Every week, sellers in the Perth Hills contact me with a number in their head. Sometimes the number is from realestate.com.au. Sometimes it’s from Domain, Cotality, or a link their bank sent them. Sometimes it’s from all of them — and the numbers are $100,000 apart from each other.
So how accurate are online property valuations, really? After more than two decades selling homes across Bedfordale, Roleystone, Kelmscott, Mount Nasura, Mount Richon and Seville Grove, the honest answer is: it depends entirely on the property, the suburb, and what the algorithm can and can’t see.
In some cases, the online estimate is close enough to use as a rough starting point. In other cases, it’s out by hundreds of thousands of dollars. And the difference between those two outcomes is almost never something the seller can predict from looking at the website.
Let me show you three real examples from properties I’ve sold.
Three Properties Where the Algorithm Got It Badly Wrong
Churchman Brook Road, Bedfordale
Both realestate.com.au and Cotality valued this property at around $1.2 million.
We listed it from $2.5 million. It sold for $3.1 million — almost $1.9 million above the automated estimate.
What did the algorithm miss? Three things that completely changed the value of the property:
– The block had mains water connected, which is rare on acreage in this area
– It was subdivisible into three separate blocks
– The blocks were dead flat — meaning the earthworks cost for any future subdivision was a fraction of what it would be on a sloping block
No algorithm can read a contract of sale, walk the land, check a deposited plan, or assess subdivision potential. It can only see what’s in the public record about square metres and recent sales. So it valued a near-million-dollar opportunity at the price of a standard residential block.
Urch Road, Roleystone
The automated valuation systems had this property at around $800,000. We listed from $900,000. It sold for $1.1 million.
The algorithm missed two things that any buyer noticed within ten seconds of walking through:
– Stunning views down the valley
– An exceptional internal finish — the kind of presentation and detail that makes a buyer want to write a cheque on the spot
You cannot put a price on the feeling a buyer gets when they walk into a beautifully finished home with a view they fall in love with. No algorithm can see views. No algorithm can see craftsmanship. They see square metres.
Blackwood Drive, Mount Nasura
Automated valuation: $750,000. We listed from $800,000. Sold for $926,000.
This property was:
– Fully renovated — exceptional presentation
– Had outstanding views
– Walking distance to local primary schools
– On the main bus route
Walkable schools, transport convenience, and a fully refreshed home are all factors that drive premium buyer demand. Algorithms don’t measure school catchments by walking distance. They don’t grade renovation quality. They average everything out.
How These Algorithms Actually Work
To understand why the algorithm gets it wrong, it helps to know what it’s actually doing under the hood.
Most automated valuation models (AVMs) work on a fairly simple formula. They take:
– The square metreage of the home’s living area
– The block size
– Land valuations of comparable blocks that have sold nearby
Some systems also use a build-cost-per-square-metre approach minus a depreciation figure, similar to what a licensed valuer might do.
Where this works reasonably well: cookie-cutter suburbs where the homes are nearly identical. A 220-240sqm four-bedroom-two-bathroom home on a 500sqm block in an estate where the only real differences between properties are paint colour, floor coverings and furniture — the algorithm can get within a workable range.
Where it falls apart: the Perth Hills, and any suburb where the homes are genuinely different from each other.
Here’s a partial list of what the algorithm cannot see on a Hills property:
– Whether the block is level or sloping — and how steep that slope is
– Side access — yes or no
– Views — and crucially, whether they face west into the afternoon sun or north toward the city
– Water connection — mains, tank, or bore
– Whether the property allows you to keep pets, horses, or livestock
– Whether you can store trucks or run a business from the property
– The presentation and finish of the home
– The quality of the street and the surrounding area
– Whether there’s a workshop — and if so, whether it has power, and whether that’s three-phase or single-phase
– Pool, established gardens, usable outdoor space
– Proximity to schools, hospitals, transport
– Which buyer group the home actually suits — first home buyer, upsizer, downsizer, or renovator
None of these factors appear in any AVM. Yet in the Hills, every single one of them can shift the sale price by tens of thousands — sometimes hundreds of thousands — of dollars.
There’s one more critical limitation worth understanding. AVMs rely on comparable past sales as the backbone of their estimate. But the algorithm can’t see what those “comparable” properties were actually like inside, in condition, in view, or in any of the factors above. So it’s comparing a flattened version of your home to a flattened version of someone else’s. Two unrelated properties are forced into a comparison they were never actually similar in.
