecard with metrics by area and feature type
- Error maps and annotated examples
- Pass or fail results against your requirements
- A documented reference dataset, where included
Good first projects
These are examples of work we can take on, not past client projects.
Vendor spot check
Test a sample of a dataset before you commit to buying it.
Model benchmark
Build a reference set and score AI output against it.
Provider comparison
Compare two or more providers against the same reference data.
See our dataset validation sample See how we track issues
Our validations are independent professional assessments. They are not legal surveys or regulatory certifications.
Built for the work your sector does
The data and the rules change from one industry to the next. The discipline of good mapping does not.
Conservation and environment
Conservation authorities and environmental teams depend on data that is current, traceable and easy to explain to regulators and communities.
- Habitat, wetland and watershed mapping
- Land cover change over time
- Maps and data for impact assessments
- Monitoring data management and QA/QC
Engineering and construction
Engineering and construction projects bring together survey, design and GIS data that often use different formats and coordinate systems.
- Site and corridor mapping
- Combining survey, CAD and GIS data
- Terrain and elevation analysis
- Project maps for planning and reporting
Municipalities and government
Ontario municipalities and other public organizations need accurate, accessible data that can be maintained long after a project ends.
- Municipal asset and land use data
- Open data preparation and publishing
- Public-facing web maps
- GIS workflow and data quality reviews
Tech companies
Product and engineering teams need geospatial depth without slowing down their release cycle.
- Map and location product strategy
- Spatial data pipelines at scale
- Geospatial AI training data and evaluation
- Automated quality checks
Other industries
If your data has a location, the same methods usually apply. Tell us what you are trying to decide.
- Utilities and infrastructure
- Transportation and logistics
- Natural resources
- Real estate and planning
We do not provide legal land surveys or stamped engineering. We work alongside your licensed professionals.
For tech companies
Geospatial expertise for teams building with maps, location data and spatial AI, without slowing your release cycle.
From raw sources to trusted data
A repeatable pipeline keeps people and AI in the right places, and it can run again whenever the data changes.
- CodeCollectBring in imagery, open data and your own files.
- CodeClean and alignFix formats, schemas and coordinate systems.
- AIExtractUse AI where it speeds up feature creation.
- Code + peopleValidateTest against rules and reference data.
- People sign offPublishDeliver data, documentation and quality results.
How we help
- Map and location strategyShape your roadmap around what your location data can realistically support.
- Data pipelines and conflationBuild repeatable workflows that combine, clean and publish spatial data at scale.
- Quality metricsDefine what good looks like, and measure it across regions and releases.
- Geospatial AI evaluationPrepare training data and test model output against reference data.
- Independent validationTest your data or a provider's data against ground truth. Learn more.
- AI training and safetyHelp your teams use AI safely, with clear rules and evaluation. Learn more.
When to call us
- You are launching a map or location feature and need a clear plan.
- You need to compare data vendors or combine several sources.
- Users keep finding errors, and you need to fix them at the source.
- A spatial AI model looks promising, but its accuracy is unproven.
What you receive
- Data source comparisons and requirements
- Metric definitions and quality baselines
- Documented pipelines your engineers can maintain
- Model evaluation reports
Good first projects
These are examples of work we can take on, not past client projects.
Data source audit
Compare open and licensed sources on coverage, quality, cost and terms.
Quality baseline
Measure your current data and set metrics to track between releases.
AI output evaluation
Test extracted features against reference data before you ship them.
For engineering and construction firms
Reliable GIS support for project teams that need survey, design and spatial data to work together.
Contours and streams in one view
Project teams can see terrain and water together for early planning. Contours here are shown every 25 m.
Screening-level sample, not for design.
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
How we help
- Site and corridor mappingClear maps of constraints, alignments and existing conditions for planning and reporting.
- Survey, CAD and GIS integrationBring data with different formats and coordinate systems into one reliable project dataset.
- Terrain and elevation analysisUse elevation data to support slope, drainage and site analysis.
- Project GIS capacityAdd experienced GIS support to your team when project demand peaks.
We do not provide legal land surveys or stamped engineering. We work alongside your licensed surveyors and engineers.
When to call us
- Project data sits in formats and coordinate systems that do not line up.
- A report or submission needs clear maps on a tight deadline.
- Your team needs GIS capacity for a single project.
- You need an independent check of spatial data before relying on it.
What you receive
- Maps and figures ready for reports
- Clean project datasets with documented coordinate systems
- Analysis results with documented methods
- QA notes that show what was checked
Good first projects
These are examples of work we can take on, not past client projects.
Corridor constraints map
Combine open and project data into one map of the constraints along a route.
CAD and GIS cleanup
Align and clean project files so that your teams work from one dataset.
Project data review
Check datums, completeness and consistency before design decisions rely on the data.
See our Humber River sample Try our QA/QC database sample
Other client groups
For conservation authorities
Watershed mapping, data harmonization and QA/QC for teams responsible for Ontario's watersheds.
Preparing for consolidation. Ontario's 36 conservation authorities are set to amalgamate into 9 regional authorities on February 1, 2027. Merging organizations often means combining datasets built to different standards. We can help you take stock of your data, align it and check it.
