Real Estate PPC Case Study in Pakistan

The Client

A family-run Karachi developer with fifteen years of residential projects behind it, currently selling two towers: a possession-ready 268-unit building with two- and three-bedroom apartments priced PKR 9.8M to PKR 16.5M, and a 190-unit under-construction tower sold on a 30-month instalment plan. The sales operation is a gallery near the site, a team of nine reps, and a phone desk — in Pakistani property sales, the deal moves on calls and site visits, and the online funnel exists to put qualified callers and visitors in front of reps.

The developer had been running Google Ads for seven months at a steady PKR 2.4M a month, managed in-house by a marketing lead who reported lead volume to the board every month. On paper the account worked: 900-plus enquiries a month at a reported cost per lead near PKR 2,600. On the sales floor it did not. Reps were calling lead after lead who wanted to rent, wanted a plot in a specific society, was shopping for a job, or — most often — could not come within PKR 4M of the entry unit. The sales manager’s spreadsheet told the real story: about 166 of those monthly enquiries were qualified buyers, which put the true cost per qualified lead near PKR 14,460 and climbing.

The brief to WeProms Digital came with a condition that shaped the whole engagement: the board would rather see 200 leads they can work than 900 they cannot, and the first deliverable should be an honest number.

The Problem

Five failures were converting PKR 2.4M a month into noise:

• One campaign for every intent. A single Search campaign with 16 ad groups, roughly 85% of keywords on broad match, served the same ads to a buyer searching “2 bed flat for sale in Gulshan”, a renter, a job-seeker, and a Lahore resident researching Karachi prices for a relocation article.
• 45% of clicks were waste. The search-term report was saturated with rental phrasing, society-plot queries from buyers who wanted land not apartments, other-city locations, and government-housing scheme lookups. Google charged for every one.
• Destination pages destroyed intent. Every click landed on the developer’s homepage or one generic project page. A possession-ready buyer, an instalment-plan buyer, and an overseas Pakistani got the same hero image, the same buried floor plans, and the same 11-field form.
• Tracking flattered the account. The only conversion was the thank-you-page view, which re-fired on reload and inflated conversions by roughly 18%. Calls — which the phone desk closed far more often than forms — were not tracked at all.
• Sales knowledge never reached the bidding. Reps dispositioned leads in their own sheets; nothing told Google Ads which keywords, ads, and audiences produced qualified buyers versus renters. The algorithm optimised toward what it could see, and what it could see was form fills.

This is the standard failure shape of Google Ads management for property in Pakistan: volume metrics that satisfy a monthly report while the cost per buyer — the only number a developer can take to the bank — stays invisible.

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

The audit began with 90 days of search-term data: 11,600 distinct queries, graded line by line against one question — could this searcher plausibly buy a PKR 9.8M apartment in Karachi this year?

The negative-keyword build. The inherited list held 60 terms. We built it to 1,240, grouped by failure mode: rental intent (rent, lease, to-let, monthly), employment intent (jobs, career, salary), plot-only intent (plot, file, society-name + plot), wrong-city queries (Lahore, Islamabad, and other cities mapped to research-not-purchase behaviour), government-scheme lookups, and broker-harvesting patterns. Karachi’s search behaviour makes this list longer than most markets — buyers search by area and society name, so the noise multiplies per locality.

Match types and geo. Broad match was pulled back to phrase and exact on commercial terms. Geo targeting was cut to Karachi for the local campaigns — an out-of-city buyer rarely completes a Karachi apartment purchase without local family ties, and those ties arrive through referrals, not search ads. Out-of-country traffic was excluded here and handled properly in Phase 2 rather than mixed into local clusters.

Honest conversion counts. The thank-you-page reload bug was fixed with a deduplicated conversion tag, and reported conversions fell about 18% overnight. The first monthly report to the board showed fewer conversions than any previous month — flagged explicitly as the first number in seven months they could trust.

Cleanup measure	Before	After Phase 1
Negative keywords	60	1,240
Irrelevant click share	~45%	~12% by week 4
Match-type mix	~85% broad	Phrase + exact led
Geo	Nationwide	Karachi (local campaigns)
Reported conversions	Inflated ~18%	Deduplicated

The arithmetic behind the cleanup was stark: 45% of a PKR 2.4M monthly spend is roughly PKR 1.08M a month buying traffic that could never produce a buyer. Cutting the irrelevant share toward 12% reclaimed about PKR 0.8M of working budget — the equivalent of funding a third campaign cluster without touching the total. This is why query cleanup came before creative, landing pages, or bidding changes: it is the only phase of a Google Ads turnaround that pays for all the other phases.

