What it does

It forecasts JadCo's effective rent growth for 2026: rent collected net of concessions, unit by unit, across the portfolio. Concession scenarios run from 1.5% to 3.8%, with a 90% error interval of 0.6% to 4.9%. Renewals and relocations are separated, and each province keeps its own rules. The TAL rate is a benchmark, not a cap, and The Met is exempt from Ontario's guideline. Results are also broken down by building and bedroom count.

How we built it

We checked what the data really meant: PromoPay is a one-time discount, not a monthly one, and we rebuilt sRentEffective to within $1.
An exact decomposition showed that 7.9 of 2023's +10.8 points came from new buildings, not price.
We fixed a selection rule before backtesting: a complex method must beat the plain historical average. Eight methods competed, including gradient boosting and TAL/CMHC-anchored models.
We tested landlord cost pass-through on 45 years of public data.

Challenges

The simplest method won. We accepted it instead of arguing around it.
Similar public data was hard to find
Short history: lease pairs only start in 2018.

What we learned

"Rent growth" covers at least five different numbers, so name yours before modelling.
A model has to beat a simple baseline in a fair test to earn its complexity.
Concessions are a pricing tool, not noise.