Bringing datasets together
A clear process helps merging organizations keep track of every dataset and decision.
- Step 1InventoryRecord datasets, owners, formats and licences.
- Step 2CompareFind overlaps, gaps and conflicting definitions.
- Step 3CrosswalkMap old fields and categories to a shared schema.
- Step 4MergeCombine the data in a repeatable workflow.
- Step 5VerifyRun QA/QC checks and record sign-off.
Humber River watershed units
Four official watershed units, with streams and waterbodies from the Ontario Hydro Network, covering 908.5 km².
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
How we help
- Data inventory and harmonizationDocument what each dataset holds, then align schemas, reference systems and standards.
- Watershed mapping and analysisBuild watershed profiles, land cover change analysis and clear maps for reports.
- Regulation mapping supportPrepare and check data and cartography for mapping under O. Reg. 41/24. Regulatory decisions stay with your team.
- QA/QC and metricsBuild QA/QC databases and metrics that track data quality over time.
When to call us
- Datasets are about to be combined across organizations.
- Your teams use different schemas, names or standards for the same data.
- You need clear evidence of how complete and current your data is.
- A watershed report or plan needs maps and analysis that can be traced to their sources.
What you receive
- Data inventories and schema crosswalks
- Harmonized datasets with metadata
- Watershed maps and statistics
- QA/QC results and data quality metrics
Good first projects
These are examples of work we can take on, not past client projects.
Data inventory
Create a clear record of datasets, owners, formats, licences and known gaps.
Schema crosswalk
Map how fields and categories in different datasets relate to a shared standard.
Watershed profile
Combine authoritative data into a checked profile, like our Humber River sample.
See our Humber River sample Try our QA/QC database sample
Other client groups
For Ontario municipalities
Accurate, accessible spatial data that your staff can maintain long after a project ends.
A simpler map for public audiences
Public maps work best when they show only what people need. This style keeps the main rivers and clear labels.
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
How we help
- Asset data supportImprove the spatial data behind asset management planning under O. Reg. 588/17.
- Open data publishingPrepare, document and publish datasets that people can find and use.
- Public web mapsBuild clear public maps with accessibility in mind. Ontario public sector websites must meet WCAG 2.0 Level AA, with limited exceptions.
- GIS workflow reviewsReview your data, tools and processes, then set practical priorities.
- AI training and safetySupport staff and council with practical training, safe-use rules and evaluation. Learn more.
- Vendor data validationCheck that purchased or contractor data meets your requirements. Learn more.
When to call us
- Asset or land use data is incomplete, outdated or spread across departments.
- You want to publish open data but it needs cleanup and documentation first.
- Residents need a clear public map.
- A small GIS team needs extra capacity or a second opinion.
What you receive
- Cleaned datasets with metadata
- Data quality reports and metrics
- Public maps and supporting documentation
- Recommendations and training for your staff
Good first projects
These are examples of work we can take on, not past client projects.
Asset data review
Measure how complete and consistent your asset data is, then prioritize fixes.
Open data readiness
Assess which datasets are ready to publish and what each one needs first.
Public web map
Turn a priority dataset into a clear public map, tested for accessibility.
See our Humber River sample Try our QA/QC database sample
Other client groups
Built to work with your team
Our founder has spent over a decade building maps inside big tech. We know how product and engineering teams work, and we can step in without slowing yours down.
For product teams
Turn location data into features your users can rely on.
- Map product strategyShape roadmaps and priorities around what the data can support.
- Data sourcingCompare open, licensed and custom data on coverage, quality, cost and terms.
- Quality metricsDefine what good looks like and measure it across regions and releases.
- Map designCreate clear styles and map experiences for your users.
For engineering teams
Add geospatial depth to the systems you already run.
- Data pipelinesBuild repeatable workflows to ingest, clean, conflate and publish spatial data.
- Data models and schemasDesign spatial structures that hold up as data and coverage grow.
- Automated QA/QCAdd validation checks and QA/QC databases that catch errors before release.
- Geospatial AIPrepare training data and test model output against reference data.
Common starting points
A new map feature
You are adding location to your product and need the plan, data and design right.
A data decision
You need to compare data sources or decide whether to build, buy or combine.
A quality problem
Users are finding errors, and you need to measure and fix them at the source.
A capacity gap
You need geospatial expertise on the team for a project or a period of growth.
How we fit in
We work your way
We can join your planning, tools and review process rather than adding a separate one.
We document as we go
Decisions, methods and code stay clear for the people who take them forward.
We leave you stronger
We hand over the work and the knowledge so your team can carry it on.
AI speed. GIS accuracy.
AI processes imagery and messy data in a fraction of the time. Every result is then checked against reference data, so you get the speed without losing confidence in the map.
We can work with any tooling, commercial or open source, and recommend what fits your project and budget.
- People
Scope
Agree on the decision the map supports and the accuracy it needs.
- AI + people
Prepare
Audit sources, licences and reference systems, including NAD83(CSRS) and CGVD2013 for Canadian work.
