Methodology · engine v1.0.0
How we work out what a property is really worth
It's a calculation, not a guess. Put the same facts in and you get the same answer every time. Every weight and threshold we use is written out below.
Principles
AI never sets the number
A calculation, not a guess
The estimate is a pure function of recorded evidence and the published parameters on this page. We use AI to help read documents and explain results in plain English. It never sets or changes the value.
No invented prices
No market price is hard-coded anywhere. If there is not enough evidence for an approach, we leave that approach out and say we're less sure. We don't fill the gap with a guess.
Versioned and explainable
Every result is stamped with the engine version (currently v1.0.0) and valuation date. Changing any parameter below bumps the version, so stored reports stay explainable.
Evidence weighting
Not all evidence counts equally
Each observation's weight = confidence weight × recency weight × locality weight (× an extra factor for asking prices). Medians are weighted medians, so a few weak observations cannot drag the result.
Confidence weights
| Confidence | Weight |
|---|---|
| High | 1 |
| Medium-high | 0.85 |
| Medium | 0.7 |
| Low-medium | 0.5 |
| Low | 0.3 |
Recency weight
| Age of evidence | Weight |
|---|---|
| Today | 1 |
| 6 months | 0.71 |
| 1 year | 0.5 |
| 2 years | 0.25 |
| 3 years | 0.13 |
| Over 5 years | Excluded |
Half-life 365 days. Evidence dated after the valuation date is never used.
Locality weights
| Match level | Weight |
|---|---|
| Neighbourhood | 1 |
| City | 0.8 |
| Lga | 0.65 |
| State | 0.4 |
| Different state | Excluded |
| Asking price (comparables) | × 0.7 |
Approach A
Land value + depreciated replacement cost
Land value is the weighted median ₦/m² at the most specific locality with enough observations (neighbourhood → city → LGA → state), times the plot size. The building is valued at today's recorded cost to rebuild per m² for its class and specification, less depreciation, plus a capped uplift for external works and services.
Cost approach parameters
| Parameter | Value |
|---|---|
| Minimum land observations at a level | 3 |
| Minimum replacement-cost rates | 1 |
| Default specification when not stated | Standard |
| Economic life (straight-line) | 60 years |
| Minimum residual value | 20% |
| Depreciation when age unknown | 15% |
| Uncompleted building discount | 40% |
| Maximum total depreciation | 90% |
| Standard plot (market pages) | 464.5 m² (50×100 ft) |
Condition adjustment to depreciation
| Condition | Percentage points |
|---|---|
| New | No depreciation |
| Excellent | -5 |
| Good | 0 |
| Fair | +5 |
| Needs renovation | +15 |
Feature uplift on building cost
| Group | Uplift |
|---|---|
| Water | 1.5% |
| Power | 2% |
| Security | 2% |
| External Works | 2.5% |
| Ancillary | 3% |
| Cap on total uplift | 8% |
A group counts once however many of its features are present. Percentages only, never naira.
Approach B
Comparable properties
Similar properties of the same type group in the same state are adjusted to match the property being valued (by size where both sizes are known, otherwise by bedrooms), scored for similarity, and the best are combined by weighted median. Duplicate listings of the same property count once; asking prices are down-weighted and never converted into assumed sale prices.
Comparable parameters
| Parameter | Value |
|---|---|
| Minimum comparables | 3 |
| Comparables used (best by similarity × weight) | 8 |
| Minimum similarity | 0.3 |
| Size adjustment elasticity | 0.85 |
| Bedroom adjustment per bedroom | 8% |
| Maximum bedroom steps adjusted | 2 |
| Similarity factor: bedroom-based adjustment | 0.85 |
| Similarity factor: unadjusted price | 0.7 |
Similarity components
| Component | Weight | Zero similarity at |
|---|---|---|
| Size | 0.3 | 3× size ratio |
| Bedrooms | 0.2 | 3 bedrooms apart |
| Condition | 0.15 | Ordinal scale |
| Age | 0.1 | 30 years apart |
| Locality | 0.15 | Locality weight |
| Distance | 0.1 | 10 km apart |
Missing attributes score a neutral 0.5. Distance is used only when both properties have coordinates.
