Online apartment valuation: what an AVM can do and where it stops

Online apartment valuation promises an answer in a minute: enter the address, area and number of rooms - and get a price. It’s convenient, and such tools really do save time. But automated valuation has limits that are rarely spoken about: it can’t see the renovation, doesn’t know the view from the window and works with listing prices, not sale prices. Let’s look at how an automated valuation model (AVM) works, what it does well, where its abilities end and how to use it so the figure helps rather than misleads.
What an AVM is and how it works
An AVM is an automated valuation model. It almost always rests on the same comparative method a realtor uses, only faster and on more data. The model finds apartments similar to yours - the same district, a close area, the same number of rooms, a similar building type - and looks at the price per square metre they’re listed at. From that sample comes a benchmark: the median and the range in which most similar offers sit.
The model works only with what’s in the data: district, area, rooms, floor, building type, price and posting date. Preparing that data matters too: if one apartment hangs in five listings, an honest model merges the duplicates first; otherwise one expensive apartment “counts” five times and shifts the benchmark.
The main advantage of this approach is volume. A person looks at a dozen listings in an evening; the model takes into account every suitable offer on the market and updates them constantly. The city benchmark that any valuation starts from looks like this:
What automated valuation does well
- Fast: a benchmark in a minute instead of an evening with listings.
- Unemotional: the model doesn’t care how much the seller put into the apartment or how attached they are.
- Market-wide: every suitable listing is counted, not just the ones you happened to see.
- Up to date: prices change, and the benchmark is recalculated with the market.
- Transparent, if the model is honest: you can see which comparables went into the sample.
These qualities make an AVM an excellent starting point. It’s especially useful when you need a quick sense of the price level: an owner thinking about selling, a buyer estimating a budget, a realtor preparing for a first meeting.
Another useful application is keeping your own expectations in check. A seller who has lived in an apartment for years finds it hard to see it through the market’s eyes. A buyer who has fallen for an option finds it hard to notice that it’s pricier than the neighbours. The model’s cool benchmark helps both stop in time and ask the right questions.

Where automated valuation stops
Every model has things it doesn’t know. It’s important to understand them so as not to mistake a benchmark for the apartment’s final price.
Scroll the table sideways
| Limit | Why | What to do |
|---|---|---|
| Condition and renovation | WhyThe model knows the area, not the quality of the finish | What to doAdjust by hand using comparables in similar condition |
| Floor, view, layout | WhyThese features are often not described in listings | What to doCompare within one building or district |
| Unusual properties | WhyToo few similar apartments to compare | What to doWiden the sample carefully and check by hand |
| Asking prices | WhyThe model sees listings, not sale amounts | What to doAllow for bargaining and days on market |
| Sharp market changes | WhyListings react to change with a delay | What to doLook at fresh listings and the trend |
None of these limits makes an AVM useless. They just show where the model’s figure needs a human eye.
The case of a small sample is especially important. If there are few similar apartments in a district, an honest model doesn’t give a confident figure but says plainly: there isn’t enough data. That’s frustrating, but far more useful than a neat number calculated from three random listings.
A range instead of a single figure
Good automated valuation shows not one price but a range: from the lower quarter of similar offers to the upper. That’s more honest, because similar apartments always have a spread, and your apartment’s real price depends on where it sits within that spread.
One figure
“Your apartment is worth X”
- Creates false precision
- The seller clings to the number and won’t hear arguments
- It’s unclear what it’s based on
- You can’t see how big the spread is
Value for the decision an illusion of an answer
A range and comparables
“Similar apartments cost from A to B”
- Shows the market’s real spread
- You can see which apartments are in the sample
- It’s easy to explain why yours is nearer the top or bottom
- Adjustments for specific differences can be discussed
Value for the decision a basis for conversation
Where your apartment lands within the range depends on how it differs from the sample. A fresh renovation, a good floor, a nice view and a well-kept building pull the price towards the top. Old finishes, the ground floor, a noisy side and an entrance in need of repair pull it towards the bottom. These adjustments are the human work on top of the model.
How to use online valuation properly
Automated valuation is the first step, not the last. Take a few more simple steps and the benchmark turns into a justified price.
From the model’s benchmark to a price
How to work with an online valuation
Get a benchmarkA minute
A price range from comparable properties by district, rooms and area.
Look at the comparablesFive minutes
Which apartments made it into the sample and whether they really resemble yours.
Make adjustmentsTen minutes
Floor, condition, view, layout - what the model doesn’t see.
Check against timeFive minutes
How long apartments at this price take to sell and whether they’ve been cut.
DecideResult
A price at the upper edge of the real range, not above it.
How to value an apartment against comparables by hand, with adjustments for floor, renovation and building, is covered in detail in our step-by-step valuation article. And when a seller is deciding whether to renovate for the sake of the price, our article on renovating before a sale will help.
The AVM and the realtor
Automated valuation doesn’t replace a realtor, but it makes them stronger. The model takes away the routine - gathering comparables and calculating the median - and the realtor adds what the model can’t see: the apartment’s condition, its strengths and weaknesses, the seller’s situation.
Valuation is especially useful in a price conversation. When a seller is convinced their apartment is worth more than the market, a range based on comparable properties moves the argument from “I feel” to “this is what’s selling nearby”. How to have that conversation is described in our article on overpricing. And how long similar apartments take to leave the market is visible in the market mirror - the second half of the argument.
And the other way round: a realtor who disagrees with the model has to be able to explain why. “This apartment is above the range because it has a new renovation and a park view” is a good argument. “I feel it’s worth more” isn’t. The model disciplines both sides of the conversation.
Honest caveats
- The valuation is built on asking prices: the final sale price is usually lower by the amount of bargaining.
- With a small sample, the model’s right answer is “not enough data”, not a confident figure.
- The model doesn’t know the apartment’s legal specifics: encumbrances, debts, registered residents.
- The benchmark is useful for a decision but doesn’t replace a valuation for a bank or a notary.
- If the comparables in the sample don’t resemble your apartment, the figure can’t be trusted - check the sample.
- A valuation last week and today may differ: the market moves, and the benchmark with it.
The live figure is the median asking price per m² for apartments for sale in the Titul mirror; these are listing prices, not transaction prices. With a small sample the benchmark isn’t shown as confident - that’s part of data honesty.