- AI + GIS
Build
Combine AI, code and GIS in a documented, repeatable workflow.
- People
Validate
Test results against reference data. A person signs off, and you receive the outputs, method and files.
What we check
Quality means different things on different projects, so we agree on the measures that matter before work starts.
Positional accuracy
Are features in the right place for the way the data will be used?
Completeness
Is anything missing, or included when it should not be?
Attribute accuracy
Are names, classes and values correct?
Logical consistency
Do geometry, topology and schema follow the rules?
Currency
Is the data recent enough, and is its age clear?
Lineage
Can each result be traced to its source and method?
How we use AI
Measured, not assumed
We test AI output against reference data and report the results.
People stay accountable
A person reviews the results and signs off on what we deliver.
Your data stays yours
We confirm how your data will be handled, and we do not use it to train models unless you agree in writing.
About Isarithm Technologies
We are a Canadian geospatial company based in the Greater Toronto Area. Our goal is to help private and public organizations turn data into decisions by strengthening their datasets, improving QA/QC and putting best practices in place. We bring over a decade of big tech map making, traditional GIS practice and modern AI to that work. Isarithm is a founder-led firm, and we bring in trusted specialists when a project needs them.
Our founder's experience
Isarithm is led by a founder with over a decade of experience building maps in big tech, supporting work at global scale. That experience is backed by a university education in geospatial studies and computer science.
Maps at global scale
Worked on consumer map products and global base map data, including roads, buildings, land use and places.
Data quality
Worked on programs covering quality metrics, validation, QA/QC and ground truth.
Imagery and AI
Worked on imagery-based mapping and AI, including machine learning, feature extraction and training data.
Hands-on and leadership
Experience spans GIS and data work, data science and engineering, technical leadership, program management, and leading teams of 10 to 25 people.
What this means for you: senior experience from a founder who has seen what breaks at scale, applied at the size your project needs.
What we value
Accuracy you can check
We show our methods and measure our results.
Plain communication
We explain spatial work in terms your team can use.
Practical choices
We recommend what fits, not what is newest or most expensive.
Lasting value
We leave you with work that your team can understand and maintain.
Our mark
Our logo comes from a basic rule of map reading. Where a contour line crosses a stream, it bends into a V that points upstream. This is known as the rule of Vs.
It is a small example of what we care about: a few simple lines can tell you how the land works, if you know how to read them. The blue line is a nod to the water behind the rule.
Tools we work with
We are not tied to any one platform. We can work with the tools you already use, or recommend and build with what fits best, whether commercial or open source.
GIS and mapping
Data and databases
Code and automation
Cloud and AI
These are examples, not limits. If your team uses something else, we can work with it.
Isarithm sample work
Humber River watershed profile
A sample of how we turn open data into decisions: official Ontario datasets, combined into one watershed profile, measured with a metric suite and checked.
908.5 km²Watershed area
4Watershed units
1,594 kmMapped streams
135 to 485 mContour elevations
Selected unit
Watershed unit statistics
| Unit | Code | Area (km²) | Share | Streams (km) | Density (km/km²) | Lakes and ponds | Open water (ha) |
|---|
Waterbodies crossing a unit boundary are counted in each unit they touch. Elevations come from contours, so true highs and lows can differ.
Data quality dashboard
This is the kind of metric suite we can build around your data. It measures 1,838.9 km of watercourse network in the Humber River watershed, including modelled flow paths.
99.5%Network connectivity
9.8 km sits outside the main network, across 10 separate groups.
100%Completeness
Required fields are filled on all 6,773 segments. 11.4% of stream length has an official name.
94.7%Freshness
Share of network length with a last geometry update in 2019, the most common year.
99.9%Recorded accuracy
Share of network length recorded as within 5 metres. This is the source’s stated value, not an independent test.
Metrics by watershed unit
| Unit | Connected | Isolated (km) | Flow gaps | Required fields | Named streams | Within 5 m | Median update |
|---|---|---|---|---|---|---|---|
| Black Creek - Humber River Outlet | 98.8% | 6.2 | 1 | 100% | 12.6% | 99.8% | 2019 |
| West Humber River | 99.8% | 0.7 | 1 | 100% | 9.5% | 100.0% | 2019 |
| Bolton Dam - Humber River | 99.8% | 0.9 | 2 | 100% | 8.7% | 100.0% | 2019 |
| East Humber River | 99.5% | 2.0 | 2 | 100% | 15.1% | 100.0% | 2019 |
The highlighted unit has the lowest connectivity.
Last geometry update
Network length by the year its geometry was last updated.
20191,741.4 km
202067.7 km
20226.6 km
202321.8 km
20251.4 km
Where to look first
Black Creek - Humber River Outlet holds 6.2 of the 9.8 isolated kilometres, so we would review it first.
What freshness can’t show
An update date shows when geometry was last edited, not whether it still matches the ground. A real project would compare it with recent imagery.
What to measure next
Independent accuracy testing needs reference data, such as survey points or LiDAR. We would add this for a client project.