Which AVMs Are Best — and Which Are Worst?
Honest answer: they’re all unreliable, but some are worse than others.
In my experience selling Perth Hills properties, the more useful ones tend to be Cotality (formerly CoreLogic) and realestate.com.au. They’re not accurate enough to base a pricing decision on, but they at least sit closer to reality more often than the alternatives.
The worst ones are usually the in-house systems some banks have built. These often produce numbers that bear no relationship to the actual market — and worryingly, banks then use those numbers for finance and bridging loan decisions.
There’s another problem worth knowing about. Run the same property through the same AVM on two consecutive days and you can get completely different numbers. I’ve seen properties where one day the AVM says $900,000 and the next day it says $1.2 million. That’s not a small error — that’s a $300,000 swing on a property that hasn’t physically changed.
I know one property in our area that recently sold for $3.5 million. The AVM for that property today, AFTER the sale has been recorded in the public register, still says $1.5 million. The algorithm can’t even self-correct from publicly available sale data.
The Algorithm Doesn’t Even Agree With Itself
Here’s something easy for you to test on your own property right now. Look up your home on any of the major AVM sites, and you’ll see they don’t give you one number — they give you a low estimate, a high estimate, and an “expected” price in the middle.
That range is often 10% to 20% wide.
On a $900,000 property, that’s a spread of $90,000 to $180,000 between the algorithm’s own low and high figures — for the same property, on the same day, from the same system.
Think about what that tells you. The algorithm itself isn’t confident in its own answer. It’s giving you a band so wide that you could drive a truck through it, and somewhere inside that band is supposedly the value of your home.
If the system that built the number isn’t sure to within $100,000-plus, why would you make a six- or seven-figure decision based on it?
Even the AVM Companies Tell You Not to Rely on the Number
This is the part most sellers never read.
I’ve called these AVM providers directly when I’ve seen valuations that were significantly out. The standard response, every time, is essentially the same:
“It’s only an algorithm working on the square metreage of the house and the block. Our terms and conditions clearly state that the valuation should not be relied on for the value of the house, or as a market value for a bank.”
Read that again, because it’s important.
The companies that produce these AVMs explicitly disclaim, in their own terms and conditions, that the figure should not be relied on to value your home — and not relied on by banks as a market value.
It’s in writing. You can go and read it yourself on any of these sites. Look for the fine print at the bottom of the valuation, or in their terms of use. The disclaimers are there.
If the people who built the tool are telling you not to trust the number, that should be the end of the conversation.
The Real-World Consequences of Trusting an AVM
This is where it stops being theoretical.
Underpricing — costing you tens of thousands. Some agents will use a low AVM as cover for listing your property cheap so they can sell it quickly. Convenient for them. Expensive for you.
Overpricing — costing you the sale. It’s not always that AVMs come in low. Sometimes they come in exceptionally high. A weak agent will see your inflated AVM and say “well, let’s try that price and see how it goes” — without explaining what happens when a home sits on the market too long. Your home goes stale, buyers wonder what’s wrong with it, and you end up below where you should have been.
Insurance underinsurance. If you insure your home at the AVM figure and your home needs to be replaced after fire or flood, you could be $200,000-$300,000 short of what you need. In the Hills, site works alone can range from $100,000 to $250,000 depending on slope and access to services — before you’ve even laid a brick.
Tax consequences. This one catches sellers regularly, particularly on acreage properties. Here’s a typical example, and please remember this is general information only — not tax advice — and you must speak to your accountant before making any decisions.
If you have a 10-acre property and it’s your principal place of residence, the first 5 acres are tax-free, and capital gains tax applies to the other 5 acres. The AVM gives you one number for the whole property. But a skilled agent can break that down properly: the house block might genuinely be worth $750,000-$850,000 on its own, and the back block (especially if landlocked) might be worth significantly less than half the total. Splitting the AVM number in two evenly — which is what many people do — can mean paying tax on money you didn’t need to.
Family disputes and divorce. AVMs hold no weight in court. If your property is part of a settlement, the court will require a licensed valuation, and often will ask for three real estate agent appraisals as well. The AVM is irrelevant to the actual legal outcome — but I’ve seen sellers and their families make poor decisions in the meantime based on those numbers.
A Cautionary Tale from Roleystone
A while back I appraised a property for a woman in Roleystone who was downsizing.
