What is an AVM?

An Automated Valuation Model (AVM) is a computer program which uses statistical methods to value properties.  An AVM provides a best estimate of current value, plus an FSD (Forecast Standard Deviation) which is a measure of the accuracy of the estimate. The lower the FSD, the more confident the AVM is of the estimate.

The IDS AVM also produces an estimate of weekly rent and annualised rental yield for the property.

How does an AVM work?

Our AVM generates estimates from multiple sub-models and then blends them together to produce a final estimate for the property.  The end result is an algorithm which scores or ranks the target property against its recently sold or listed neighbours, then blends this price information with the target’s attributes and specific components of value from the statistical analysis, plus an indexed previous sale and/or current list price of the target (if available).

How we use historical sales data

For each historical sale in the database, we know the address, property type, sale price, sale date, land area, number of bedrooms, bathrooms and car spaces, plus the listing price prior to sale.

This lets us to determine, for each property type and suburb how, at any point in time, prices change on average with land area, number of bedrooms, bathrooms and car spaces.  For example, on average, how much extra was a 3rd bedroom or 2nd car space worth in July for Bondi units?

By analysing large numbers of sales, statistical methods allow us to separate the effects on price of the different components of value.  Moreover, we can see how those components of value have changed over time. For example, how the value of being in a particular suburb or the value of car parking in a suburb has changed over time.

How we use sales and listings data for the property

When valuing a property, professional valuers don’t just use the property's current attributes.  They also use the sales and listing history of the property, plus recent sales and listings of comparable neighbouring properties.  Our AVM does this too.

How should I interpret an AVM?

An AVM provides a property has a value estimate (AVM) and a % FSD. Think of them as together providing a price range which depends on the information available about the target property, including recent sales and listings of comparable neighbours, plus any previous sales of the target (and their age).

You should think of the price range as being the AVM ± 1 FSD. For example, an AVM of $800,000 with an FSD of 10% should be thought of as a range of $720,000 – $880,000, with a best estimate halfway in between ie. $800,000.

• A property might have an FSD of 5% because it has a more recent sale of its own and is more similar to recently sold or listed neighbours. That higher confidence will give its estimate a smaller range.

• A property in a metro fringe suburb, with a more sparse sales market and which last sold 20 years ago could easily have an FSD of 20%.

How you interpret an AVM depends on the purpose for which you are using it:

• Point of Sale - When using an AVM to advise a consumer about the value of their property for sale or loan application purposes, it is worth discussing the comparable sales the model used with the customer.  They may have a different view of what is comparable to their property.

• Decisioning - When using an AVM to make a credit decision, the lower end of the range should probably be used to calculate an LVR. If the AVM estimated LVR is above the policy threshold, the loan application or limit increase may then need to be escalated to a valuation that is performed by a professional valuer.

• Portfolio Valuation - When using an AVM to provide regular updates of the market value of a property over time eg. for a SMSF, it is best to use the AVM estimate, although showing the range in addition is sensible.

What if the attributes of the property recorded in your database are wrong?

We go to great lengths to ensure the maximum accuracy of the property data in our database.  However, property data is not perfect, or always up to date, so we sometimes get it wrong.

If you would like our AVM to estimate the value of the property based on different attributes, then you can provide our AVM with modified attribute values and it will provide an estimate based on those values.  This functionality is available through our API.