Approach C
Rental income
Value = weighted median annual rent ÷ weighted median gross yield, both from recorded evidence. Without yield evidence we leave the income approach out. We never assume a yield. Not applied to bare land.
Income approach parameters
| Parameter | Value |
|---|---|
| Minimum rental observations | 3 |
| Minimum yield observations | 1 |
| Weight for rentals ±1 bedroom | 0.5 |
| Weight when bedrooms unknown | 0.6 |
Approach D
Market momentum
The median ₦/m² of the latest window is compared with the window before it. Momentum is shown for context. It never changes the value.
Momentum parameters
| Parameter | Value |
|---|---|
| Window length | 6 months |
| Minimum observations per window | 2 |
| Reported as flat below | 3% |
Reconciliation
One central estimate, one honest range
Each available approach gets its base weight scaled by the quality of its evidence, then weights are normalised. The range around the central estimate widens when confidence is lower or the approaches disagree.
Base approach weights
| Approach | Base weight |
|---|---|
| Comparables | 0.5 |
| Cost | 0.3 |
| Income | 0.2 |
Range half-width
| Confidence | ± around central |
|---|---|
| High | 6% |
| Medium | 10% |
| Low | 16% |
| Extra per 1% disagreement | +0.25% |
| Maximum | 30% |
Approaches differing by more than 20% of the central estimate are flagged in the report.
Rounding
| Amount | Rounded to |
|---|---|
| ≥ ₦10,000,000 | ₦100,000 |
| < ₦10,000,000 | ₦10,000 |
Confidence
How confident is the estimate?
A 0–100 score combines evidence coverage, evidence confidence, recency, agreement between approaches and how many approaches were possible.
Confidence score components
| Component | Weight |
|---|---|
| Coverage | 0.3 |
| Evidence Confidence | 0.25 |
| Recency | 0.15 |
| Agreement | 0.15 |
| Approaches | 0.15 |
Agreement reaches 0 at 50% disagreement; a single approach scores 0.4.
Labels and caps
| Rule | Value |
|---|---|
| High confidence | score ≥ 70 |
| Medium confidence | score ≥ 45 |
| Low confidence | score < 45 |
| Fewer than 50% completed-sale comparables | score capped at 69 |
| Strong data coverage | ≥ 70% |
| Moderate data coverage | ≥ 40% |
Full-coverage targets
| Evidence | Observations |
|---|---|
| Land | 5 |
| Replacement Costs | 2 |
| Comparables | 6 |
| Rentals | 6 |
| Yields | 3 |
Data confidence levels
Every observation carries a confidence level
Assigned when evidence is recorded and shown next to every figure on our market pages.
| Level | Typical evidence | Weight |
|---|---|---|
| HighHigh | Verified supplier, government or statutory source, verified transaction, strong professional evidence | 1 |
| Medium-highMedium-high | Verified professional contributor, multiple independent observations | 0.85 |
| MediumMedium | Current verified listing, established agent submission | 0.7 |
| Low-mediumLow-medium | Unverified agent submission, single marketplace listing | 0.5 |
| LowLow | Anonymous or unverified observation, old stale data | 0.3 |
Limitations
What an automated estimate can't do
- Automated market estimate, not a statutory valuation. It is not a substitute for a valuation by a registered valuer.
- Asking prices are not completed sale prices. Where comparables are mostly asking prices, confidence is capped below high.
- No physical inspection: size, condition and features are taken as reported.
- Title, legal status and encumbrances are not verified by the estimate.
- Coverage varies by location. We launched in Edo State and say so wherever data is limited.
- Results reflect evidence available on the valuation date; markets move, so re-check before acting.
See the evidence for yourself on our market price pages.
Price check
Know what a property in Benin City is really worth.
We check it against land prices, building costs, similar homes and rents, and show you the numbers.