Google Ads for Real Estate: Buying Intent, Not Browsers

The short answer

Real estate is the vertical where the gap between an enquiry and a transaction is widest, and where the largest share of the available traffic is people who enjoy looking at property and have no intention of buying or selling anything. Both of those facts point at the same conclusion, and it is not the one most property advertising is built around.

The job of a property ad account is not lead generation. It is filtering — distinguishing the small number of people with a live transaction from the very large number without one, as cheaply and as early as possible. Every structural decision in the account should be read against that: does this make the filter sharper, or does it just add volume? Most of the standard advice adds volume, because volume is what the platform reports and what agencies get thanked for.

The two completely different businesses inside "real estate"

This is the structural mistake that costs the most and gets noticed the least. An estate agency is two businesses wearing one brand. One business wins listings from owners who want to sell or let. The other finds buyers and tenants for listings it already holds. They have different searchers, different vocabulary, different conversion actions, different timelines and different economics — and a great many agents advertise as though they are one business with one budget.

On the seller side the customer is the property. The person searching is an owner working out what their home is worth, whether now is a sensible time to sell, or which agent to instruct. The conversion is a valuation request or an appointment. The value of winning that customer is the instruction itself, which then generates its own buyer-side demand. The competitive set is other agents, and the deciding factors are local credibility, evidence of recent activity in that specific area, and how quickly someone calls back.

On the buyer side the customer is the person, and the inventory is fixed — you can only sell what you hold. The person searching is looking for a place that matches a set of constraints, and your ability to satisfy them depends entirely on whether one of your listings fits. That is a completely different advertising problem: it is inventory-matching, not persuasion. If you have nothing in their area at their price, no amount of ad quality helps, and spending to reach them is spending to be told no.

Why they should not share a campaign

Put them in one campaign and the budget will quietly drift toward whichever side produces cheaper form fills. That is almost always the buyer side, because buyer-side enquiries are easy to generate and easy to generate badly. Meanwhile the seller side — where each win brings an asset that generates buyer demand for free — gets starved by an automated bidding decision nobody made consciously.

The vocabulary problem compounds it. Seller-side and buyer-side searches share words. "House price Lalitpur" could be an owner checking their value or a buyer checking affordability. In a mixed campaign you cannot tell which side of the business a click belongs to, which means you cannot tell which side is working, which means every optimization decision is made on blended data that describes neither.

The practical instruction is simple. Separate campaigns, separate budgets, separate conversion actions, separate negative keyword lists, and a rule that neither side is allowed to be judged against the other's cost per enquiry. They are not comparable numbers and treating them as comparable is the source of the drift.

Reading intent in a property search

Property is unusually generous with intent signals, because people carrying a real requirement describe it. They know the area they need to be in, roughly what they can spend, and how many bedrooms the family requires. Browsers do not volunteer constraints, because they do not have any. That asymmetry is the cheapest qualification tool available and most accounts throw it away by bidding broadly.

The specific query

A search that names an area, a property type and a further constraint — something in the shape of "3 bedroom flat Lalitpur" — is doing your qualification work unpaid. Three separate decisions have already been made and stated: where, what kind, and how big. The person has a requirement concrete enough to describe. They may still be early, they may still be unfinanced, but they have a specification, and a specification is the thing you can match inventory against.

These queries are worth paying properly for, and they are worth building an ad group around rather than folding into a general one, because the ad and the landing page can both answer the specification directly. A searcher who typed three constraints and lands on a page showing properties matching all three has had a very different experience from one who lands on a search results page and has to re-enter what they already told Google.

The bare area query

"Property in Pokhara" or "houses in Kathmandu" is a genuinely ambiguous signal and should be treated as one. It contains real buyers at the beginning of their search, sellers researching the market, tenants, people who moved abroad and follow prices at home, students on a project, and neighbors being nosy. All of them look identical in the auction.

This is not a keyword to exclude, but it is a keyword to fund differently. Lower bids, a landing experience designed to collect a requirement rather than to extract a phone number, and an honest expectation that most of this traffic will never identify itself. Treating bare area queries with the same bid and the same page as specific queries is one of the most common and most expensive structural errors in property accounts.

