YouTube Demand-Gen for a Lahore Construction SaaS

The Client

A Lahore-based vertical SaaS startup selling construction project management software to mid-size Pakistani contracting firms — bill-of-quantities and estimation tools, site progress tracking, material procurement ledgers, and digital labour attendance registers. Founded in 2021 by two civil engineers, the company had grown to roughly 300 paying firms, PKR 29 million in annual recurring revenue, and an average contract value near PKR 96,000 a year. Gross margin sat at 74% and the average customer stuck around two years, which put lifetime value near PKR 142,000 per account on a gross-profit basis.

The buyer structure was the part a generic B2B playbook would misread. The signer is the firm’s owner or a director — usually someone running two to four concurrent projects, keeping the real material accounts in a register and a WhatsApp group, and deeply sceptical of software that promises to “digitise” what they can already do on paper. The internal champion is typically a quantity surveyor or project engineer in their late twenties who shortlists tools in English and demos them to the owner. Any acquisition channel that could not speak to both of those people at once was going to waste money speaking to neither.

Sales were field-led. A six-person inside team booked site and office demos, ran them with a screen share and, where possible, a walkthrough of a live project, and closed over a six-to-ten-week cycle. The product held up in evaluations and retention was healthy. The constraint was upstream: there were not enough qualified demos entering the funnel, and the ones that did arrive were getting more expensive every quarter.

The category itself was the structural problem. Construction management software has almost no search demand in Pakistan. Very few contractor owners type “construction management software” — in English or Urdu — into Google in any given month. The firm had been forced to buy borrowed, adjacent keywords, and it was paying borrowed-keyword prices for borrowed-keyword lead quality.

The Problem

The diagnostic surfaced six issues, and only two of them were media questions:

1. Blended CAC at PKR 51,000, up from PKR 37,000 in under two years. Payback had drifted past eight and a half months against a board target of six. The trendline was worsening every quarter, not stabilising.
2. Search spend on borrowed keywords. About 71% of a PKR 2.2 million monthly paid budget sat in Google Search on broad terms like “project management software” and “construction software free”. A search-terms pull showed 58% of clicks coming from queries that would never buy: students researching tools, job-seekers, and free-download hunters.
3. Junk lead ads. Meta lead ads consumed PKR 0.48 million a month and produced form-fills that sales could not use — roughly seven in ten were unreachable by phone or openly looking for employment, a pattern anyone running lead-gen forms in Pakistan will recognise.
4. No video and a failed LinkedIn experiment. The account held two search ad variations and zero video assets. A three-month LinkedIn trial had produced nine demos at PKR 19,000 each before being shut down.
5. Flat brand demand. Brand-search volume had sat at roughly 1,750 queries a month for four consecutive quarters. Nobody who was not already a customer had heard of the product.
6. Measurement gaps. Enhanced conversions were off, the Zoho CRM was not feeding demo outcomes back to the ad platforms, and every demo was credited last-click — which meant the team could not even see which half of the borrowed-keyword spend was doing the borrowing.

The strategic read was blunt: the team was competing for intent that barely existed, in an auction designed for a different market, with no mechanism for creating next quarter’s demand. The buyers did exist — they were on YouTube every evening, watching saria rate updates and estimation tutorials.

Phase 1 — Diagnosis and Measurement Rebuild (Weeks 1-3)

The engagement opened with measurement and audience research, not media, because a demand-gen channel judged on last-click attribution would be declared dead before it had a chance to compound.

Conversion tracking rebuild. We implemented enhanced conversions in Google Ads, defined the qualified-demo event in GA4 (a demo held for at least 25 minutes with a decision-maker or champion present), and built a Zoho offline conversion import that pushed demo-held, qualified, and closed-won stages back against both click and view interactions. This alone reattributed a meaningful slice of pipeline that had been silently parking itself on “direct”.

CAC tiering. A single blended number hides the mechanics. We split reporting into lead CAC, qualified-demo CAC, and closed-won CAC, and baselined the blended closed-won figure at PKR 51,000. The internal dashboard had been tracking lead CAC — which flattered the junk lead ads and masked the deterioration.

Search waste audit. We exported six months of search terms and tagged every query against a simple buyer/not-buyer rubric. The verdict: 58% of clicks were non-buyer traffic, concentrated in three broad-match campaigns. Exact-match brand defence and a handful of genuinely commercial terms were carrying the entire channel.

Audience research. This was the phase’s real deliverable. We compiled an inventory of 40-plus Urdu-language YouTube channels and content formats the buyer actually consumes: daily steel (saria) rate updates, brick and concrete estimation tutorials, site vlogs from housing schemes in Lahore and Gujranwala, and machinery reviews. Combined view counts on construction-topic Urdu content run into the millions monthly. None of the three funded competitors were running YouTube. It was open whitespace.

