How we calculate our estimates

Where the estimate comes from

There is exactly one source of the headline estimate: the Funda WaardeCheck automated valuation model (AVM), the same consumer estimate Funda publishes. We send it the address, the living area, the plot area, the building type and the year of construction, and it returns a value, a lower bound, an upper bound and its own confidence level. We store all four unchanged. The number you see is Funda’s number, not a HomeWend calculation.

An AVM is a statistical model over registered transactions and property attributes. It is not a person and it has never seen the home. It also has no way to know the condition of the interior: we do not ask you to rate it and we do not guess, so every request is sent with a neutral condition assumption.

We do not use comparable sold transactions from the Land Registry (Kadaster). We used to. That pipeline was removed because the comparable sales we could obtain carried no genuine asking price, which meant they could not measure the one thing they were there to measure, namely how homes in an area actually price against what they were listed at. Rather than keep a table that looked like local evidence and was not, we took it out. Nothing in the product currently reads sold comparables.

Around the AVM we add two checks of our own. First, we compare the estimate to the asking price: when the two disagree by more than 20 percent we mark the estimate provisional, stop showing it as a bare number beside the asking price, and stop letting it influence the suggested bid. Mass-appraisal models price floor area rather than land, so they read low on a large plot or a free-standing house. Second, the code carries a WOZ (municipal assessed value) cross-check that flags a divergence beyond 35 percent. Be aware that we currently have no WOZ data source wired up, so in practice that cross-check does not run and no WOZ figure appears.

An estimate is fetched once and reused for up to two weeks, so the figure you are looking at may be up to fourteen days old. If Funda has no valuation for an address, or the listing is missing the living area or the year of construction, there is no estimate at all and we show none rather than inventing one.

• It cannot see inside the home. Renovations, finish quality, damp, a new kitchen or a dated bathroom are not inputs.
• It is not based on nearby sold prices. No Land Registry comparable-sales data reaches the estimate today.
• It cannot value an atypical home well. A large plot, a free-standing house or a one-of-a-kind property is where this kind of model is weakest, and we flag those rather than fix them.
• It is not live. The figure can be up to fourteen days old, and it does not react to what happened on the street last week.
• It is not available for every home. Where the model declines an address or the listing lacks the required facts, we show no estimate.

Why we show a range, not a single number

The low and high figures are the AVM’s own bounds, and the confidence label beside them is the model’s own assessment of how sure it is. We pass all of it through rather than collapsing it, because a single number implies a precision that no automated model has.

A range is also the honest shape of the answer. Two homes on the same street with the same floor area can sell tens of thousands of euros apart on condition, layout, orientation and how many people wanted them that month. A model that reports one number is not more accurate than one that reports a range. It is the same model with the uncertainty deleted.

Treat the range as a sanity check on the asking price, not as a target. If your thinking changes materially between the low end and the high end of the range, the estimate is not precise enough to settle the question and you need a valuer.

• The range is not a guarantee that the sale price falls inside it.
• A narrow range does not mean the estimate is correct. It only means the model was confident.

What the estimate is not

It is not a taxatie. A formal valuation in the Netherlands is a physical inspection by a qualified valuer, written up as a taxatierapport and validated by an institute, in practice the NWWI. That is a different exercise carried out by a person with professional liability, and it is the only kind of valuation a Dutch lender will lend against.

You cannot use anything on this page for a mortgage application. Your lender will require an NWWI-validated report, dated within six months of the reference date, from a valuer independent of the purchase, the sale and the financing. Our estimate has no standing in that process whatsoever.

It is also not advice. We do not tell you what a home is worth to you, what to offer or whether to buy.

• It cannot be submitted to a lender, and no lender will accept it.
• It cannot be used to challenge a valuation report or a WOZ assessment.
• It carries no professional liability. A valuer signs their report. A model does not.

How the bid confidence score works

The confidence score in the bidding simulator is a transparent weighted heuristic. It is deterministic, versioned, and not an AI model. It reads four things about the bid you typed, and nothing else: how your offer sits against the asking price, how much of your own money you are bringing, whether you keep the financing reservation (financieringsvoorbehoud), and whether you keep the building inspection clause (bouwkundige keuring).

The weights are fixed and public in that order of influence: price carries the most weight, then the financing condition, then own money, then the inspection clause. Price strength runs from two percent under asking (no strength) to eight percent over asking (saturated), so bidding further above eight percent stops moving the score. Own money saturates at roughly eight percent of the asking price. Keeping the financing reservation costs you less the less you actually rely on a mortgage, because a near-cash buyer has little to finance. The score is then clamped so it never reads as zero and never reads as certain.

The percentage figure shown next to it is derived from that same score plus a bounded nudge for bidding above the estimated value. It is a presentation of the score, not an independent forecast. Both numbers are capped so a weakly structured bid can never read as near certain, however high the price.

The crucial point: these constants were calibrated by hand against how competitive Dutch resale bidding generally works, in a market anchored to typical Randstad dynamics. They were not fitted to observed acceptances. We do record the outcome of a bid when you tell us what happened, but nothing feeds those outcomes back into the weights today. So the number is a way of comparing your own bid options against each other. It is decision support, not a measurement.

The monthly payment shown alongside is an indicative annuity calculation at a single fixed rate over thirty years. It is arithmetic on an assumed rate, not a quote, and the assumed rate is displayed so you can see what it was.

The risk warnings on a bid are separate and are always shown to everyone, on every tier. They describe consequences (what waiving the financing reservation exposes you to, what happens when your mortgage exceeds the appraised value) rather than recommending a course of action.