By the end of week 2 the junk share of leads had already begun to fall, before a single new ad or landing page existed.

Phase 2 — Intent Clusters and Landing Pages (Weeks 3-5)

The restructure replaced the single campaign with five clusters, each with its own budget, keywords, ads, and destination:

Cluster	Intent captured	Destination page
Project brand	Developer and tower names	Tower pages with unit availability
High-intent area search	”flats for sale in Gulshan”, “apartment Scheme 33”	Area-matched landing pages
Possession-ready	”ready flat”, “immediate possession Karachi”	Ready-tower page with floor plans and price table
Payment-plan	”apartment on instalments”, monthly-plan queries	Under-construction page with 30-month plan calculator
Overseas Pakistanis	US, UK, UAE buyers, English + Roman Urdu	Remote-booking page in USD/GBP with video walkthrough

The area-search cluster deserves its own note. Karachi buyers search by locality — “flats for sale in Gulshan”, “apartment Scheme 33”, neighbourhood names strung together with unit types — and each area query implies a budget band and a shortlist of projects. Splitting these into their own cluster with area-matched landing pages moved them from the account’s most expensive keyword group to its most productive: cost per qualified lead in the area cluster ultimately ran 22% below the account average.

Landing pages that qualify. Each page carried the specifics a buyer needs to self-select: floor plans, a payment schedule table, and a financing calculator showing bank and developer instalment maths side by side. The form was cut from eleven fields to six, and two of the six were qualification questions — a budget band and a “financing needed?” toggle. Mismatched buyers filtered themselves out at the form; the ones who submitted arrived pre-qualified.

Overseas Pakistanis as their own cluster. Diaspora buyers behave differently from local buyers: they browse in US and UK evenings, think in dollars and pounds, and cannot visit the site before committing. The dedicated page showed USD pricing, a video walkthrough, verification documents, and a callback scheduler set to the buyer’s timezone. Campaign scheduling followed the target cities’ evenings, not Karachi’s.

Responsive search ads per cluster. Headlines were tested against cluster intent: the possession-ready cluster led with “Immediate Possession — Ready 2 & 3 Bed Apartments” (+22% CTR against the generic control), while the payment-plan cluster led with “30-Month Payment Plan — Book with 20% Down” (+34% CTR). Same towers, different doors.

Two implementation details mattered more than they look. Every landing page was built light — under 1.5MB with the form above the fold — because Karachi property research happens overwhelmingly on mid-range Android phones over mobile data, frequently after 9pm; a page that stalls loading a floor-plan gallery loses the buyer to a competitor’s faster page. And the financing calculator defaulted to showing the monthly instalment figure before the total price, because the monthly number is what Pakistani apartment buyers actually negotiate around.

Phase 3 — Offline Conversions and Smart Bidding (Weeks 4-8)

With clean clusters and honest on-site conversion counting, the next job was teaching Google what a buyer looks like.

Call tracking. Ad-specific phone numbers were provisioned so calls from ads could be counted — with a 60-second minimum duration threshold to screen misdials and wrong numbers. Calls of 60 seconds or more converted at nearly three times the rate of form fills, which confirmed the sales team’s long-standing claim about where the serious buyers were.

Offline conversion import. Working through call tracking and offline conversion integration, we mapped two CRM dispositions to Google: Qualified (budget matched, timeline inside 12 months, financing path confirmed) and Site Visit (appointment attended). Each carried the GCLID captured at form or call, with a 28-day import window. For the first time, the bidding algorithm saw which keywords and ads produced buyers weeks after the click — not just form submissions minutes after it.

tCPA on the qualified signal. Once each campaign accumulated roughly 50 qualified conversions, bidding moved to target CPA against the imported Qualified event, and budget shifted toward the clusters producing them: the high-intent area cluster grew from 34% to 61% of spend across weeks 6 to 8. The overseas campaign became the account’s efficiency leader at a PKR 7,800 cost per qualified lead — 22% of spend producing 31% of qualified enquiries by week 12, including 14 remote booking deposits in the final five weeks.

Negative audiences. Closed-lost leads with “out of budget” dispositions were built into exclusion lists, so the account stopped paying to re-reach the people sales had already ruled out.

