Can one celebrity beat that engine
on cost per enrolment?

Curious Jr already runs a mature engine of UGC, faculty-led, creator and AI ads: 400 ads live in the window, 357 of them non-celebrity.

ClientCurious Jr

CelebrityShailesh Lodha

WindowJun 1 – Aug 11

01THE SETUP

One shoot day,
built like a system.

Curious Jr already runs a mature engine of UGC, faculty-led, creator and AI ads: 400 ads live in the window, 357 of them non-celebrity. The question: can one celebrity beat that engine on cost per enrolment, not on reach?

STEP 011Shoot day
modular shoot architecture, built for variants
STEP 0214Scripts
8 objection-led concepts + 6 short awareness scripts
STEP 032–3Hooks per concept
h1 / h2 / h3 openings tested independently
STEP 0443Live ad assets
full-length, 15-second, 6-second and internal cuts

The celebrity got no exemption from testing discipline: every hook launched, measured and killed on the same data as the UGC engine. 31 of the 43 assets never took more than 1.7% of celebrity spend between them.

02THE HEADLINE, QUANTIFIED

27% of the money.
32% of the enrolments.

SHARE OF VIDEO COHORT · META

Shailesh Lodha ads vs. every other video ad in the account, Jun 1 – Aug 11.

AND THE RESULTS CAME CHEAPER

−22%

cost per enrolment

−44%

cost per completed view

A bigger share of results than of budget, at a lower cost per result, and a far bigger share of attention than of spend.

Sources · Meta Ads Manager. Enrolments = pixel-tracked transaction & batch-enrolment success.

03META · EFFICIENCY

Enrolments 22% cheaper.
Attention 44% cheaper.

META · COST INDEXED, REST OF ACCOUNT = 100

100

rest of accountShailesh Lodha ads

cost / enrolmentindexed · 10 delivering assets of 43

100

rest of accountShailesh Lodha ads

cost / completed viewindexed · steepest gap

The attention gap is the tell: a face India codes as the reasonable, un-hysterical parent earns the stop before the pitch, so the same media money buys far more watched seconds and a 10% cheaper CPM.

The single biggest ad in the whole account carried Shailesh Lodha’s face.

Blended vs. all non-celebrity video in the same conversion cohort, same period.

04META · FULL-FUNNEL

The advantage held at every stage.

STOPHook rate · 3-sec plays / impressionsA recognised face earns the pause in-feed.+55%
WATCHThruPlays per impressionMore impressions became finished views, not just stops.+61%
HOLDCompletion rate · ThruPlays / 3-sec playsThose who stopped stayed: the script held them.+4%
BUYCPM · cost per 1,000 impressionsMeta priced the creative cheaper because people watched it.−10%
PAYCost per enrolmentAttention converted into paid enrolments, not just views.−22%

Not a top-of-funnel novelty spike. The lift compounds from first impression through to the paid enrolment at the end of it.

Relative deltas vs. non-celebrity video · same cohort, same period · blended across all celebrity video, including the scaled winner

05THE HERO AD

One script became the
biggest ad in the account.

One ad (the “khelna-koodna band” narrative, hook 1) absorbed 10% of all account spend and returned 13% of all enrolments, running across 13 ad sets.

−21%

cost per enrolment vs. account average, while running as the most-scaled ad on the platform.

−25%

cheaper per enrolment than the biggest non-celebrity ad, at comparable spend.

Illustrative: actual values indexed.

In its three biggest audiences it ran 9–16% below even its own blended cost per enrolment. Concentrating budget behind it made it cheaper, not dearer.

06WHY IT WORKED

Celebrity as a performance asset,
not a billboard.

01
Credibility transfer
For a product sold to anxious parents, we cast the face India already codes as the calm, reasonable head of the household. The trust bought the pause: a 55% higher hook rate and 44% cheaper attention.
02
Objection-led scripting
14 scripts mapped to the real parent objections (play-time vs. study-time, ratta-maar learning, Olympiad ambition, personal attention, practical concepts), so every segment met the celebrity in its own argument.
03
Hooks tested like UGC
Three hooks per concept, launched and killed on data. 31 of 43 assets never cleared 1.7% of spend between them; the winner was allowed to absorb 63%. Strip that winner out and the rest ran near parity, which is the point. The system exists to find the one and fund it hard.
04
Length earned its place
Full-length narrative indexed at 100 on cost per enrolment; 15-second awareness cuts 118, 6-second cuts 135, internal re-cuts 199. Trust needs room to land, so we scaled the long cut and stopped funding the short ones.

07THE TAKEAWAY

Trust converts.

Celebrity creative is usually bought for brand and forgiven on performance. Run through a performance system (objection-mapped scripts, tested hooks, modular cuts, ruthless kill rules), it beat a mature 357-ad engine on its own scoreboard, in the same auctions, on the same days.

−22%

CPO

−44%

cost / view

32%

of enrolmentsfrom 27% of spend

Sources · Meta Ads Manager, Jun 1 – Aug 11 2026

08SHAILESH LODHA