YouTube Ads Case Study in Pakistan

The Client: A Karachi Professional Training Institute Network

A network of three professional training campuses across Karachi — Gulshan, Tariq Road, and Bahadurabad — offering IT bootcamps (full-stack web, data analytics, cybersecurity), a one-year digital marketing diploma, professional certification preparation (PMP, ACCA fundamentals), and IELTS. The institute enrolled roughly 1,400 students per batch across its programs, with average program fees between PKR 28,000 for short courses and PKR 145,000 for the year-long diploma. Annual revenue sat around PKR 380M, with growth targets tied to opening a fourth campus in Hyderabad within twelve months.

Their acquisition mix leaned heavily on Meta (Facebook and Instagram) lead-form ads, Google Search for branded and high-intent commercial keywords (“digital marketing course in Karachi”), walk-ins from campus visibility, and a small, unstructured YouTube spend that a previous agency had set up as a single broad awareness campaign. Enrolment cycles were long and considered — students typically compared three to five institutes, attended counselling sessions, and decided over two to four weeks. Counsellors were the real conversion engine: once a qualified student reached a phone or campus conversation, roughly one in five enrolled.

When the leadership team approached WeProms Digital, the problem was narrow and specific. Their Meta cost per lead had risen 38% year-on-year as competing institutes flooded the auction, branded search was tapped out, and the YouTube line item was burning PKR 900,000 a month with no measurable enrolment tied to it. They wanted YouTube to do what it should do for an education business — build demand at the top of the funnel so that search, retargeting, and counsellor outreach closed enrolments more cheaply. Instead, YouTube was a black box. The work was scoped as a YouTube ads strategy and management engagement built around enrolment outcomes rather than views.

This page reviews the diagnostic, the creative and targeting rebuild, the offline conversion tracking fix, and the measurable outcomes at the 90-day mark.

The Problem: Views Without Enrolments

Four blockers were holding the YouTube channel back:

• No enrolment-level tracking. YouTube was optimised for views and “engagement” by the previous agency. Nothing connected a video view or a click to a form submission, a counselling call, or an enrolled student. Google Ads reported 1.2M impressions and 380,000 views a month; the CRM reported nothing attributable.
• A single bloated awareness campaign. All programs, all audiences, all locations in one ad group, served against a generic brand film and three recycled television commercials. Frequency sat at 14+ against the same warm audience and near-zero against the cold prospects who were actually in-market.
• Creative that didn’t match the search for education. The ads sold the institution (“25 years of excellence, ISO-certified”). Students evaluating a course were weighing curriculum, instructors, placement outcomes, and fees. The creative answered none of the actual buying questions.
• A broken measurement loop with the counselling team. Counsellors logged leads in spreadsheets and a legacy CRM with no source attribution. When YouTube-sourced leads did arrive, they were coded as “Facebook” or “direct,” making it impossible to optimise bidding or audiences against real enrolment outcomes.

The net effect: at PKR 900,000 a month, YouTube was contributing an estimated 6% of enrolments — but because nothing was measured, leadership could not tell whether to scale it, fix it, or kill it.

Phase 1 — Diagnosis and Offline Tracking Rebuild (Weeks 1–2)

The first two weeks were not creative work. They were plumbing. Without enrolment-level data, every later optimisation would be guesswork.

Offline conversion import setup. The decisive move was connecting the institute’s CRM to Google Ads through offline conversion tracking. Every form submission and counselling-call booking was tagged with a GCLID and a landing-page session ID. As a student moved through the counselling pipeline — enquiry → demo class attended → enrolment — those CRM stage changes were pushed back into Google Ads as offline conversions at defined values:

CRM stage	Imported as	Conversion value
Qualified enquiry (counsellor reached)	“Lead”	PKR 1,200
Demo class attended	”Qualified”	PKR 4,500
Enrolled (first fee instalment paid)	“Enrolment”	Average program fee

This gave YouTube’s bidding system real enrolment signals instead of vanity views. We set the “Enrolment” conversion as the primary bid goal for max conversion value bidding.

GA4 and consent hygiene. We rebuilt the GA4 configuration to fire a generate_lead event on form submission with program-specific parameters, and stood up enhanced conversions for leads so email and phone hashes could stitch cross-device journeys. A consent mode implementation kept the setup privacy-compliant without gutting measurement.

Audience and source attribution cleanup. The counselling team’s CRM intake form was restructured so source capture was mandatory and structured (a dropdown, not free text), with “YouTube ad” as an explicit option. Within two weeks the share of leads coded as “direct / unknown” dropped from 31% to 14%.

Creative and message audit. We reviewed eighteen months of creative alongside the actual search-intent data: what queries brought students to the site, what they read, where they dropped. The gap was obvious — the ads talked about the institute; the students were researching outcomes.

Phase 1 produced no creative output, but it set up everything that followed. By the end of week 2, we had a clean conversion pipeline feeding Google Ads enrolment data with a 72-hour attribution lag.