Metric graphs
Each graph answers a practical question. All of them are calculated from the official datasets used in this sample.
Stream density in 2 km grid cells
Grid metrics show variation hidden by unit totals. Across 247 cells, the median is 1.73 km/km² and the highest is 4.27 km/km². A low value can reflect the landscape or less detailed mapping, so unusual cells are good candidates for review.
Stream density, km/km²
01234.3
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project. Cells with under 25% of their area in the watershed are left out.
Bolton Dam has the densest mapped streams, at 2.07 km/km²
Stream length per square kilometre, by unit.
Bolton Dam2.07 km/km²
West Humber1.99 km/km²
East Humber1.92 km/km²
Black Creek1.30 km/km²
Black Creek has the highest modelled share, at 19.9%
Network length by type. Modelled paths connect flow through features such as waterbodies.
- Streams
- Modelled flow paths
- Value shows modelled share
Black Creek19.9%
Bolton Dam14.4%
East Humber9.7%
West Humber7.8%
Black Creek holds the most isolated network, at 6.2 km
Length outside the main connected network, by unit. These are candidates for review.
Black Creek6.2 km
East Humber2.0 km
Bolton Dam0.9 km
West Humber0.7 km
Bolton Dam reaches the highest ground, at 485 m
Range of contour elevations in each unit. True highs and lows can differ.
Bolton Dam215 to 485 m
Black Creek135 to 365 m
East Humber145 to 365 m
West Humber140 to 315 m
100200300400500 m
East Humber has the largest open water share, at 2.5%
Mapped waterbody area as a share of each unit.
East Humber2.5%
Bolton Dam2.4%
Black Creek1.5%
West Humber0.9%
89% of mapped waterbody pieces are under 0.5 ha
Count of clipped waterbody pieces by size. Small features are often the hardest to keep complete.
Under 0.1 ha2,226
0.1 to 0.5 ha1,776
0.5 to 1 ha296
1 to 5 ha186
5 ha or more30
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
Metrics we can output
The graphs above show part of what is possible. We can build a suite from any of these measures, or improve the metrics you already use.
Completeness
Finds what is missing or extra compared with the reference.
OutputOmission and commission rates by area or feature type
Positional accuracy
Measures how far features sit from their true location.
OutputRMSE, offset direction and error maps
Logical consistency
Checks geometry, topology and schema against the rules.
OutputRule pass rates, gaps, overlaps and invalid records
Thematic quality
Tests whether classes, names and attributes are right.
OutputConfusion matrix and accuracy by class
Temporal quality
Shows how current data is, and how quickly changes arrive.
OutputAge by region, update lag and freshness trend
Coverage
Shows where quality or detail changes across an area.
OutputGrid scores, regional rankings and heat maps
Model agreement
Compares AI or vendor output with ground truth.
OutputPrecision, recall, F1 score and error examples
Error resolution
Tracks whether quality work is making progress.
OutputOpen issues, time to fix and sign-off rate
These are illustrations of how each metric works, not measured results. We tailor every suite to your data and goals, and can align it with quality elements from ISO 19157-1.
QA/QC results
We ran these checks on the datasets used for this sample. Items marked Review would go to a person on a real project. You can review them yourself in our QA/QC database sample.
- PassBoundaries are valid
4 of 4 watershed units have valid geometry.
- PassNo overlaps or gaps
Units overlap by 0 m² and combine into one polygon with 0 holes.
- PassAreas reconcile
Our equal-area calculation gives 908.48 km², matching the 908.48 km² in the source attributes.
- PassRecords are complete
We removed 2,989 repeated records from the paged download. Final counts match the source service: 10,781 watercourses, 5,494 waterbodies and 17,277 contours in the search area.
- PassRequired attributes are filled
6,773 clipped watercourse segments checked. Blank permanency values: 0.
- ReviewDisconnected segments
99.5% of the network connects as one system. 9.8 km across 10 separate groups does not. These may be real, or may result from clipping or gaps in the data. They are highlighted on the map.
- ReviewFlow gaps
6 segments are classed as flow gaps in the source. A person would confirm how these should be treated.
- NoteStream names
11.4% of stream length carries an official name. The rest would rely on other identifiers in reporting.
How it was made
- Download official watershed, watercourse, waterbody and contour data.
- Remove repeated records and confirm counts against the source.
- Clip the data to the four Humber River watershed units.
- Calculate area in an equal-area projection and length in UTM Zone 17N.
- Run the QA/QC checks and simplify the geometry for the web.
What a client project adds
- Metrics chosen around your decision, such as flood risk or restoration priorities.
- Your own monitoring, asset and field data.
- LiDAR elevation, culverts, dams and storm sewers.
- Review of every flagged item by a person.
- Deliverables in the formats your team uses.
Data sources
Ontario Watershed Boundaries, Ontario Hydro Network (Watercourse and Waterbody) and Contour, from Land Information Ontario, retrieved November 2026. Contains information licensed under the Open Government Licence – Ontario. Map geometry is simplified for display.
