Real Estate Google Ads Case Study in Pakistan

The Client

An Islamabad-based property developer midway through launching a mixed-use mid-rise in a developing sector of the capital. The project combined one-, two-, and three-bedroom apartments above ground-floor retail and office space, with unit prices ranging from roughly PKR 12M for a one-bed to PKR 45M for a three-bed corner unit. The developer had a credible brand locally, a finished sales centre, and a small in-house sales team that ran on site visits and phone conversations rather than online forms.

When they engaged WeProms Digital, they were spending around PKR 1.9M a month on Google Ads and generating a steady stream of leads that the sales team openly distrusted. The marketing dashboard showed healthy lead volume; the sales floor told a different story. Half the enquiries came from people who could not afford the entry unit, were looking to rent, were based in another city, or were other brokers fishing for commission splits. The developer was paying Google a fortune to populate a CRM the sales team did not want to call.

The brief was specific and honest: we do not need more leads, we need more qualified enquiries, and we need to know what a lead actually costs once the junk is removed.

The Problem

Four failures were turning ad spend into noise:

1. Broad-match keyword waste. The account ran almost entirely on broad-match terms like “apartments Islamabad” and “property for sale.” Broad match merrily served the ads to people searching for rental flats, property-management jobs, news about the sector, and competing projects. The search-term report was full of queries that had no chance of producing a buyer, and Google was charging for every one.

2. One landing page for everything. A single generic page described the whole project. A searcher looking for a one-bed apartment and a searcher looking for commercial office space landed on the same page, saw the same hero image, and filled the same generic form. Intent was destroyed at the door, and the sales team had to re-qualify every lead from scratch.

3. Performance Max and Search were tangled. Performance Max was running alongside Search campaigns with shared budget and no separation by inventory type. PMax’s opaque placements were consuming budget without any visibility into which units the leads were interested in, and it was competing with the Search campaigns for the same high-intent queries.

4. Bidding optimised toward the wrong event. The only conversion being fed back to Google was the form submission. Calls — which the sales team considered far more serious — were not tracked at all, and the sales-qualification outcome (whether a lead was actually a qualified buyer) never reached the bidding algorithm. Google was dutifully optimising toward cheap form fills, which is exactly what it delivered.

The result was a lead pool where 47% was waste, a cost per qualified lead that had crept up to PKR 4,650, and a sales team that had stopped believing the marketing numbers.

Phase 1 — Account Audit and Query Cleanup (Weeks 1-2)

The first move was not creative or strategic — it was janitorial. We pulled 90 days of search-term data and went through it line by line.

Negative-keyword build. We built a negative-keyword list of roughly 380 terms grouped into intent buckets: rental intent (rent, lease, to let, monthly), employment intent (jobs, career, salary), informational intent (news, review, complaint, scam), geographies outside the catchment (other cities, overseas locations that mapped to relocation-not-purchase intent), and explicit broker terms. This alone stripped a large share of the wasted spend in the first ten days.

Match-type tightening. Broad match was replaced with a combination of phrase and exact match on the commercial-intent terms that actually produced buyers, with broad match retained only for carefully controlled discovery campaigns. Query coverage dropped, but the queries that remained were far more likely to be buyers.

Geo and audience layers. We layered location targeting to prioritise Islamabad and the affluent adjacent catchments, with bid adjustments rather than exclusions for second-tier cities so the account could still learn from them. Audience observation (without restrictive targeting) was added for in-market property buyers and high-income household segments so we could read which audiences converted without choking volume early.

Cleanup lever	Before	After
Search-term waste share	~38% of clicks	<7% of clicks
Negative keywords	40 inherited	~420 active
Match types	~85% broad match	Phrase + exact led
Geo targeting	Whole country bid equally	Islamabad + catchment prioritised

By the end of week 2 the lead pool was already cleaner, and the sales team’s grumbling about junk leads had thinned noticeably — all before any structural rebuild.

Phase 2 — Restructure and Unit-Specific Landing Pages (Weeks 3-5)

With the query base clean, we rebuilt the account architecture so budget and intent could be controlled at the unit level.

