What is an automated valuation model (AVM) and is it accurate?

Type an address into a property portal and a value appears almost instantly. Behind that number sits an automated valuation model, or AVM: software that estimates what a home is worth from data alone, with no one walking through the door. AVMs power the consumer estimates buyers obsess over, the valuation checks lenders run on low-risk loans, and the engines agents use to start a pricing conversation. They are fast, cheap and everywhere. They are also frequently misunderstood. This guide explains what an AVM actually does, how accurate it really is, and when you should trust the number, with a worked example.

What an AVM is and how it works

An AVM is a statistical model that turns property data into an estimated value. It does not inspect the home or judge whether the kitchen is dated or the view is blocked. Instead it draws on three ingredients: recent comparable sales in the area, the attributes of the subject property (living area, plot, bedrooms, build year, location), and a model that has learned how those attributes map to price across thousands or millions of past transactions. Most AVMs blend two classic approaches: a hedonic model that prices each feature of a home, and a comparable-sales model that anchors the estimate to nearby transactions. The output is a single value, ideally accompanied by a confidence score and a value range.

AVM vs CMA vs appraisal: three different tools

These three get muddled constantly, so it is worth being precise. An AVM is an instant, data-only estimate with no human and no certification. A comparative market analysis (CMA) is an agent's market estimate, built by hand-picking comps and adjusting for differences, used to price listings and advise on offers. An appraisal is a certified valuation from a licensed appraiser who inspects the property and carries legal responsibility for the figure, usually required before a lender funds a mortgage. Roughly speaking, the AVM is the fastest and cheapest, the appraisal is the most rigorous and legally weighty, and the CMA sits in between as the everyday working tool. If you want the full method behind a CMA, see our guide on how to create a CMA.

How accurate is an AVM, really?

The honest answer is: it depends, and the spread is large. AVM accuracy is usually described with two numbers. The first is the median error, which tells you how far off the typical estimate is. The second is the hit rate, the share of estimates that fall within, say, 10% of the eventual sale price. In dense urban markets full of recent, near-identical sales, a strong AVM can post a median error in the low single digits. In rural areas, for unique or heavily renovated homes, or in fast-moving markets where last quarter's comps are already stale, the error can climb to 10-20% or beyond. The model is only ever as good as the data feeding it: thin or outdated transaction records mean a confident-looking number that is quietly unreliable.

This is why a serious AVM never gives you a bare figure. It gives you a confidence score and a value range. A range of plus or minus 3% says "trust this." A range of plus or minus 15% says "treat this as a rough starting point and do real work before you rely on it." Ignoring that confidence signal is the single most common AVM mistake.

A worked example

Suppose three different AVMs value the same three-bedroom house. The figures here are illustrative, not market data, but they show how to read the output:

• AVM A: €418,000, confidence high, range €405,000 to €431,000 (about plus or minus 3%).
• AVM B: €395,000, confidence medium, range €360,000 to €430,000 (about plus or minus 9%).
• AVM C: €440,000, confidence low, range €385,000 to €495,000 (about plus or minus 12%).

The naive reading is "the house is worth somewhere between €395k and €440k, so call it €418k." The smarter reading weights by confidence: AVM A's tight, high-confidence range is doing most of the real work, while AVM C's wide, low-confidence estimate is barely more than a guess. You would anchor on roughly €410,000 to €420,000 and then verify with hand-picked comparable sales before advising anyone. The disagreement between the models is not noise to average away; it is a signal that this property deserves human attention.

When to trust an AVM, and when not to

Lean on an AVM when the property is mainstream (a standard home in a liquid market), when you need a fast first number to triage a lead or sanity-check an asking price, and when the confidence score is high. Be far more cautious when the home is unusual, recently and heavily renovated, in a thin or volatile market, or when the AVM itself flags low confidence. And never present an AVM figure to a client as if it were an appraisal: it has no inspection behind it and no legal standing. Used as a starting point it is brilliant; used as a final answer it gets people into trouble.

How agents use AVMs without outsourcing their judgement

The professionals who get the most from AVMs treat them as an accelerant, not an autopilot. The AVM does the cold-start work in seconds: it pulls the data, surfaces the comparable sales and proposes a value. The agent then does what software cannot: walks the street, judges condition, knows that the quiet side of the block sells for more, and hand-selects the comps that genuinely match. That is the move that turns a raw estimate into a defensible price. If you want the discipline behind choosing comps well, our guide on finding comparable sales covers the three filters that matter most.

This is exactly where Biedradar fits. You enter an address and the platform runs the automated valuation, pulls comparable sales and market signals, and produces a branded property analysis report in minutes rather than hours. The AVM gives you speed; the report gives you something client-ready; and your local judgement stays where it belongs, on top of the data instead of buried under spreadsheet formatting.

The bottom line on AVMs

An automated valuation model is a powerful, instant estimate of a home's value built from data and statistics. It is not magic and it is not an appraisal. Its accuracy swings with the quality of the underlying data and the type of property, which is why the confidence score matters as much as the headline number. Read AVMs as a fast, honest starting point, verify the ones that matter with real comps and local knowledge, and you get the best of both worlds: the speed of the machine and the judgement only a professional can supply.

Frequently asked questions