Isarithm sample work
QA/QC database sample
Find issues, review them, record each decision and show what was checked. Try the workflow below using issues from the Humber River data.
| Issue | Check | Severity | Unit | Size | Status |
|---|
How the database is structured
Each issue links back to its dataset and the check that found it. Every change is kept as an event, so nothing is lost when a status changes.
PostGIS example
CREATE TABLE issues (
issue_id text PRIMARY KEY,
dataset_id integer NOT NULL REFERENCES datasets (dataset_id),
check_id integer NOT NULL REFERENCES checks (check_id),
unit_code text NOT NULL,
severity text NOT NULL CHECK (severity IN ('High', 'Medium', 'Low')),
status text NOT NULL DEFAULT 'Open',
geom geometry(Point, 26917) NOT NULL,
created_at timestamptz NOT NULL DEFAULT now()
);
-- Issues still waiting for sign-off, by watershed unit and severity
SELECT unit_code, severity, count(*) AS waiting
FROM issues
WHERE status <> 'Signed off'
GROUP BY unit_code, severity
ORDER BY unit_code, severity;What this gives your team
One record of quality
Issues, decisions and sign-offs sit in one place, not across spreadsheets and email.
Decisions you can defend
The audit trail shows who decided what, when and why.
Metrics that stay current
Counts and progress come directly from the records, so reports always match the data.
Sample work
See how we approach real problems. These are samples we built to show our methods, not client projects, and each one states what is real and what is simulated.
Real open data
Humber River watershed
Official Ontario data, profiled with an interactive map, dashboard and graphs.
View sampleReal open dataDataset validation
A dataset scored against reference data, with metrics and a field verification queue.
View sampleReal issues, practice workflowQA/QC database
Review real issues, record decisions and keep an audit trail.
View sampleReal open dataMunicipal utility data
A municipality's published water, sanitary and storm networks, with a detailed map, main break analysis and quality checks.
View sampleReal run on open dataData pipeline
A tested pipeline from public web service to published data, with lineage, data contracts, damage tests and reproducible results.
View sampleYour projectTest your own data
Tell us what you need to check, and we can suggest a focused first project.
Contact usIsarithm sample work
Dataset validation sample
How we test a dataset against reference data, score it against a requirement and decide where field verification is worth the trip.
100%Precision at 50 m
Share of the test data found near the reference.
80.2%Recall at 50 m
Share of the reference found near the test data.
89.0F1 score at 50 m
A single score that balances precision and recall.
2.1 mMedian offset
Typical distance from a test line to the nearest reference line.
Example requirement
Precision and recall of at least 95% and 90% at 50 m
Precision: passRecall: below requirement
The test data sits close to the reference, but about one-fifth of the reference network has no match within 50 m. It would not meet this example requirement for detailed network work, though it could still suit overview maps.
Compared
- Data under test
- 1,292.6 km
- Reference
- 1,838.9 km
- Measured every
- 50 m
- Test sample points
- 25,414
- Reference sample points
- 35,137
Validation metrics
Precision stays near 100% while recall climbs with tolerance. That pattern points to missing detail rather than misplaced lines.
Recall rises from 73.7% to 96.0% as tolerance widens
Results at five match distances.
Bolton Dam has the lowest recall, at 74.4%
Share of each unit's reference network matched within 50 m.
West Humber84.6%
Black Creek82.1%
East Humber79.7%
Bolton Dam74.4%
| Unit | Test data (km) | Precision | Recall | F1 score | Median offset |
|---|---|---|---|---|---|
| Black Creek - Humber River Outlet | 370.8 | 100.0% | 82.1% | 90.2 | 2.1 m |
| West Humber River | 335.7 | 100.0% | 84.6% | 91.7 | 2.1 m |
| Bolton Dam - Humber River | 299.7 | 100.0% | 74.4% | 85.3 | 2.1 m |
| East Humber River | 286.5 | 100.0% | 79.7% | 88.7 | 2.1 m |
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only, not a client project.
Three methods working together
Each method answers a different question, and we use the lightest one that can answer it with enough confidence.
Used in this sample
Ground truth data
Compare the test data with reference data selected for the job, and document its limits.
Used in this sample
Metrics
Measure precision, recall, F1 score and offset, by area and at several tolerances.
Planned, not carried out
Field verification
Visit the sites where desk methods cannot settle what is on the ground.
Field verification queue
These 12 sites were chosen from the largest differences, at least 1.5 km apart. A real plan would also confirm safe, legal access.
| Site | Why it was selected | Unit | Distance to test data | Status |
|---|---|---|---|---|
| Named stream not in test data (Emery Creek) | Black Creek | 249 m | Not visited | |
| Named stream not in test data (Humber River) | Black Creek | 143 m | Not visited | |
| Stream not in test data | Black Creek | 527 m | Not visited | |
| Stream not in test data | East Humber | 509 m | Not visited | |
| Stream not in test data | Black Creek | 504 m | Not visited | |
| Stream not in test data | Bolton Dam | 493 m | Not visited | |
| Stream not in test data | Black Creek | 484 m | Not visited | |
| Stream not in test data | East Humber | 483 m | Not visited | |
| Stream not in test data | West Humber | 483 m | Not visited | |
| Stream not in test data | Black Creek | 482 m | Not visited | |
| Stream not in test data | West Humber | 475 m | Not visited | |
| Stream not in test data | Bolton Dam | 470 m | Not visited |
What a field crew records
- GPS position and its accuracy
- Dated photos
- Whether a channel exists, and whether water is flowing
- Conditions such as culverts or recent construction
- Any access limits or safety concerns
What you would receive
- A scorecard against your requirement
- Metrics by area and tolerance
- A map and list of differences
- Field records, where field work is included
- A documented reference dataset, where included
Isarithm sample work
Municipal utility data sample
A profile of one municipality's published water, sanitary and storm networks, built from real open data and measured with a metric suite.