Search versus Performance Max separation. We split the account into a Search arm, organised by unit type (one-bed, two-bed, three-bed, commercial retail, office) and a Performance Max arm that ran the project’s asset group as a brand-and-inventory awareness engine. Budget was no longer shared between them, and PMax was set up with asset-group signals tied to in-market property audiences so its opaque placements were at least steered toward likely buyers. For context on why PMax needs this kind of guardrailing rather than blind trust, our note on Performance Max as the single real lever in Pakistani Google Ads performance is relevant reading.

Unit-specific landing pages. Instead of one generic page, we built five landing pages — one per unit type — each carrying the relevant floor plan, price band, possession timeline, and a form whose fields matched that unit’s buyer (for example, commercial pages asked about business type and required area, residential pages asked about family size and preferred possession date). This is the change that did the most for lead quality: when the page matches the search, the form fills itself with relevant information and the sales team’s first call starts from a real conversation.

Conversion actions redefined. We defined a clear hierarchy of conversion events — form submission, phone call, WhatsApp enquiry — and weighted them by seriousness, with the call set as the primary optimisation event rather than the form.

The full Google Ads management and optimization restructure meant that by the end of week 5, each unit type had its own campaigns, its own landing page, and its own conversion accounting. Lead quality kept climbing.

Phase 3 — Call Tracking and Offline Conversion Import (Weeks 4-7)

This is the phase that changed what the bidding algorithm was actually optimising toward, and it is the part most Pakistani property accounts skip entirely.

Call tracking. We deployed dynamic call tracking so every ad-click phone call was attributed back to the keyword and campaign that produced it. For a vertical where the serious buyer almost always picks up the phone, not tracking calls means under-counting exactly the leads that matter. The call-tracking and offline conversion integration also captured call duration as a weak quality signal.

Offline conversion import. The real unlock was feeding the sales team’s qualification outcome back into Google. Once a lead was contacted, the sales team tagged it in the CRM — unqualified, qualified, site-visit booked, site-visit completed. We imported those outcomes back into Google Ads as offline conversions of different values, attached to the original click via the GCLID. Suddenly Google was not guessing which form fills were good; it was being told.

tCPA Smart Bidding on qualified signal. With qualified-enquiry and site-visit conversions flowing back, we moved the Search arm to a target cost-per-acquisition bidding model where the “acquisition” was a sales-qualified lead, not a form fill. The algorithm now had permission to pay more for a click that produced a qualified buyer and less for one that produced a tyre-kicker. Paired with the conversion tracking setup work, the bidding system finally had a definition of success that matched the sales floor’s.

By the end of week 7, cost per qualified lead was already falling, and — critically — qualified enquiry volume was rising at the same time, which is the opposite of what usually happens when you tighten quality.

Phase 4 — Smart Bidding, Dayparting, and Scale (Weeks 6-12)

The final phase turned a working system into a compounding one.

Dayparting and device bidding. Call data revealed that qualified enquiries concentrated in weekday evenings and weekend mornings — the times when salaried buyers were actually free to talk. We shifted budget weight toward those windows and reduced it during low-conversion daytime hours when broker activity was highest. Device adjustments reflected that tablet and desktop traffic converted to qualified enquiries at a noticeably higher rate than mobile-only sessions.

Creative testing tied to unit type. Ad copy was tested per unit type rather than across the whole project. Commercial-unit copy led with possession timeline and footfall; residential copy led with floor plan and family fit. This sounds obvious, but the original account had run one set of ads for everything, and the per-unit testing lifted click-through and qualification together.

Pacing and budget reallocation. As the Search arm proved out at a lower cost per qualified lead, budget was gradually shifted out of PMax and into the Search campaigns by unit type, with PMax retained at a measured level for awareness. Monthly spend stayed close to the original PKR 1.9M; the gains came from reallocation and efficiency, not from new money. Within Search itself, budget migrated toward the two-bed campaigns, which the offline-conversion data showed were producing the highest site-visit rate at the lowest cost per qualified lead, and away from the one-bed campaigns where broker interference remained highest despite the cleanup.

