Home

Glossary of terms

Key concepts and definitions in the ASINT system

Automated Valuation Model (AVM)

a system of mathematical and statistical algorithms that computes an asset's value at a given point in time without human involvement once the process is started.

Asset

any item of economic value owned by a person or company; in this context primarily real estate and vehicles that confer the right to economic benefit.

Algorithmic Bias

the risk that racial or socio-economic inequality arises in automated systems by repeating historical patterns of discrimination, even when such parameters are not entered directly.

Depreciated Replacement Cost (DRC)

the current cost of creating a modern equivalent of the asset, less depreciation (physical deterioration) and obsolescence.

Comparable Sales (Comps)

recently sold properties similar to the subject asset in location and characteristics, used as the basis for adjustments in market valuation.

Ensemble Methods

machine-learning methods that combine several decision trees (such as bagging, boosting and random forest) to improve the stability of predictions.

Basis of Value

the fundamental premise on which the estimated value rests (e.g. market, investment or equitable value).

Boosting

an ensemble-building method in which models (trees) are created iteratively and weights are adjusted to correct the errors of earlier stages.

Bagging / Bootstrap Aggregating

a machine-learning method that uses random independent samples of data to reduce noise and instability in predictions.

P(Gain)

a financial metric that estimates the probability of recovering the invested capital and earning a positive return.

Volatility

a statistical measure of the dispersion of an asset's prices or returns, used to assess market risk and uncertainty.

Input

information, assumptions or adjustments that a specialist considers significant for quantifying value.

Hedonic Pricing Model

a theoretical framework that treats an asset as a bundle of characteristics and estimates the contribution of each individual attribute to the total value.

Geographically Weighted Regression (GWR)

a local regression method that accounts for spatial heterogeneity by varying coefficients according to geographic location.

GeoScore

a rating of the quality of a property's location, based on infrastructure, safety, ecology, proximity to services and climate risks.

GIS (Geographic Information System)

a computer system for collecting, storing and analysing data tied to geographic coordinates on the Earth's surface.

Hypothetical Condition

a condition that does not correspond to reality on the valuation date but is assumed for the purposes of analysis (e.g. valuing an unfinished property as completed).

Deep Learning

advanced machine learning that uses multi-layer neural networks to process massive and complex datasets.

Data

quantitative and qualitative information available to an appraiser or algorithm for forming a conclusion about value.

Call Detail Records (CDR)

mobile-network data that allows analysis of population mobility and social behaviour to forecast property prices.

Attributes Data

a property's characteristics, including both continuous variables (area, age) and categorical data (number of rooms, presence of parking).

Dynamic Market Indicator (DMI)

an asset's estimated market value, updated in real time on the basis of new data.

Dynamic Market Rental Indicator

a property's estimated monthly rental value, computed by algorithms in real time.

Confidence Interval

the range of values within which the asset's true value lies with a given probability.

Assumption

a supposition taken as true and not requiring verification by the appraiser.

Income Approach

a valuation method in which value is determined by converting forecast future income or cash flows into a single present amount.

Opinion of Value

the final result of a professional valuation carried out by a specialist in person, as opposed to the "estimated value" produced by an AVM.

Cost Approach

a method based on the principle of substitution: a buyer will not pay more for an asset than the cost of acquiring or building a property of equivalent utility.

SHAP values (Shapley Additive Explanations)

a method for interpreting machine-learning ("black box") models that computes the contribution of each feature to the final prediction.

Investment Value / Worth

the value of an asset to a specific owner or investor, based on their individual objectives or operational needs.

Indicative Value

a preliminary, quick and inexpensive estimate of value based on basic information.

Mortgage Lending Value (MLV)

value determined through a prudent assessment of a property's future marketability, taking long-term sustainable factors into account.

Computer Vision

the field of AI that uses deep learning to extract metadata from images (e.g. to assess the condition of a finish or a facade).

Quality Control (QC)

procedures to ensure the reliability of AVM valuations, protect against data manipulation and eliminate conflicts of interest.

Disparate Impact

a situation where seemingly neutral rules or algorithms have a disproportionately negative effect on protected groups of the population.

R-squared

a statistical measure showing the proportion of the variation in the dependent variable (price) explained by the independent variables in the model.

Coefficient of Dispersion (COD)

a measure of the uniformity of valuations, expressed as the average percentage deviation from the median assessment-to-sale-price ratio.

Liquidation Value

the amount that can be obtained from selling an asset under conditions where the seller is forced to complete the transaction within a short time.

Marginal Price

the change in market value caused by changing one characteristic of a property by one unit of measurement.

Mass Appraisal

the process of valuing a group of properties on a specific date using standardised statistical methods.

Mahalla

a traditional residential neighbourhood in Uzbekistan, used as an important micro-location factor in determining value.

Multiple Regression Analysis (MRA)

a statistical method for determining the relationship between price and several independent property characteristics.