• It does not predict whether this seller will accept your bid. It has no information about this seller.
• It is not a probability derived from real seller behaviour. It is a weighted heuristic with hand-set constants.
• It does not know what anyone else is bidding, how many other bidders exist, or whether there are any.
• It cannot see the seller’s reserve price, their deadline pressure, or their preference for a particular completion date.
• It cannot tell you what you can afford. It uses the financing figures you type in and does not verify them.
• It does not improve as more people use HomeWend. Nothing in the product retrains these weights.

How the suggested bid range is built

The suggested range is a separate model from the confidence score, and it works the other way round: rather than scoring a number you typed, it proposes one. It starts from a price segment prior. Under the NHG limit the buyer pool is deepest and bidding over asking is normal, while above roughly one million euros the pool thins, listings sit longer and closing below asking is ordinary. Applying one national over-bid figure to both would be wrong, so the model uses a segment table instead.

The prior is then shifted by how long the home has been listed relative to what is typical for its segment, by the gap between the AVM estimate and the asking price (capped, and skipped entirely when the estimate is provisional or the home is atypical), and finally by your own stated financing position, which can only ever cap the suggestion downward. Every shift is shown to you with its own contribution in percentage points, and those contributions sum to the number, so the arithmetic on screen is the arithmetic the model ran.

Two things the model is built to use are not currently supplying anything. A "recent sales nearby" factor exists and is wired in, but with the comparables pipeline gone it never fires, so the range never reflects what homes near this one actually sold for. A WOZ factor exists for the case where a home is listed below its assessed value, and it never fires either, because we hold no WOZ data. In both cases the model reports them to you as missing rather than pretending they were considered.

The width of the range comes from how much the model knows, not from the price. Fewer usable inputs produce a wider band. Because two of the inputs above never arrive, the band you see is usually a wide one, which is the correct response to knowing less.

• It is not evidence about this specific home. It is a segment-level pattern adjusted by a handful of facts.
• It does not currently reflect actual sold prices in the neighbourhood, because no comparable-sales source is connected.
• It cannot tell you the winning bid. Sealed Dutch bidding means nobody outside the process knows that number.
• It is not a promise that a bid inside the range wins, or that one below it loses.

How we rank buying agents, valuers and mortgage advisors

None of these rankings are paid placements. The ranking code reads no payment, referral or commission field, and no such field is populated anywhere in the product. Nobody can buy a position, and we do not take a fee from any professional shown to you. Position is a function of the factors described here and nothing else. Every ranking is deterministic, which means the same inputs always produce the same order and the same stated reasons.

Buying agents are real offices from the NVM directory, synced daily and enriched with Google Places ratings where they exist. The score is an area-fit measure: how close the office is to the areas you are searching, whether it advertises buy-side work, the size of the office, and Google review quality as a bonus that only counts when there are enough reviews to mean something. An office with no reviews is not penalised for it.

Be aware of how thin the directory data is. For scraped offices we hold no fee, no languages and no specialisms, so the fee, language and case-fit factors in the ranking are constant across every office and separate nobody. In practice the order you see is driven by proximity, buy-side capability, office size and Google reviews. Fees are shown as "on request" because no office publishes one to us.

Valuers are real offices from the same directory that list taxation as a service, ranked by proximity to the home, any taxation credential the directory lists, whether the office does valuation work only, its size and its reviews. We do not verify NWWI registration, and we make no claim about price or turnaround, because we hold no data for either. Check the register yourself before you book.

Mortgage advisors are a curated list of real Dutch firms that we maintain by hand. They are ranked on how well their stated specialisms match your financing situation (an expat purchase, the 30 percent ruling), on review reputation weighted by review volume so that a high rating backed by thousands of reviews outranks the same rating backed by dozens, and on whether an independent broker or a single bank suits your stated preference for rate versus speed. Without your consent to use your financing profile, the list is not personalised at all and is simply ordered by reputation.

• A ranking is not a recommendation, a vetting or an endorsement. We have not audited anyone’s work.
• We do not verify NWWI registration, AFM licensing, insurance or credentials. Check the public registers yourself.
• Buying-agent fees are not shown, because no office publishes one to us. Ask directly and get it in writing.
• Ratings are third-party Google reviews. We neither collect nor moderate them.
• The directory can be out of date or incomplete. An office missing from the list is not a judgement on it.

What we do not know

Some of the most important facts in a Dutch purchase are ones no outside party has, and we would rather name them than let a confident interface imply otherwise.

We do not see other buyers’ bids. Dutch bidding is sealed, so there is no feed of competing offers, no count of how many people bid, and no way to know the winning number afterwards. Anyone claiming to model this is guessing.

We do not see the seller’s side at all: their reserve price, their instructions to the selling agent, their timeline, or how they weigh conditions against price. We do not see the condition of the interior, any structural problem, ground lease terms, VvE finances for an apartment, or anything a survey would find. And we hold no WOZ data at present.

• We cannot tell you what the other bids are, or whether there are any.
• We cannot tell you what the seller will accept.
• We cannot tell you what the home is worth after the survey finds something.

Who to rely on for the decision

HomeWend organises a purchase. It does not replace the people who carry professional responsibility for it, and it is not financial or legal advice.

For what you can borrow and what a mortgage will actually cost you, rely on a mortgage advisor. For what a home is formally worth, rely on a qualified valuer and their NWWI-validated report. For the deed, the transfer and what you are signing, rely on the notary. Those three are the people whose figures carry weight, and any of them may reasonably disagree with a number on this site.

One consequence worth stating plainly: an accepted bid in the Netherlands is a binding purchase agreement. Discuss the number and the conditions with your buying agent or advisor before you send anything.

• HomeWend is not a mortgage advisor, a valuer, a notary or an estate agent, and is not regulated as any of them.
• Nothing in the product is financial, legal or tax advice, and none of it should be the last thing you check before you bid.