The research query

"Is it a good time to buy", "how much deposit do I need", "what does a valuation involve" — these are informational and they are not lies about intent, they are just early. Someone asking what a deposit requires is potentially a buyer, eventually. What they are not is a person who will respond usefully to an ad for a specific three-bedroom flat today.

The honest position is that most research queries do not belong in a search campaign at all. They belong in content, where they cost nothing per visit, and the visitor can be picked up later through remarketing when their search turns specific. Bidding on research terms in a conversion campaign is paying full price to reach someone months before they can act, then judging the campaign on their failure to act.

The browser problem, and what to do about it

Property listings are entertainment for a substantial share of the population. That is not a criticism of anyone — looking at homes is enjoyable — but it has a direct consequence for anyone paying per click. A property portal or a listing page attracts traffic that behaves beautifully by every engagement metric. Long sessions, multiple pages, returning visitors, photo galleries viewed end to end. None of it means a transaction is coming.

Worse, this traffic converts. Not into sales, but into enquiries — because an enquiry form on a beautiful property is a low-cost, low-commitment way to satisfy curiosity, and because "can I see more photos" and "what is the actual price" are perfectly natural things for a browser to ask. So the browser problem is not solved by conversion tracking. Browsers convert. They just do not transact.

Filter one — query specificity

The first and strongest filter is upstream of everything else: bid for constraint-carrying queries and be sparing with everything else. Exact and phrase match on specification queries, tight ad groups per area and property type, and broad match used only where you have the negative keyword discipline and the conversion signal to control it. This is unglamorous and it is where most of the improvement lives.

Filter two — negatives against research and curiosity

Property generates an unusually rich set of terms that look commercial and are not. Valuation-curiosity language, price-history and sold-price lookups, tax and legal process questions, career and job searches inside the industry, and the entire class of educational queries about how buying works. Each of these deserves a considered decision rather than a blanket block — some agencies genuinely want valuation-curiosity traffic on the seller side — but the decision has to be made per campaign, not once for the account.

• Research and education language — how to, guide, process, explained, meaning, requirements — belongs in content, not in a conversion campaign.
• Price-history and sold-price lookups are usually curiosity or valuation research, and they will not behave like buyer traffic no matter how the ad reads.
• Rental terms in a sales campaign and sales terms in a rental campaign are the most common cross-contamination, and both are pure waste.
• Industry-internal searches — jobs, licensing, commission structures, agency software — arrive constantly and convert never.
• Investment, auction, repossession and land-only terms may or may not be relevant; decide deliberately, because they behave nothing like residential buyer traffic.
• Area names you hold no stock in should be negatives on the buyer side even when they are core to the seller side.

The search terms report is the only honest source for this list, and in property it needs reading more often than in most verticals because seasonal and news-driven curiosity spikes arrive without warning. A single local news story about prices in an area will pour non-transactional traffic into a campaign for a week. The general method for building and maintaining these lists is covered in the negative keyword list for service businesses, and the same discipline applies here with one addition: in property, the terms you exclude on the buyer side are often the exact terms you want on the seller side, so the lists must be campaign-scoped rather than shared at account level.

Filter three — use the form to qualify, not to maximize submissions

Every piece of standard conversion advice says shorten the form. In property that advice is actively harmful on the buyer side, because a two-field form is exactly what a browser will fill in and a serious buyer will fill in either way. The form is the last filter before a human spends time, and it should be doing work.

The two questions that separate transactors from browsers are timeline and financing readiness. Someone who is buying in the next few months and has spoken to a lender will answer both without hesitation. Someone browsing will either abandon the form or answer in a way that immediately tells you where they are — and both outcomes are useful. You will get fewer submissions. You will get more viewings, which is the number that matters.

The long cycle, and what it does to your attribution

A property transaction is one of the largest financial decisions a household makes, and it moves at a speed that reflects that. Months pass between a first search and a completed transaction. During those months the person searches repeatedly, visits portals, talks to friends, watches an area, changes their mind about the area, and revises their budget — usually without identifying themselves to anyone.

The enquiry, when it finally arrives, arrives late. By the time someone fills in a form about a specific property they have typically done a great deal of unattributed work. This creates a specific and systematic distortion: the click that gets credited is the last one, near the end of a long process, and it is very often a branded or highly specific search that the earlier, broader work made possible. Last-click attribution will tell you the branded campaign is your best performer and the discovery work is waste. In this vertical that reading is close to guaranteed to be wrong.