Phase 2 — Creative and Audience Restructure (Weeks 3–5)

With measurement in place, we rebuilt the YouTube programme from the ground up.

Campaign restructure. We retired the single awareness campaign and built a three-layer structure:

• Demand-gen layer (the workhorse). A YouTube demand-gen campaign (instream plus in-feed) segmented by program vertical: IT bootcamps, digital marketing diploma, professional certifications, and IELTS. Each vertical had its own ad group, its own creative, and its own audience.
• High-intent layer. A YouTube campaign targeting custom-intent audiences built from the institute’s highest-converting search keywords (“best data analytics course Karachi,” “PMP certification cost Pakistan,” “ACCA vs CA”) — capturing students actively researching.
• Retargeting layer. Light-touch remarketing to website visitors and engaged-viewer audiences (people who watched 30 seconds or more of an ad) with counselling-booking hooks.

Audience strategy. The previous setup served everyone in Karachi. We narrowed to in-market and affinity audiences that mapped to education decisions: “Education,” “Job Seekers,” plus custom segments around specific career transitions (recent graduates and mid-career professionals reskilling). We layered demographics carefully — 20–32 for the diploma and bootcamps; 26–42 for PMP and management certifications.

Creative rebuild — answer the buying questions. This was the highest-leverage change. We produced 14 new video assets across four formats, built around the actual decision criteria students used:

• Outcome-led instream ads (15s and 30s) opened with the job outcome (“From commerce graduate to data analyst in six months”) and backed it with placement numbers and a graduate testimonial.
• Curriculum walkthrough ads offered a fast-cut tour of what a student would actually learn week by week, addressing the “what will I be able to do” question.
• Instructor credibility ads were short profiles of lead instructors with their industry background.
• Fee-transparency ads were a 20-second spot acknowledging cost openly, covering the instalment plan and the money-back guarantee for the first module.

We deliberately cut the “ISO-certified, 25 years of excellence” framing from primary rotation. It moved to a 6-second bumper used purely for remarketing frequency. Roughly 40% of creative ran in Urdu (or mixed Urdu-English) for broader reach, and 60% in English for the professional and certification audiences who research in English. Captions were burned in — a meaningful share of YouTube viewing in Pakistan happens on muted mobile devices.

Phase 2 launched at the end of week 5. We held budgets flat for the first ten days to let the bidding system calibrate against the new offline conversion signals before scaling. The counselling-to-enrolment flow this feeds into is part of a broader lead generation systems and funnel building discipline that WeProms applies across education clients.

Phase 3 — Optimise and Scale on Enrolment Signals (Weeks 4–8)

Once the new structure had two weeks of clean conversion data, we moved from stabilise to scale.

Bid strategy migration. We migrated the demand-gen layer from target-CPA (set against lead volume) to a max conversion value strategy with a target ROAS floor, now that enrolment values were flowing back into Google Ads. This single change shifted delivery toward the audiences and creatives that produced enrolled students — not just form-fillers.

Creative performance divergence. The early data was unambiguous about what worked:

Creative format	Cost per enrolled lead	View-to-enrolment rate
Outcome-led instream (testimonial)	PKR 7,950	0.41%
Curriculum walkthrough	PKR 8,640	0.33%
Fee-transparency 20s	PKR 10,200	0.28%
Instructor credibility	PKR 11,400	0.22%
Legacy brand film (control)	PKR 22,600	0.06%

The outcome-led testimonial ads delivered enrolled students at roughly one-third the cost of the legacy brand film. We reallocated 70% of spend toward the top three formats and paused the brand film from cold prospecting. We also stood up a YouTube Shorts component for the IT bootcamp vertical, where the target audience skewed younger and Shorts consumption was highest — a dedicated Shorts ad group drove cheap qualified views and a measurable lift in branded search for that program.

Audience pruning and expansion. The “Job Seekers” affinity segment looked strong on cost-per-view but produced low enrolment quality — students without the means or readiness to commit to a year-long diploma. We pruned it for the diploma ad group and narrowed to custom-intent and in-market Education audiences. Conversely, the custom-intent audience built from ACCA and PMP keyword lists performed 2.3x better than any affinity segment for the certification vertical, so we expanded it.

Frequency capping and decay. We capped cold-prospect frequency at 4 per week (down from the prior unmanaged 14+) and moved audiences into the retargeting layer after a 30-second engaged view, where cheaper bumper and in-feed formats carried them to a counselling booking.

Counsellor feedback loop. A weekly 30-minute review with the counselling lead fed qualitative signal back into creative — which ads students mentioned on calls, which objections recurred. Two new creatives came directly from counsellor-flagged objections: “the diploma is too long” produced a fast-track 8-month-path creative, and “I’m not sure I’ll get placed” produced a placement-stats creative.

Phase 4 — Measure, Validate, and Compound (Weeks 8–12)

The final phase turned the rebuilt channel into a predictable, compounding acquisition engine.