Multimodal Machine Learning

an approach that integrates different types of data (tables, text, images, GIS) to improve valuation accuracy.

Observable Data

information about real events or transactions that is available to market participants.

Highest and Best Use

the use of an asset that is physically possible, legally permissible, financially feasible and produces its maximum value.

Intangible Asset

a non-monetary asset with no physical form (e.g. goodwill, brands or domain names).

Neoclassical Economic Theory

the foundation of valuation theory, based on the interaction of supply, demand and equilibrium in a competitive market.

Residuals

the difference between the actual sale price and the value predicted by the model; they represent the unexplained portion of the variation.

Valuation Model

the quantitative implementation of a valuation method, converting inputs into an output to form a conclusion about value.

Appraisal Function

a mathematical relationship (often linear) establishing a cause-and-effect link between price and a property's characteristics.

Late Fusion

a data-processing method in which different types of information are processed independently and their results are combined at the final stage.

Bias

a preference or inclination that impedes impartiality, independence and objectivity in carrying out a valuation.

Forecast Standard Deviation (FSD)

a measure of statistical uncertainty showing the AVM's level of confidence in a specific valuation.

Professional Judgement

the use of accumulated knowledge and logic to make well-founded decisions during the valuation process.

Disparate Treatment

the deliberate infringement of a person's rights by an appraiser or model on a protected characteristic (e.g. race or religion).

Equitable Value

the estimated transaction price between specific, informed and willing parties, reflecting their individual interests rather than the market as a whole.

Early Fusion

a method in which raw data from different sources is combined into a single representation before the model begins processing it.

Non-Disclosure State

a jurisdiction where transaction prices are not officially published, forcing an AVM to rely on indirect data.

Elo Rating

a measure of visual quality obtained through pairwise comparisons; it turns subjective judgements about images into numerical data.

Risk of Decline / Depreciation

the probability that an asset's value decreases over time due to market changes, physical wear or economic obsolescence.

Market Rent

the estimated amount for which a property could be rented on the valuation date on market terms.

Market Area

a geographic area that groups homogeneous market segments for applying a model.

Market Value

the most probable sale price of an asset in an open and competitive market between a willing buyer and seller.

Synergistic Value / Marriage Value

additional value that arises when two or more assets are combined and their total value exceeds the sum of the separate parts.

Control Systems

internal audit and quality control used by an organisation to assess the effectiveness of valuation processes.

Random Forest

a machine-learning model that uses an ensemble of decision trees and averages their results to improve accuracy.

Special Assumption

an assumption that differs from the real facts on the valuation date, often used to analyse "what-if" scenarios.

Sales Comparison Approach

a valuation method based on comparing a property with similar sold assets and making adjustments for differences.

Fair Value

under IFRS 13, the price that would be received on selling an asset in an orderly transaction between market participants on the valuation date.

RMSE (Root Mean Square Error)

a measure of the magnitude of errors between the model's predicted values and the actually observed data.

MAE (Mean Absolute Error)

a common AVM quality metric representing the average magnitude of errors across a set of predictions.

MAPE (Mean Absolute Percentage Error)

a measure of forecast accuracy expressed as a percentage; it allows comparison of different models or markets.

Days on Market (DOM)

the number of days a property stays for sale until the contract is signed; an indicator of liquidity and demand.

Discount Rate

the rate of return used to bring future cash amounts to their present value.

Capitalization Rate

the ratio of a property's income to its market price, used to convert income into a capital value.

Similarity

a quantitative measure of the comparability of the subject property and a comparable, needed to select appropriate transactions.

Stochastic Modeling

an analysis method that uses probability distributions instead of static values, allowing a range of outcomes to be computed with a confidence level.

Relevant Characteristics

features that affect a property's value or liquidity (legal, economic, physical).

Current Value

the most recent valuation of an asset at the time of valuation, reflecting current market conditions.

Occupancy Type

a classification of how a property is used (owner-occupied, tenant-occupied or vacant), which affects value and risk.

Property Type

a classification of an asset by purpose or physical properties (residential, commercial, industrial or land).

Accuracy

a characteristic of data and models indicating the absence of errors and bias and the correct reflection of the properties being measured.

Price

the specific amount requested, offered or paid for an asset; unlike value, price is a historical fact.

Extraordinary Assumption

an assumption directly related to a specific assignment, the falsity of which could materially change the appraiser's conclusions.

Management Fees / Operating Fees

monthly contributions by owners for the upkeep of common areas in residential complexes or gated communities.

Effective Age

the age of an asset based on its current condition and utility rather than its actual construction date.

BERT (Bidirectional Encoder Representations from Transformers)

a modern natural-language-processing method used to turn textual property descriptions into numerical vectors.

iBuyer (Instant Buyer)

a business model in which companies use an AVM to buy homes directly from sellers with a view to reselling them.