Two consequences follow. First, a thirty-day reporting window is too short to judge a property account, and monthly performance reviews in this vertical are largely reading noise. Second, the conversion lag has to be understood before any bidding decision is made — pausing keywords that have not converted "yet" in a market where "yet" means several months is how accounts get cut back to nothing but brand terms.

None of this is fixable by choosing a different attribution model in a dropdown. What it requires is measurement that survives the gap between click and outcome, which means click identifiers stored at enquiry, outcomes recorded in the CRM, and those outcomes sent back. The mechanics of setting that up correctly are covered in the conversion tracking guide, and in property they are not optional refinements. Without them the account is being steered by a signal that describes the last few days of a process that took half a year.

The CRM is the campaign's other half

In most verticals I would call the CRM an important adjacent system. In property it is half the campaign, and an account run without it is being flown blind by definition rather than by neglect.

The first reason is speed. A property enquiry has a short window in which it is worth what you paid for it. Someone enquiring about a listing is very likely enquiring about several listings from several agents in the same session, and the first substantive human response frames the entire relationship. An enquiry answered the following week is not a slightly degraded version of the same lead — it is a fraction of its value, because by then someone else has booked the viewing. Buying enquiries faster than you can answer them is a way of spending money to hand prospects to competitors.

The second reason is measurement. The transaction closes so far downstream that nothing inside the ads platform can see it. The only place the sequence — enquiry, response, viewing, second viewing, offer, agreed sale — exists in one record is the CRM, and only a reconciliation between that record and the ad account can answer which campaigns produced revenue rather than which produced form fills. That is the entire basis on which budget should be allocated in this vertical, and it lives outside the tool most people spend their day inside.

Both halves of that depend on plumbing that is genuinely difficult to get right. The routing side — making sure an enquiry reaches a person who is available, with a fallback when they are not, and an escalation when nobody has touched it — is covered in the post on building a lead routing system that does not drop leads, and property is the vertical where its failure modes bite hardest because the response window is measured in hours. The data side is covered in why most CRM implementations fail, and the specific failure that ruins property reporting is the one that looks harmless: outcomes recorded inconsistently, so nobody can tell whether a lead went cold or was simply never updated. A pipeline stage that means different things to different agents makes offline conversion import worse than useless, because it feeds confident but wrong signals back into bidding.

Remarketing on specific properties — effective, and easy to push too far

Remarketing genuinely works in property, and it works for a reason specific to this vertical: the consideration period is long, the decision involves more than one person, and people return to the same properties repeatedly before acting. Reminding someone of a listing they engaged with, while they are still deciding, is a legitimately useful service. Very few tactics in advertising can be described that way honestly.

It is also the tactic most easily pushed past the point where it helps. There is a real boundary here and it is worth stating plainly: following someone around the internet with a property they viewed once is intrusive, and past a certain frequency it stops reading as a helpful reminder and starts reading as surveillance. People notice property ads more than they notice most ads, because the subject matter is personal and because a specific home they looked at is instantly recognizable. The negative reaction is proportionally stronger.

• Frequency capping is not optional. Decide the cap deliberately, set it low, and treat a rising impression-per-user figure as a problem rather than as increased coverage.
• Keep membership windows short and tied to the consideration period, not to the maximum the platform allows. Someone who looked at a flat months ago has almost certainly resolved their situation one way or the other.
• Remove people who transacted, and remove people who enquired and were disqualified. Both are pure waste and the first is actively embarrassing.
• Do not remarket a property that is under offer or withdrawn. This is the single most common failure and it produces enquiries that cannot be satisfied, which is worse than no enquiry at all.
• Segment by depth of engagement. Someone who viewed one page and someone who returned three times and opened the floor plan are not the same audience and should not receive the same weight of contact.

The listing-status point deserves emphasis because it is where remarketing and inventory management collide. An ad account does not know a property went under offer. If nothing connects listing status to audience membership and ad serving, you will spend money advertising properties you cannot sell, generate enquiries you have to disappoint, and train a segment of the local market to associate your brand with a wasted afternoon.

