A band that is read, not fitted
One five-year price-growth band for Kenya, stated by the framework rather than curve-fitted to whichever analogs happen to carry a number.
Kenya real-estate market intelligence for investors, developers, lenders, and advisors.
We do not start with property prices. We start with the country — its growth, urbanization, credit, and infrastructure — then look at countries that once looked like Kenya, observe what their property markets did next, and apply only the patterns that repeated.
Not a valuation. Not a price target. Not a timing bet.
The method in five steps
Selection happens before any property outcome is examined. Each step is a gate, not a menu.
The platform is built around a few disciplined rules that keep the output honest and usable.
One five-year price-growth band for Kenya, stated by the framework rather than curve-fitted to whichever analogs happen to carry a number.
We build Kenya's macro baseline — growth, urbanization, credit, and infrastructure — before looking at a single property outcome.
Countries matched on income, urbanization, credit depth, and mortgage depth — not geography — and locked before any property outcome is examined.
No ordinal rank without measured skill, no weights without a judgment panel, no worst case without a figure to stress against. Every gap is stated on the page it would have appeared on.
Each monitor carries a pre-committed level that flips the call, the headroom to it, and the consequence — set before the number moved, on its own axis.
A driver enters the Kenya forecast only if it recurs across multiple comparable countries. One market is an anecdote; recurrence is a mechanism.
Two halves: stages 1–2 build the foundation and choose the comparison set; stages 3–5 measure what happened and translate it into an outlook.
Build Kenya's 5–10 year macro base case
Output: A defensible Kenya baseline
Select comparable countries; set the anchor date
Output: A fixed country × time cohort
Observe the cohort's property response after the anchor
Output: Segment-level evidence
Keep only drivers recurring across analogues
Output: Validated mechanisms
Apply validated mechanisms to the baseline
Output: Decision-ready forecast
Each stage removes candidates that cannot be defended. Nothing is added back later. What reaches the final band is only what survived every filter — which is why the range is read off the evidence rather than fitted to a target.
Select any gate to see the test applied, what survives it, and a worked example — or start the walkthrough to step through the cascade one gate at a time.
1 of 6
Candidate countries
Test applied
Does the country have at least ten unbroken years of house-price, credit, and income data from an official or multilateral source?
What is kept
Any market with a continuous, sourced record — rich or poor, successful or not.
Worked example
Colombia stays in on a full central-bank price index; a peer with a three-year gap drops out.
The comparison set and its start dates are locked first. That firewall is what stops the answer from being reverse-engineered from the result we wanted.
Comparable countries that disappointed are kept and shown. The band widens honestly instead of narrowing around the success stories.
Each surviving driver is replayed on out-of-sample history. If it would have missed then, it does not get to speak about Kenya now.
The platform uses a reinforcement learning approach: every data point is tested against out-of-sample analogs and real outcomes before it is trusted. Once fully verified, a series is promoted to its own dataset and feeds the next cycle.
Ingest
Raw series enter with source tags and known coverage gaps
Verify
Back-tested against historical analog outcomes
Promote
Fully verified data becomes its own dataset