Sales-team feedback loop. The technical work only holds if the sales team keeps tagging outcomes honestly. We installed a lightweight CRM habit: every contacted lead got one of four tags within 48 hours, and a weekly ten-minute review compared the marketing dashboard against the sales floor’s actual experience. When the two disagreed for a given campaign, that campaign was flagged for re-examination rather than left to drift. This loop is what kept the bidding model honest over time — an offline conversion import is only as good as the tagging discipline behind it.

Optimisation lever	Effect on cost per qualified lead	Effect on lead quality
Negative keywords + match types	Largest single reduction	Removed broker and rental waste
Unit-specific landing pages	Moderate reduction	Largest single quality lift
Offline conversion import	Large reduction	Taught bidding the real definition
Dayparting + device bidding	Moderate reduction	Concentrated spend in serious-buyer windows
Per-unit creative testing	Small reduction	Improved intent match at the ad level

Final Results at 90 Days

Metric	Before	After	Change
Cost per qualified lead	PKR 4,650	PKR 3,070	-34%
Qualified enquiries per month	142	224	+58%
Junk and broker lead share	47%	9%	-38 pts
Lead-to-site-visit rate	14%	29%	+15 pts
Cost per site visit	PKR 33,200	PKR 11,800	-64%
Monthly ad spend	PKR 1.9M	PKR 1.9M	Flat

The number the developer cared about most — cost per site visit — fell 64%, because cost per qualified lead fell and a far larger share of qualified enquiries actually turned into visits. The sales team went from distrusting the marketing pipeline to asking for more.

What Made This Work

1. We cleaned the signal before touching the bidding. Smart Bidding is only as good as the conversions it learns from. By removing broker and wrong-city waste through negatives and geo layers first, we ensured that when tCPA switched on, it was optimising toward real buyers rather than reinforcing the junk-lead pattern the account had been training on for months.

2. The offline conversion loop closed the gap between marketing and sales. Most real estate accounts feed Google a form-fill and hope. Importing the sales team’s qualification outcome meant the bidding algorithm and the sales floor were finally optimising toward the same definition of success. That single loop did more for cost per qualified lead than any creative change.

3. Unit-specific landing pages matched intent to inventory. Real estate intent is specific — a buyer wants a particular unit type at a particular price. One generic page flattens that intent and forces the sales team to rebuild it on the phone. Five matched pages preserved the intent the search had already expressed, which raised both conversion rate and lead quality simultaneously.

4. Calls were treated as first-class conversions. In a phone-first market, not tracking calls means under-counting the serious buyers and over-weighting the casual form-fillers. Attribution that ignores calls will always misprice property traffic.

What Teams Can Apply

For Pakistani property developers running Google Ads:

1. Read your search-term report before you change your bids. If 30% or more of your clicks are rental, job, or out-of-catchment queries, no bidding strategy will save you. A disciplined negative-keyword and match-type cleanup is usually the single highest-return move available, and it costs nothing but attention.

2. Build one landing page per unit type. A buyer searching for a one-bed apartment and a buyer searching for commercial office space are not the same lead. Matching the page to the search intent is the cheapest way to lift both lead quality and conversion rate at once — far cheaper than buying more traffic.

3. Import your sales qualification back into Google. If the only conversion Google sees is a form fill, it will optimise toward form fills. Feeding qualified-enquiry and site-visit outcomes back through offline conversion import, tagged to the GCLID, turns Smart Bidding into a partner rather than a volume machine.

4. Track calls, and treat them as the serious conversion. In Pakistani property, the serious buyer calls. An account that does not track calls is optimising toward the least committed leads in the pool.

5. Separate Search from Performance Max by inventory type. PMax is powerful but opaque, and mixed with Search it hides which units your money is actually selling. Run them as separate arms with separate budgets so you can read and reallocate.

This framework is built for the Pakistani real estate context — broker-heavy lead pools, phone-first buyers, Urdu and Roman-Urdu search behaviour, and a sales cycle anchored on the site visit. The specific keywords, landing-page set, and qualification rules change with every project, but the sequence is the same: clean the query base, match intent to inventory, close the offline conversion loop, and only then let Smart Bidding work.