Her daughter was helping her with the sale. The daughter had already taken out bridging finance on a property in another suburb based on the bank’s valuation system, which had the Roleystone home at $1.1 million to $1.2 million. The agent they were leaning toward had appraised it at $1.25 million.
I appraised it at $850,000 to $900,000 — which is genuinely where the market sat for that property.
The daughter would not accept my appraisal. She trusted the bank’s valuation system. They went with the agent who matched the AVM.
The property went on the market at $1.25 million. Four months and multiple price drops later, it sold for $875,000.
That’s below the bottom of my original appraisal range. The home went stale chasing a number that was never real. Meanwhile the family had four months of bridging finance interest stacking up on the daughter’s other property.
Everyone in that family was acting in good faith. The daughter was trying to protect her mum. But the AVM and the agent who matched it caused real, expensive damage.
So What Should You Actually Do?
If you’re thinking about selling, here’s the honest path forward.
Use AVMs only for very early curiosity — say you’re 12 to 18 months out from selling and you just want a rough sense of where the market sits. Even agents look at AVMs to see what buyers will see when they search your property online. So they’re not useless. They’re just not the basis for a decision.
When you’re getting serious — typically around six months out, sometimes sooner — get at least two, ideally three real estate agent appraisals. Free, no obligation, takes 15-30 minutes per agent.
Do not pick the agent who quotes the highest number. This is the single most important point in this entire article.
Many agents are trained in a technique called conditioning. They tell you what you want to hear to get you to sign up for 90 or 120 days. Around week six, they tell you the market has changed and you need a “price adjustment.” Now your home has been on the market for six weeks with the wrong price, it’s gone stale, and you’ve lost negotiating leverage.
A well-priced home in the Perth Hills should sell within three to four weeks. If it’s not selling in that window, the price was wrong from the start — and the agent who told you the high number knew it.
The right pricing strategy is to meet the market — or, even better, to price slightly below the market to create genuine buyer competition.
I’ll be blunt about something most agents won’t say out loud:*you and the agent don’t set the price of your home. The buyers do. There’s no retail price on a property. There’s no manufacturer’s recommended figure. If an agent guarantees you a specific price, they’re either lying to win your listing or they don’t understand the market.
What you can do is set up a process that lets buyers reveal what they’re actually willing to pay.
Why the Select Date Sale® Method Solves This Problem
Buyers buy on feeling. They walk in the door, and they either fall in love with the home or they don’t. No algorithm captures that. No fixed asking price captures it either.
Our Select Date Sale® method is built specifically to solve this. It lets buyers compete with each other to pay what THEY are willing to pay — without knowing what other buyers are offering.
That last part is the critical mechanism. In a normal sale process, if a buyer knows what other buyers have offered, they only need to bid slightly higher. You never find out what they were actually willing to pay. The gap between “slightly higher” and “their genuine top price” can be $10,000, $20,000, $30,000 — sometimes $100,000 or more.
The competition should be between the buyers. Not between you and the buyers.
In a traditional sale, the buyer offers low, the agent suggests meeting halfway, the seller agrees, the buyer pushes again, halfway again — and three rounds of “halfway” later, you’ve quietly given away 10-15% of your home’s value. You weren’t negotiating with the market. You were negotiating against yourself.
Select Date Sale® takes the seller out of that fight. The buyers compete blind. The market reveals the real number — which is almost always significantly higher than the AVM, and often higher than even the most optimistic agent appraisal.
The Bottom Line
Online property valuations are okay for a very rough idea. They’re built on data and algorithms that can’t see most of what makes a home valuable — and in the Perth Hills, that’s almost everything. They can be 10%, 20%, or even 50% out from the actual sale price.
So why wouldn’t you make a phone call?
A real appraisal from an experienced local agent takes **15 to 30 minutes**. It costs you nothing. You get a properly considered figure based on what your specific home is actually worth in today’s market — including all the factors no algorithm will ever see.
Get a Real Answer from a Real Agent
If you’ve got a number in your head from an online valuation and you’re not sure whether to trust it — give us a call, send an email, or drop a text.
We’ll come out, have a look, and give you an honest read on what your property is genuinely worth in today’s market. No pressure, no obligation. It doesn’t matter if you’re thinking of selling next month, next year, or just curious for now — that’s what we’re here for.
Brendan Leahy, Naked Real Estate
08 6254 6333
brendan@nakedrealestate.com.
Unit 1/198 Brookton Highway, Kelmscott WA 6111