939 kmWater mains
1,720 kmSanitary and storm sewers
2,930Recorded main breaks
156,524Records analyzed
Central area network map
Selected record
Showing about 3.9 km by 4.2 km of the central area: 11,446 building outlines, 3,836 water mains, 7,074 sewer pipes and 776 main breaks. Zoom in for street names and individual assets. Building colours follow the published category: grey for residential, tan for commercial, brown for industrial and utility, and blue-grey for institutional and recreational. Line geometry is lightly simplified for display.
Citywide inventory
| Network | Pipes | Length (km) | Related assets |
|---|---|---|---|
| Water | 16,214 | 938.6 | 8,401 valves and 4,860 hydrants |
| Sanitary | 14,755 | 884.7 | 13,451 manholes and 25 active pumping stations |
| Storm | 21,863 | 835.6 | 18,662 manholes, 15,227 catch basins, 1,228 outlets and 169 facilities |
Water network by pressure zone
| Zone | Length (km) | Mains | Mean age | Valves | Hydrants | Main breaks, 2016 to 2025 |
|---|
Zone and owner names are generalized for this sample. Mean age is weighted by length. Each break is assigned to its nearest current main. Zones with under 1 km of main are left out.
Data quality dashboard
This is the kind of metric suite we could build for a utility. These checks use only the published open data, so private fields a city holds internally could answer some of them.
99.8%Linked water assets
Valves and hydrants whose related main ID exists in the published mains layer.
94.0%Sewers with inverts
Gravity sewers with both an upstream and a downstream invert.
374Possible adverse slopes
1.1% of gravity sewers with inverts record a lower upstream end.
17.4%Storm material code XXX
Share of storm pipes with this code, compared with 0.5% of sanitary pipes.
Published data checks
Review means a person would check these items. It does not mean the records are wrong.
Metric graphs
Each graph answers a practical question, using municipal open data retrieved on November 5, 2026.
Main breaks per km of main, 2016 to 2025
Across 740 main breaks in 10 years, this grid shows where recorded breaks concentrate. It shows history, not a forecast, and small cells with little main can look high.
Breaks per km
0
Point to a grid cell to see its values.
Cells with under 200 m of main are left out. Lines show mains of 300 mm or larger.
99.5% of buildings are within 150 m of a hydrant
Straight-line distance from 16,767 published building outlines to the nearest hydrant, leaving out sheds, garages and utility buildings. This is a simple coverage measure, not a fire protection assessment.
Utility metrics we can output
We can build a suite from any of these, or improve the metrics your team already uses.
Network connectivity
Finds broken ends, loose stubs and isolated parts of a network.
OutputConnected share, gaps by zone and traceable length
Attribute completeness
Measures required fields for each asset type and owner.
OutputFilled share by field, zone, asset class and age
Positional confidence
Compares records with as-builts, GNSS points or field checks.
OutputOffset statistics and the source of each location
Hydraulic consistency
Tests sewer slopes, inverts, flow direction and connections.
OutputAdverse slopes, invert errors and low-capacity segments
Isolation analysis
For authorized staff, shows the valves needed to isolate each main, and who loses service.
OutputValves to close, customers affected and critical sites
Service coverage
Measures reach from hydrants, facilities or other service points.
OutputCoverage by area, and places outside the target distance
Renewal risk
Combines age, material, break history, size and consequence.
OutputRanked renewal list, with likelihood and impact scores
Real-time reliability
Tracks delays, rejected readings and events without an asset match.
OutputLatency by source, match rate and feed health
What the checks suggest
On a client project, each of these would become a conversation with staff who know the system.
- StrongWater asset links are very complete
Only 9 of 8,401 valves and 24 of 4,860 hydrants refer to a main not found in the published layer.
- StrongRecorded pipe lengths match the map
No sanitary or storm pipe differs from its mapped length by more than our tolerance.
- ReviewStorm material has the largest gap
3,801 storm pipes, or 17.4%, use the code XXX. Filling this gap could improve renewal planning, if these are confirmed as unknown.
- ReviewSome sewers appear to slope the wrong way
129 sanitary and 245 storm pipes record a lower upstream invert. These could be real conditions, survey differences or data entry issues.
- ReviewMany storm pipes refer to structures outside the manhole layer
4,213 storm pipes, or 19.3%, have an end ID not found in the storm manhole layer. Many may connect to catch basins, inlets or outlets, which a next step could test.
