A forecast built to predict how an office gets used
An NDA Canadian employer needed to understand real office usage before making leasing decisions. We built a machine learning model that forecasts usage patterns with 97% accuracy.
97%
Forecast Accuracy Rate
1
Dedicated Engineering Team
1 month
From Kickoff To Completion
3x
ML Models Built & Deployed
AI & Automation
About The Company
They needed a system that could take what had already happened and forecast what would happen next, so future office decisions could be grounded in evidence rather than assumption.
The Challenge
Usage patterns were a guess
Employers needed to understand how their office space was actually being used day to day, not assume it based on lease terms or headcount alone.
30% sat underused
Without visibility into real usage patterns, the client risked paying for capacity that didn't match how employees actually worked - locking in cost for space nobody touched.
Demand swung 100%+
Office usage wasn't steady week to week: some weeks saw double the activity of others, which made any static planning approach fall apart almost immediately.
Data sat unused
The raw location and usage data existed, but nothing was forecasting from it, so past behavior never translated into a usable view of future demand.
No numbers to lease by
Office selection and leasing calls were being made without a clear, data-backed view of demand, leaving real financial decisions resting on instinct.
The Approach
Audit & Prepare
Reviewed employee location data and office usage data, cleaned and structured it for modeling, and mapped out the patterns worth forecasting.
- Employee location data reviewed and structured
- Office usage data cleaned and prepared
- Key usage patterns and variables identified
Build & Train
Built the machine learning model and trained it on historical usage data to forecast future patterns.
- Forecasting model architecture built
- Model trained on historical location and usage data
- Initial accuracy benchmarks tested
Validate & Deliver
Validated forecast accuracy, packaged the output into footprint insights, and handed off a model ready to inform leasing decisions.
- Forecast accuracy validated at 97%
- Footprint insights generated for office selection
- Model delivered, ready to inform leasing decisions
The Results
97%
Forecast accuracy confirmed against historical office usage data.
1 month
From scoping to validated delivery: one dedicated team, start to finish.
1 model
Single forecasting algorithm replacing guesswork across office locations.
0 guesses
Every leasing and office selection decision now backed by data and not assumption.
Want to forecast how your space gets used?
Precision targeting is the whole game.
Case Studies
Results that Compound
Built ML that Underwrites a Mortgage in Seconds
Two Canadian banks were issuing mortgages in weeks. We built the ML system that automated underwriting end to end.
