Search across all content
projectWaives site visits for low-risk builds by predicting which inspections will pass.
Prior to 2019, the City of Edmonton faced an administrative bottleneck due to a rapidly expanding residential housing market. Under the traditional operational framework, municipalities tasked human inspectors with conducting site visits for every single stage of new-build construction. Provincial guidelines delineated certain safety code evaluations as strictly mandatory, while others were designated as discretionary based on the local authority's perceived risk. Lacking a data-driven method to assess compliance probabilities dynamically, the city defaulted to an analogue paradigm, scheduling physical site visits across 11 to 15 different construction checkpoints per house.
This blanket approach quickly proved untenable. The necessity of sending expert human resource professionals to every minor, routine intervention created extensive scheduling backlogs. Builders experienced costly project delays waiting for field sign-offs, while municipal expert labour was systematically misallocated to low-risk construction sites that were statistically highly likely to pass initial checks. The baseline status quo handcuffed municipal agility: the city was forced to process over 50,000 unique inspection requests annually, distributed across a highly consolidated landscape where the top 100 contractors generated fully 80% of total inspection volume. Faced with clear structural and resource constraints, the municipality required an objective system capable of screening out statistically certain compliance passes to unlock latent inspector capacity.
To transition into a modern, data-driven regulator, the City of Edmonton’s Data Science and Research (DSR) team developed the Safety Code Inspection Efficiencies (SCIE) model. Engineered completely using open-source utilities within preexisting budgetary constraints, SCIE calculates the probability of an inspection passing on its first attempt, automatically waiving low-risk physical site visits so personnel can prioritise high-risk, complex infrastructure.
The project mechanics operate across two core components:
Part 1: Predictive Engine & Automated Data Ingestion
The technological core relies on a Random Forest machine learning algorithm trained extensively on historical inspector data. When a builder logs into the self-service portal to request a check-off, the model processes real-time inputs across four primary categories to yield a predictive assessment:
To eliminate structural bias, sensitive attributes like individual inspector identities are purposefully omitted or pruned during model training.
Part 2: Implementation Infrastructure & Quality Control Workflow
The SCIE model integrates directly with POSSE, the city's underlying permit and licensing workflow database. The end-to-end automation cycle operates under a rigorous logic gate structure:
How the SCIE model routes each inspection request:
If the system determines that a project exhibits an exceptional likelihood of compliance, it registers a "No Inspection Required" outcome, sending an automated notification to the builder at 10:00 AM on the day of the requested check. This functions as an immediate, zero-delay approval to proceed with construction.
To guarantee ongoing structural integrity, the model incorporates Supervised Adjustments. First, contractors with poor or inconsistent compliance histories are hard-coded to receive 100% physical site inspections. Second, a strict 5% random physical safety audit is automatically triggered on dropped clearances, ensuring the predictive logic remains accurate and bad actors cannot exploit the automated system.
Project Evolution and Iteration
The initiative has deliberately scaled across successive operational phases to secure trust and validation:
The SCIE deployment has transformed residential permitting in Edmonton, driving quantifiable budget efficiencies and operational optimisation.
1. Financial Efficiencies The initiative yields an annual cost avoidance of $150,000 for the City of Edmonton, achieving significant cumulative savings since its operational inception.
2. Resource Optimisation Across the primary target inspections, the city maintains an average historical reallocation rate of 22%. During peak expansion periods, the system successfully processes and automates up to approximately 40% of standard residential files on peak mornings.
3. Inspector Capacity Reclaimed In its expanded iteration, the model successfully frees an average of 1,838 highly skilled staff hours annually (increasing from 1,327 hours in the initial layout), redirecting 100% of these expert hours toward high-risk, non-compliant, or highly complex construction projects.
4. Reduced Friction for Industr. The program translates to an average of 4,138 site inspections dropped per year under expanded operations. This structural change eliminates construction lag, allowing premium local homebuilders to bypass physical inspection wait times and keep field crews active without localised delays.
5. Strategic Team Growth Crucially, the deployment was not leveraged to downsize the civil service; instead, data-driven savings allowed the city to permanently secure funding to expand its technical field presence, adding 10 Safety Codes Officers (SCO I & II), 1 Specialised SCO, and a General Supervisor to focus explicitly on complex oversight.
For public sector leaders looking to adapt Edmonton's framework, three transferable insights are evident.
Launch year: 2020
Connect with 500,000+ public servants solving your hardest challenges.
Connect with 500,000+ public servants solving your hardest challenges.
Help public servants worldwide learn from your work, what worked, what flopped and what you'd do differently
Share your project
Log in or sign up to continue the conversation