Baseline dimension	Finding at end of Week 3
Enhanced conversions	Off → on, with CRM offline import live
CAC tiers reported	Lead only → lead / qualified-demo / closed-won
Search clicks from buyer-intent queries	42% (58% waste tagged)
Brand-search volume	~1,750 queries/mo, flat four quarters
Urdu construction channel inventory	40+ channels mapped for placement and segment build
Competitor YouTube presence	None — whitespace confirmed

Phase 2 — Urdu Creative System and Audience Build (Weeks 3-7)

With measurement honest and the audience mapped, we built the engine. In demand-gen, creative is the primary lever — more than targeting, more than bidding — and for this buyer it had to be native to how the audience actually talks about the job.

Four creative pods, Urdu-first. All dialogue in Urdu, with English captions so the quantity-surveyor champions could evaluate silently at a desk:

• Problem-aware (28s, in-stream). A site-office scene: material disappearing from the ledger, an overrun discovered at billing time, the owner finding out weeks too late. No product pitch in the first fifteen seconds.
• Estimation demo (35s, in-stream). A screen capture building a BOQ for a small commercial unit in four minutes, with the Urdu voiceover counting what the manual version costs in estimator-hours.
• Social proof (20s, in-feed and retargeting). A Faisalabad contracting firm owner (an illustrative, anonymized account) describing, in plain Urdu, what changed in their monthly reconciliation.
• Shorts hook (12s, vertical). A single sharp idea — the day’s material ledger, on the owner’s phone before the mistri signs off — built for Shorts and in-feed discovery.

Custom segments from real search behaviour. Rather than importing generic B2B interest categories, we built custom segments from the Urdu YouTube search behaviour the research had surfaced: people who had searched for steel rate updates, BOQ tutorials, brick estimation formulas, and site-vlog queries. We layered the channel inventory as managed placements and used in-feed formats to reach viewers browsing construction topics.

Suppression before reach. Current firms, open pipeline, and recently churned accounts were suppressed across all paid media via customer match. Demand-gen budget that reaches people already in the CRM is budget that cannot show return and inflates apparent reach while hiding the real cost of net-new demand. The full creative-and-targeting architecture follows the scope we run for YouTube ads strategy and management engagements — the pairing of one pod to one audience is what keeps cost-per-qualified-demo interpretable in weekly scoring.

Objective discipline. Every campaign optimised to the qualified-demo event, never to views, reach, or impressions. A construction-category demand-gen campaign can hit any vanity number cheaply and still produce zero pipeline.

Phase 3 — Launch, Test, and Reallocate (Weeks 6-14)

Phase 3 is where spend moved and the CAC curve began to bend.

Sequenced reallocation. We moved PKR 0.55 million a month — about 25% of the search budget — out of the three waste-tagged broad campaigns into the YouTube layer, leaving exact-match brand defence untouched. In parallel, the Meta lead ads were cut entirely and a small retargeting layer was built on site visitors and video viewers instead. The point was not to abandon search; it was to stop renting irrelevant intent and start creating relevant intent where the buyers already were.

Weekly pod scoring. Every pod was scored weekly on cost-per-qualified-demo, combining click and view-through conversions, and on watch-time and view-through rate. The bottom quartile was cut or refreshed each week. Over eight weeks the blended cost-per-view fell from PKR 5.60 to PKR 3.40 as bidding learned and the creative improved — well inside the CPV range Pakistani in-stream inventory typically clears at.

Early demand signals. By week ten, brand-search queries were up 21% against a baseline that had been flat for a year, and direct traffic to the pricing page had picked up — the classic spillover of a demand-gen layer that is working before its attribution says so.

Pod performance at end of Phase 3:

Pod	Format	Cost-per-view	Cost-per-qualified-demo	Decision
Problem-aware	In-stream	PKR 4.90	PKR 4,100	Scale
Estimation demo	In-stream	PKR 4.10	PKR 2,700	Scale
Social proof	In-feed	PKR 2.60	PKR 3,300	Scale
Shorts hook	Shorts	PKR 1.90	PKR 7,400	Refresh hook, keep testing

Phase 4 — Scale and Compound (Months 3-6)

The final phase ran through the second quarter and turned a working test into a durable efficiency gain.