The property page is the landing page

Buyer-side traffic should almost never land on a homepage or a generic search results page. The person told the ad platform what they wanted, the ad implied you had it, and the page has to confirm that within a second or the handoff fails. In this vertical the landing page usually already exists — it is the listing — which is convenient and also the reason it is so often left unexamined.

• The photography is the page. Real, current, honestly lit images of the actual property, not renders that flatter it and not a stock exterior. In property the visual is the product description.
• Price must be visible. "Price on application" filters out serious buyers as efficiently as browsers, and a searcher who arrived with a price band in their query will leave rather than ask.
• Location has to be specific enough to be trusted and precise enough to be useful. Vague area labeling reads as concealment even when it is not.
• The specification the searcher typed must be visible without scrolling — bedrooms, size, type, tenure. They are checking whether their constraints hold.
• The enquiry action has to be present on the first screen and again after the gallery, because the two moments people decide to enquire are before they look and immediately after.
• It has to work on a phone, on an ordinary connection, with a large gallery. This is where most property pages actually fail, and it fails silently.
• Related listings matter more here than on most pages, because a near-miss is common and a buyer who rejects this property may take the next one.

The broader mechanics of page-level performance for paid traffic — message match, first screen, form design and speed — are covered in the post on landing page optimization for Google Ads. The property-specific addition is that a listing page has a shelf life. It becomes wrong the moment the property is under offer, and a landing page that is wrong is worse than a landing page that is slow.

Advertising claims, and why honesty is the commercial position here

This section is short and it is firm. Property advertising has a persistent temptation toward invention, because the individual transaction is large enough to make a small exaggeration feel worth it. It is not, and the reasoning is not only ethical.

• No invented sold prices. Do not cite a sale figure you cannot evidence, and do not imply a track record you do not have. This is checkable, and in a small market it will be checked.
• No fabricated scarcity. "Only two units left" when it is untrue is a misrepresentation, and countdown urgency on a property that has been available for months is transparent to anyone who has been watching the listing.
• No misrepresenting condition. Photography that hides a defect does not prevent the buyer from standing in the room. It just moves the conversation from the property to your credibility.
• No borrowing a location. Describing a property as being in an area it borders rather than the area it is in is the oldest one, and it is the one buyers spot fastest and resent longest.
• No implied guarantees about value, appreciation or rental yield. These are forecasts about someone else's money and stating them as facts is a serious exposure.

The policy risk is real — advertising platforms treat misrepresentation seriously and account-level enforcement is not a proportionate response to a single ad, it is the standard one. But the reputational risk is the larger one and it is specific to this industry. Property runs on referral and repeat instruction. The seller who instructs you in five years is somebody's friend today. An agency that overstates in its ads acquires a reputation faster than it acquires listings, and the reputation outlasts the campaign.

Running this from Nepal

Three platform facts change what is buildable for a Nepal-based advertiser, and they are worth knowing before designing an account rather than after. Google Merchant Center and Shopping are not available for Nepal, which closes any property advertising route that depends on a Merchant Center feed. Lead form assets are not available in Nepal, so in-ad capture is impossible and every enquiry has to be taken on your own landing page and your own form. Call reporting and Google forwarding numbers are not available either — a phone number can still appear on the ad, but calls generated by Google Ads cannot be measured natively, which matters in a market where the phone is a dominant enquiry path.

The practical effect is that a Nepali property advertiser is pushed toward exactly the setup this post argues for anyway: own landing pages, own forms, own tracking, and outcome reconciliation in the CRM. The wider set of constraints — targeting, language support, billing and the regulatory picture — is covered in the Google Ads in Nepal guide, which is the place to start if you are setting up an account here rather than adapting one.

Where this breaks in practice

• Advertising stock you do not have. Buyer-side campaigns running against areas and price bands with no matching listings generate enquiries you can only disappoint. Inventory has to constrain the campaign, and nothing does that automatically.
• Listings advertised after they go under offer. The ad account has no idea. Without a process connecting listing status to campaign and audience state, this happens continuously and nobody reports it.
• Enquiries arriving faster than anyone answers them. Scaling spend before response capacity is the most reliable way to convert budget into competitor viewings.
• One campaign for both sides of the business, with a shared budget quietly reallocating away from listing acquisition because its form fills cost more.
• Optimizing bidding toward enquiry volume. The bidding system will find you people who fill in forms, and in this vertical those people are disproportionately browsers.
• Judging the account monthly. With a cycle measured in months, a monthly review will cut the campaigns that were working and protect the branded search that was harvesting their output.
• Remarketing with no frequency cap, on a personal subject, to an audience that has already moved on. This is the failure people actually complain about out loud.