- ReviewBreak causes are often unspecific
77.1% of main break records have a blank or unknown break nature. Better categories could make root-cause analysis more useful.
- LimitThis is not a true network test
The published metadata says water features are not modelled as a true network, and that pipes are broken at road blocks. We therefore flag only 298 main ends that stop within 10 m of another main, as candidates for review.
Improving a utility database
A practical cleanup improves the data in stages, and every change stays traceable. Here is how these stages would apply to patterns found in this data.
Standardize
Move mixed material values, such as Concrete and CP, to one list.
Complete
Fill missing inverts and materials from drawings, CCTV and field records.
Connect
Link pipes to all structures, and test them as a true network.
Maintain
Add validation rules and QA/QC so new errors are caught early.
Fields we used
- Water mainsID, pressure zone, size, material, install date, owner and geometry
- Sanitary and storm pipesID, size, material, install year, inverts, slope, length, owner, connected manholes and geometry
- Valves and hydrantsID, related main and location, plus valve type and size
- Main breaksDate, type, nature, cause, pipe size, material, year and location
- Base mapRoads, buildings, parks, water, rail, trails and neighbourhoods
Example validation rule
-- Flag gravity sewers that may slope the wrong way
SELECT sanpipeid
FROM sanitary_pipes
WHERE category = 'GRAVITY'
AND up_invert < dn_invert;This rule found 129 sanitary pipes in the published data.
What we left out on purpose
- Street addresses of main breaks
- Service pipes to individual properties
- Published criticality and hazard fields
- Shutdown tracing on a real water system
- Rankings of individual pipes for renewal
Adding real-time data
This sample has no live feed. With a municipality, a real-time map would connect events to trusted asset records in five steps.
- SourcesCollectSensors, SCADA exports and work orders send updates.
- CheckValidateReject bad readings and match events to asset IDs.
- StoreUpdateSave events and keep the full history.
- ShowMapDisplay current status for approved staff.
- LearnMeasureTrack data quality, delays and response times.
Who could use it
- Operations staff tracking active work
- Engineers planning renewals
- Asset managers measuring data quality
- Emergency teams finding nearby valves
- Leaders reviewing performance
What real deployments need
- Security review and restricted access
- Separation from control systems
- Rules for sensitive infrastructure data
- Backup plans if a feed fails
- Clear ownership of each dataset
How it was made
- Download 21 layers from the municipality's public web services.
- Confirm that each layer's record count matches the service.
- Keep only the fields needed, and leave out sensitive fields.
- Project the data to UTM Zone 17N, and measure lengths from geometry.
- Run the checks, then simplify the central area for the map.
What a client project adds
- Internal GIS, CCTV, asset management and work order records.
- Rules, standards and scoring agreed with your staff.
- As-built research and real field verification.
- Network modelling, shutdown analysis and renewal priorities.
- Secure delivery for approved staff in the tools your team uses.
Data source
Open data published by a Canadian municipality under its open government licence, retrieved 2026-11-05. The municipality provides this information as is, and makes no warranty about its accuracy. Zone and owner names are generalized for this sample, and material descriptions are our reading of the published codes.
Isarithm sample work
Data pipeline sample
The pipeline behind our Humber River sample, from public web service to checked website data. We reran and tested it, and this page shows what happened.
36,545Records downloaded
33,556Unique records
16 of 16Pipeline checks passed
11 of 13Damage tests caught
Step through the pipeline
- Input
- Output
- Automatic check
- Measured time
Python excerpt
Data lineage
Lineage shows where every output came from and what depends on it. Select an item to trace it through the pipeline.
A fingerprint is a short code calculated from a file's exact contents. If one byte changes, the code changes.
What the pipeline produced
All of these figures were measured. Times are the median of three runs in a small cloud workspace with 2 processors and 7 GB of memory.
Data shrank from 146.9 MB to 0.9 MB for the web
Total size of the data at each stage.
The stages take about 25 seconds in total
Clipping takes the longest. This excludes the original download and the separate tests.
Records at each stage
A record that crosses a unit boundary becomes more than one clipped part. The web map leaves out waterbodies under 0.5 ha, and shows only 25 m contours.
Data contracts
A data contract states what each input must contain, so a change at the source is caught before it reaches a report or map. These profiles come from the source files in this run.
Pipeline checks
These checks test whether the pipeline did its job. They are separate from the 20 data issues in the Humber sample, which a person would review.
Testing the checks
A check that always passes proves nothing. We deliberately damaged copies of the data, then ran the matching check.
The 2 misses are expected. These checks test structure and logic. They cannot tell that a believable position or value is wrong. That takes comparison with reference data or a field check, as shown in our dataset validation sample.
Choosing the web detail
Simplifying lines makes a map fast, but it also moves them and shortens their mapped length. We tested seven tolerances on all 236,075 stream points to see how the 15 m tolerance used on our map compares.
Why 15 m
At 15 m, the map keeps 9.0% of stream points, and no line moves more than 15 m. Going to 50 m would remove only 3.8% more of the original points, but would lose a further 5.8% of mapped length.