Closed-loop incrementality check. At week 10 we ran a geo-holdout test — pausing YouTube demand-gen spend in the Hyderabad catchment (where they had no campus yet but were capturing online enrolments for the diploma) while holding spend steady in Karachi. Over 14 days, the held-out region saw a 23% drop in diploma enquiries with no compensating lift from Meta or search — a clean signal that YouTube was generating net-new demand rather than cannibalising other channels. Spend resumed with confidence.

Budget reallocation. With a clear efficiency advantage, the leadership team shifted budget within the paid mix: Meta lead-form spend was reduced 18%, and YouTube demand-gen spend was scaled from PKR 900,000 to PKR 1.6M per month — funded entirely by Meta savings rather than new budget.

View-through attribution. Because education journeys are long and often span devices, we set view-through windows to 7 days and modelled view-through enrolment contribution. By week 12, view-through accounted for 27% of YouTube-attributed enrolments (up from a negligible baseline), confirming that many students watched, left, searched the brand, and enrolled later — value the previous view-only setup had never captured.

Compounding effects on branded search. The demand-gen layer lifted branded search query volume for the institute’s name by 34% over the 90 days, at a CPC roughly half of their non-branded course keywords — YouTube was feeding the search funnel more cheaply than search could feed itself. This kind of demand-gen-to-search compounding is one of the recurring patterns we see across digital marketing for educational institutions, where considered enrolment decisions benefit from a top-of-funnel video layer.

Final Results at 90 Days

Metric	Before	After (90 days)	Change
Cost per enrolled lead	PKR 14,200	PKR 9,080	-36%
Qualified enquiries (monthly)	1,180	1,665	+41%
Marketing-attributed ROAS (tuition)	2.9x	4.6x	+59%
Enrolment rate from enquiry	18%	24%	+33% relative
YouTube share of enrolments	6%	19%	+13 pts
View-through enrolment share	~0%	27%	—
Branded search volume	Baseline	+34%	—
Monthly YouTube spend	PKR 0.9M	PKR 1.6M	Reallocation

The 36% reduction in cost per enrolled lead and the lift to 4.6x ROAS were achieved at a higher absolute spend — meaning enrolment volume grew while unit economics improved, which is the only kind of efficiency gain that matters for a growth-stage institute.

What Made This Work

1. Enrolment signals beat view signals. The single most important change was importing offline enrolment conversions back into Google Ads. Until YouTube’s bidding system could “see” an enrolled student and its value, it was optimising against views — which optimises for cheap attention, not paying students. The Phase 1 plumbing unlocked every later decision.
2. Creative that answered the buying question. Education is a considered purchase with specific decision criteria — outcomes, curriculum, instructors, cost. The outcome-led testimonial creative outperformed the legacy brand film by 3x on cost per enrolled lead because it spoke to what students were actually evaluating. Brand films have a role; it is remarketing, not cold acquisition.
3. Segmentation by program, not by institution. A diploma student, an IT bootcamp student, and a PMP candidate are three different buyers with different search behaviour, different audiences, and different objections. One campaign could not speak to all three. Vertical-level ad groups and creative were non-negotiable.
4. A real counselling-team feedback loop. The weekly check-in with counsellors surfaced objections that produced two of the best-performing creatives. Creative generated in a vacuum underperforms creative pressure-tested against the people closing enrolments every day.
5. Incrementality over attribution. The geo-holdout in Phase 4 was what gave leadership the confidence to scale. Platform attribution alone is unreliable for long-cycle education journeys; a clean holdout proved the channel was generating net-new demand.

What Teams Can Apply

For Pakistani education providers — training institutes, test-prep centres, private schools, certification bodies — looking to make YouTube earn its budget:

1. Wire offline enrolment into Google Ads before anything else. If your CRM stage changes are not being imported as conversions, you are optimising YouTube blind. This is the highest-ROI two-week project available to most education marketing teams.
2. Build creative around student decision criteria, not institutional pride. Audit your current YouTube ads against the four questions students actually ask: outcome, curriculum, instructor, cost. If your ads answer none of them, rebuild.
3. Segment by program. Resist the one-campaign-fits-all structure. Different programs serve different audiences with different economics; structure your account to reflect that.
4. Set a frequency cap and use a demand-gen to retargeting flow. Burning impressions against the same warm audience at 14+ frequency is wasted budget. Cap cold frequency, graduate engaged viewers to remarketing, and let cheaper formats carry the bottom of the funnel.
5. Validate with a geo-holdout before you scale. Attribution dashboards will tell you YouTube is working whether or not it is. A simple held-out region test tells you whether the spend is net-new demand or borrowed from another channel — and it costs you nothing but two weeks of patience.

WeProms Digital has applied this YouTube demand-gen framework across Pakistani education providers in Karachi, Lahore, and Islamabad — from professional training networks to test-prep franchises and private higher-education institutions. The program-level segmentation, creative angles, and tracking architecture adapt to each vertical, but the core principle holds: measure enrolment, answer the buying question, prove incrementality.