The metric that lies

Enquiry volume. It is the number in every report, it is the number the platform optimizes toward if you let it, and in property it can rise steadily while the business gets worse.

The mechanism is straightforward. Broaden the targeting, shorten the form, soften the ad copy toward curiosity, and enquiries go up. Every one of those moves also lowers the average intent of the person enquiring. The report shows a rising line. The agents show a rising workload. The viewing diary shows nothing new at all.

The pairing matters more than either number alone, because each is misleading by itself. Viewings alone will fall in a slow market regardless of the account. Enquiries alone will rise whenever the filter loosens. Together they describe whether the filtering is working, which is the only question the account is actually answering. The broader argument for pairing every headline metric with the number that catches its failure mode is in the bid governance framework, and property is the clearest case for it I know of.

When an agent should not run Google Ads

There are situations where the honest recommendation is not to spend, and they are common enough to state directly.

• You cannot respond to an enquiry within the working day, reliably, including when the person who normally handles them is out showing a property. Buying enquiries you cannot answer is worse than not buying them, because it also costs you the reputation.
• You have very few listings. Paid search on the buyer side is inventory-matching, and with thin inventory most of the traffic you pay for will be looking for something you do not have.
• Nothing records what happened after the enquiry. Without outcome data you cannot distinguish the campaigns producing transactions from the ones producing browsers, and in this vertical those two look identical for months.
• Your listings are already carried on a portal that dominates local search, and your realistic differentiator is service rather than stock. In that case seller-side advertising and local content may be the better use of the same money.
• The budget only supports one side of the business and you have not decided which. Split thin budgets produce two campaigns with too little data to steer either.

In several of those cases the money is better spent on the plumbing first — response capacity, a CRM that records outcomes, listing pages that work on a phone. That is not a delay tactic. Those are the components that make the advertising legible later, and buying traffic into a system that cannot measure or service it is the most common way property budgets disappear without anyone being able to say what happened.

A starting sequence

1. Decide which side of the business you are advertising this quarter — listing acquisition or buyer demand. Pick one to start. Running both badly is worse than running one properly.
2. Fix the response path before spending. Establish who answers an enquiry, within what time, with what fallback when they are unavailable, and what happens if nobody has touched it by end of day.
3. Get outcomes into the CRM with consistent stages: enquiry received, contacted, viewing booked, viewing attended, offer, agreed. Consistency matters more than granularity.
4. Install conversion tracking that stores the click identifier at enquiry, so outcomes can be matched back later. Do this before the first click, not after the first confusing report.
5. Build the campaign around specification queries — area plus type plus constraint — in tight ad groups, with bare area queries funded separately and lower.
6. Write the negative keyword lists per campaign, not per account, starting with research language, industry-internal terms, cross-contamination between sales and lettings, and areas you hold no stock in.
7. Make the enquiry form ask timeline and financing readiness. Accept that submissions will fall. Watch viewings, not submissions.
8. Audit the listing pages you are sending paid traffic to, on a phone, on a normal connection. Fix price visibility, first-screen specification and gallery weight before increasing budget.
9. Add remarketing only after the above works, with a low frequency cap, short windows, and a process that removes properties the moment they go under offer.
10. Review on a quarter, not a month, and review against viewings and instructions rather than enquiry counts. Then reconcile the CRM outcomes back against campaigns and reallocate from that, not from the platform's default report.

Where to go from here

The one decision that changes a property account most is the first one in that list: stop treating listing acquisition and buyer demand as a single business with a single budget. Everything downstream — keyword structure, conversion actions, landing pages, what you measure — resolves more easily once that split exists, and stays confused for as long as it does not.

After that, the work is filtering rather than generating, and the honest test of whether it is working is not in the ads platform at all. It is whether the diary is fuller. I build these measurement and routing systems through Arcetis, and in property the pattern is consistent enough to be a rule: the accounts that work are the ones where the CRM and the ad account are reconciled against each other, and the accounts that quietly fail are the ones judged entirely on the number of forms submitted.

Related reading