The transition to tCPA was staged rather than flipped. Each campaign stayed on Maximize Clicks until it held roughly 50 imported qualified conversions — below that threshold a target-based strategy is just guessing — and the initial targets were set loose at PKR 11,000 before tightening to PKR 9,500 over three weeks as the algorithm accumulated signal. Search partners and the display network were switched off for every Search campaign; both had been quietly absorbing spend with close to zero qualified output. Rushing the switch is the classic error in this sequence: tCPA trained on a few weeks of thin data optimises confidently toward the wrong buyer.

Phase 4 — Sales Feedback Loop and Scale (Weeks 8-12)

The final phase made the system durable rather than lucky.

A weekly disposition review with the sales manager became the account’s steering mechanism: every lead marked junk produced a search-term check, and roughly 40 to 60 new negatives were added each week as Karachi query behaviour shifted with the market cycle. The qualification definition itself was tuned twice — first to exclude buyers with a timeline beyond 12 months, then to fast-track cash buyers — and both changes flowed into the offline import within a week.

The board pack changed shape too: spend, qualified enquiries, cost per qualified lead, and cost per site visit by cluster — no raw lead counts, because raw counts had been the problem for seven months. The under-construction tower, until then carried only as a payment-plan cluster, got the full treatment in week 10 and produced its first site-visit bookings within twelve days. Those bookings came in at a cost per site visit 18% below the possession-ready tower’s first-month figure, because the payment-plan cluster inherited the full negative list, the qualification form, and the offline-conversion plumbing from day one — the second project paid none of the first project’s tuition.

By the 90-day mark the account was running at PKR 2.5M a month — 4% above the starting spend — with a forecast floor for cost per qualified lead near PKR 8,200 once the second tower’s campaign set fully matured.

Final Results

Measured at the 90-day mark against the pre-engagement baseline:

Metric	Before	Day 90
Monthly spend	PKR 2.40M	PKR 2.50M
Raw enquiries	923 / month	806 / month
Qualified enquiries	166 / month (+18% of raw)	262 / month (33% of raw)
Cost per qualified lead	PKR 14,460	PKR 9,540 (-34%)
Junk and broker share	55%	19%
Site visits booked	35 / month	89 / month
Cost per site visit	PKR 68,600	PKR 28,100 (-59%)
High-intent impression share	19%	44%

Raw lead volume fell 13% on purpose. The account stopped paying for renters, job-seekers, and wrong-city traffic, and reinvested the reclaimed budget into clusters where the sales floor was closing. Fewer leads, more buyers, lower cost per buyer — which is the entire argument for optimising real estate marketing in Pakistan against qualified counts rather than raw ones.

What Made This Work

1. The bidding target changed before the bidding strategy did. Importing Qualified and Site-Visit events from the CRM meant tCPA optimised toward the sales floor’s definition of a buyer. Smart Bidding on a form-fill proxy just buys more form fills, efficiently.
2. The negative list was written in Karachi’s query language. Rental phrasing, plot-only society searches, and wrong-city variants are market-specific noise. A generic real estate negative list imported from a Western template would have missed most of it.
3. Qualification moved upstream into the form. Budget band and financing questions let the wrong buyers disqualify themselves before a rep spent a call — the qualified share rose from 18% to 33% without shrinking reach.
4. Overseas Pakistanis were treated as a separate market, not a keyword. Own timezone schedule, own currency, own landing page — and the cluster became the account’s cheapest source of qualified buyers.
5. Sales dispositions flowed back weekly. The loop between the CRM and the ad account kept the definition of “qualified” honest as market conditions shifted, instead of letting it decay into a dashboard fiction.

What Teams Can Apply

1. Compute cost per qualified lead, not cost per lead. Pull your sales dispositions and divide spend by qualified count only. For most Pakistani developers the honest number is 3 to 6 times the reported one — you cannot fix a number you do not publish.
2. Build the negative list from your own city’s query noise. Grade 90 days of search terms against one question — can this buyer afford my entry unit this year — and mine the failures into buckets. Expect the list to keep growing weekly.
3. Put budget and financing questions in the form. Two fields cut junk leads at the source and cost nothing. Reps spend their day on buyers instead of apologising to renters.
4. Separate the diaspora. Overseas Pakistanis convert on trust assets — USD pricing, walkthrough video, verification documents, timezone-matched callbacks — none of which fit a local campaign. Give them a cluster and measure them apart.
5. Feed dispositions back or the account learns the wrong lesson. Without offline conversion imports, Google optimises toward what it can see. What it can see unaided is the cheapest form fill in the market — which is exactly the junk you are trying to escape.