Scaling with a refresh cadence. The three winning pods scaled to roughly PKR 0.9 million a month in YouTube spend. Creative fatigue was the constraint — watch-time on a winning pod decayed measurably after about three weeks — so we held a fixed refresh rhythm, rotating new hooks into proven structures rather than restarting from zero. End-state budget: PKR 0.85 million in trimmed exact and brand search, PKR 0.9 million in YouTube, PKR 0.2 million in retargeting, PKR 0.15 million held for tests.

Honest attribution. We reconciled view-through and data-driven attribution against Zoho monthly. YouTube-influenced demos closed at 22%, against 14% for cold search — the pre-educated buyer arrives knowing what the BOQ screen does and asks procurement questions instead of first-principles questions. The optimisation target moved from lead CAC to qualified-demo CAC, junk bookings fell, and the average cycle shortened by roughly nine days. The economics of that shift — payback, LTV:CAC, and tiered CAC — follow the same discipline we apply in marketing unit economics and CAC payback modeling work.

Brand demand as the compounding asset. Brand-search volume finished quarter two at 64% above the pre-launch baseline. Brand clicks are the cheapest clicks in any account, and for a category with no organic search demand, grown brand search is functionally new market creation. By the end of the engagement the firm was no longer renting all of its demand from borrowed keywords.

Institutionalising the system. The pod library, weekly scoring sheet, and refresh calendar were handed over as a running operation, with a quarterly review that re-checks the segment inventory — Urdu construction viewership shifts with steel prices and housing-cycle news, and a custom segment built in January can be stale by Eid season. Two pods were retired in month five not because they failed, but because the placement inventory underneath them had moved on. Demand-gen decay is normal; the defence is a cadence, not a budget increase.

Final Results Over Two Quarters

Metric	Before	After (two quarters)	Change
Blended CAC (closed-won)	PKR 51,000	PKR 35,200	-31%
Qualified site-demos per month	41	56	+36%
Demo-to-close rate (overall)	14%	19%	+5 pts
Brand-search queries (monthly)	~1,750	~2,870	+64%
Cost-per-view (blended)	PKR 5.60	PKR 3.10	-45%
CAC payback	~8.5 months	~6 months	-2.5 months
LTV:CAC	2.8x	4.0x	+1.2x

Every line traces to a phase: the CAC drop to the waste coming out of broad search and the cheaper net-new demand in Phase 3, the close-rate lift to pre-educated buyers from the Urdu creative system in Phase 2, brand-search growth to two quarters of consistent demand-gen presence, and the LTV:CAC expansion to CAC falling while retention held.

What Made This Work

1. The channel went where the buyers already were. Construction decision-makers in Pakistan were never going to generate search demand for this category in volume — but they were watching Urdu construction content nightly. Meeting the buyer in their existing media habit beat trying to manufacture search behaviour that was not coming.
2. Creative was native, not translated. Urdu dialogue with English captions let one asset move both the owner and the quantity surveyor. A dubbed English script with an Urdu subtitle track would have read as foreign in the first three seconds and skipped.
3. Segments came from observed behaviour, not imported categories. Custom segments built from steel-rate and BOQ searches reached a construction audience with a precision that “business software interests” never could, at costs that made the maths work.
4. Suppression protected the blended number. Keeping current firms and open pipeline out of the audience meant every rupee of demand-gen spend was aimed at net-new logos, which is the only way a CAC reduction survives contact with the CRM.
5. The objective never drifted to vanity. Optimising to qualified demos — never views or reach — meant the platform’s learning loop was solving the problem the sales team actually had.

What Teams Can Apply

For a Pakistani B2B SaaS team watching CAC climb in a low-search-demand category:

1. Audit search waste before blaming the channel. Tag six months of search terms as buyer or non-buyer. If more than half your clicks are students and job-seekers, the problem is not Google — it is that you are buying someone else’s keywords.
2. Check whether your buyers live on YouTube before assuming they live on search. Urdu-language professional content — rates, tutorials, site vlogs — has real, measurable audiences in construction, agriculture, textiles, and trading verticals. Where search demand is thin, video demand is often not.
3. Build the offline conversion loop first. Without CRM-fed demo and closed-won stages plus view-through windows, any demand-gen test will be judged last-click and killed before it compounds. This is a two-to-three-week prerequisite.
4. Optimise to the qualified event, not the lead. Cutting the junk lead ads and moving to qualified-demo CAC shortened the sales cycle and freed rep capacity — a bigger blended-CAC lever than any bid change.

This illustrative composite reflects the patterns WeProms Digital sees across Pakistani vertical SaaS companies — the personas, budgets, and creative angles shift with each buyer profile, but the measurement-first, native-creative, suppress-the-CRM sequence stays consistent. The acquisition-cost pressure facing tech startups in Pakistan is precisely what this demand-gen layer is built to relieve.