The web version is for display only. Lengths and areas are always calculated from the full geometry.
Humber River data from Land Information Ontario. Contains information licensed under the Open Government Licence – Ontario. Sample only.
Can the results be reproduced?
Yes, for every stage after the download. Three runs produced identical files, which also match the files from the original run. A full rerun of the original script rebuilt the published web data in 23 seconds.
The web test file comes from the timed test stage. Check C15 separately confirms that the published web data can be rebuilt exactly.
Problems found during the original run
- Early stops in pagingThe first downloads stopped at 1,904 of 10,781 watercourses, 1,856 of 5,494 waterbodies and 1,802 of 17,277 contours. The count check exposed the shortfalls, and the paging rule was fixed.
- Repeated recordsThe corrected download repeated 572 watercourses, 492 waterbodies and 1,925 contours. Removing them by ID made all counts match the service.
Finding what changed
A production pipeline should not rebuild everything when little has changed. To test the method, we changed 25 records, removed 3 and added 5 in a copy of the watercourses.
- Changed records found
- 25 of 25
- Removed records found
- 3 of 3
- New records found
- 5 of 5
- Share needing updates
- 0.31%
- Comparison time
- 0.43 s
This tests the comparison only. These were test changes, not real updates from the source.
From a one-off run to production
The results above are from a tested one-off run. This is how we would propose running a similar pipeline for a client. It has not been deployed or operated.
| Situation | Starting rule | Proposed response |
|---|---|---|
| Source is unchanged | Every record fingerprint matches the last run | Record the check and end the run |
| Source has small changes | Fingerprints differ for some records | Process affected records, then run all checks |
| Download is incomplete | Record count differs from the count reported by the source | Retry, then stop and alert the owner, keeping the last good data live |
| Source changes sharply | Record count moves by more than 2% between runs | Hold publication until a person reviews it |
| Source structure changes | Any data contract check fails | Stop before processing and alert the owner |
| Run becomes slow | Run time exceeds three times its baseline | Warn the owner, while allowing the run to finish |
| Output is too large | Web file exceeds its 1.5 MB budget | Stop publication and keep the last good version |
Each rule and threshold would be agreed with your team, based on how often your sources change.
Scheduling
Run on a schedule or when new data arrives.
Monitoring
Alert the right person when a step fails or a metric drops.
Version history
Keep earlier outputs and record what changed.
Handover
Provide documentation and tests your team can maintain.
How this was tested
- Keep the original downloaded files as source snapshots.
- Rerun the deduplicate, clip, measure, write and simplify stages three times.
- Compare fingerprints across runs and against the original outputs.
- Run 16 checks on real data, then 13 tests on damaged copies.
- Test seven simplification tolerances on all clipped streams.
What this does not show
- The download was not repeated, so current source data was not checked.
- No scheduler, alerting or monitoring system has been operated.
- Passing these checks does not prove that the source data is accurate.
- Times are specific to this workspace and data size, and will differ elsewhere.
Data sources
Ontario Watershed Boundaries, Ontario Hydro Network (Watercourse and Waterbody) and Contour, from Land Information Ontario, retrieved November 2026. Contains information licensed under the Open Government Licence – Ontario.
Contact us
Whether you need one map, a dataset or a full mapping product, we would like to hear about it. It is fine if the scope is still rough. Working that out is part of what we do.
Emailhello@isarithm.ca
Based inGreater Toronto Area, Ontario, Canada
Working withTeams of any size, across Canada
Privacy policy
This policy explains what personal information Isarithm Technologies collects through this website and how we handle it.
Last updated November 5, 2026
What we collect
If you use our contact form or email us, we collect what you send: your name, email address, organization, your answers about your project, and your message. Please don't include sensitive personal information in your message.
Like most websites, our hosting and font services can receive technical details when you visit, such as your IP address, browser type and the pages you request.
How we use it
We use your details to read your enquiry, reply to you, prepare any proposal you ask for, keep basic business records and protect the form from spam. We don't sell your personal information, and we don't add you to a marketing list because you contacted us.
Cookies and analytics
This site's own code doesn't set cookies, and it doesn't use analytics or advertising trackers. If that changes, we will update this policy first.
Services we use
- Netlify hosts this website, and receives and stores contact form submissions. See Netlify's privacy statement.
- Google Workspace provides our email. See Google's privacy policy.
- Google Fonts supplies the fonts on this site, so your browser requests them from Google.
These providers may store or process information outside Canada, including in the United States, where it may be subject to local laws. We may also disclose information if the law requires it.
How long we keep it
We keep your information only as long as we need it for the purposes above, including our business records, and then delete it.
Your choices
You can ask to see or correct the personal information we hold about you, withdraw your consent, or ask us to delete it, subject to any records we must keep by law. Email hello@isarithm.ca, and we will respond within 30 days.
Questions or complaints
Please contact us first at hello@isarithm.ca. You can also complain to the Office of the Privacy Commissioner of Canada.
Changes to this policy
We will post any changes on this page and update the date above.