Work

Creative, built to be bought.

Short-form video our team produced for paid placements, built to sell products, services and subscriptions: client spots in English and Spanish, product campaigns across personas and accents, and animated product demos. Plus a look inside the accounts they run in.

Client work

Made for real businesses.

Spanish-language spots for a South Florida car-inspection company, written for Miami's Spanish-first buyers. Each piece shows the idea behind it and a full transcript.

Language test

Spanish-first buyers, in their own Spanish.

Variable: Language and dialect: native Cuban Spanish for Miami, with the brand URL kept in English.
Held constant: The same true-story structure as the English versions: the loss, the villain, the new rule.
How we call it: Cost per lead and booked-inspection rate by language and audience. In Miami, Spanish-language creative often reaches buyers English ads never do.

A bilingual market is two markets. Translating isn't enough; the speaker, dialect and story have to belong to the audience.

Creative testing

One variable at a time.

This is how we structure creative for scale. Each set changes one thing (the persona, the creator, the angle, the format or the proof) and holds everything else constant, so the result tells us something we can act on. The videos are product campaign creative our team produced.

Persona test

Same scare, two different messengers.

Variable: Who delivers the hook: a rural Southern dad or an Atlanta creator.
Held constant: Same offer (vehicle protection), same moment (the repair estimate), same 7–8 second length.
How we call it: Hook rate and cost per lead by persona, then lead quality downstream. The winner tells us who the message actually reaches, not who we assumed would buy.

On Meta, the face and voice in the first second choose the audience. Testing personas is audience targeting done through creative.

Creator test

One product, his voice and hers.

Variable: The creator: a woman in the gym versus a man before leg day.
Held constant: Same pre-workout, the same three benefits (pump, focus, no crash), same setting.
How we call it: Thumb-stop rate, hold rate and CPA by creator. If both win, they scale in parallel to reach different buyers instead of competing for the same impressions.

Creator fatigue is the first thing to cap scale in fitness and supplements. A bench of proven creators keeps spend growing.

Angle test

Same jar, different reason to buy.

Variable: The problem the product solves: camera-ready skin versus sun and salt damage.
Held constant: Same skincare product, the same UGC format and length.
How we call it: CTR and conversion rate by angle. Angles that convert become separate concepts with their own hooks; angles that only earn clicks get cut.

Most products have several buyers with different reasons. Finding the reason that converts is worth more than polishing the ad.

Format test

Talking head or tactile demo.

Variable: Format: a presenter talking to camera versus a claymation product demo.
Held constant: Same category (skincare, hydration), same 7–8 second slot, the same simple benefit.
How we call it: Hook rate and cost per purchase by format. Animation often wins attention in feeds full of faces; the question is whether it also wins the sale.

Format diversity fights ad fatigue and gives the delivery system genuinely different creative to work with.

Proof test

Five ways to make a claim believable.

Variable: The type of proof: insider (the clerk who restocks it), referral ("my boy showed me"), savings, a founder's workflow, and the checkout moment.
Held constant: UGC format, one speaker, one line, under 8 seconds.
How we call it: Which proof style lifts conversion rate per vertical. Insider proof tends to work for retail products, referral for considered purchases, savings for budget-led buyers.

Proof is the gap between a click and a purchase. Matching the proof type to the buyer is a structural choice, not a copywriting flourish.

Inside the accounts

The structure behind the creative.

Great ads still need the right account around them. These views show how we split, test, cap and scale on Google and Meta, and how we check the results against the business.

Google Ads E-commerce brand · United States Last 30 days

Conversions44.7K

Cost$1.79M

Cost / conv.$40.14

Conv. value / cost3.10

Google Ads campaigns, Last 30 days
Campaign	Status	Budget	Impr.	CTR	Cost	Conversions	Cost / conv.	Conv. value / cost
SRCH \| Brand \| Exact¹Search · Target impr. share Capped	Limited by budget	$3,500.00/day	1,380,000	23.04%	$96,400.00	9,820	$9.82	12.22
SRCH \| Non-brand \| High intent²Search · Target CPA Scaling	Eligible	$22,000.00/day	6,910,000	7.05%	$611,800.00	13,640	$44.85	2.78
SRCH \| Non-brand \| Category \| Broad³Search · Target CPA Active	Eligible	$10,000.00/day	9,240,000	3.81%	$288,300.00	5,060	$56.98	2.09
SRCH \| Competitor \| ExactSearch · Max. clicks (capped CPC) Limited	Eligible (Limited)	$2,500.00/day	1,120,000	3.72%	$64,200.00	820	$78.29	1.44
SHOP \| Hero SKUs \| Margin tier A⁴Shopping · Target ROAS 340% Scaling	Eligible	$12,000.00/day	14,800,000	2.72%	$341,000.00	8,910	$38.27	3.48
PMAX \| Feed only \| Brand excluded⁵Performance Max · Target ROAS 260% Testing	Eligible (Learning)	$9,500.00/day	22,600,000	1.31%	$274,100.00	5,380	$50.95	2.49
DGEN \| Prospecting \| Video⁶Demand Gen · Max. conversions Active	Eligible	$4,000.00/day	31,400,000	0.38%	$118,200.00	1,060	$111.51	1.02
Total: all campaigns		$63,500.00/day	87,450,000	2.30%	$1,794,000.00	44,690	$40.14	3.10

Why it's built this way

1Brand is split out and capped. Branded clicks are cheap and convert well, so blending them in makes non-brand look better than it is.
2High intent is organized by search theme, not by product SKU. Each ad group points at a page that answers the query.
3Broad match only runs where conversion volume can steer it, with search terms reviewed weekly and negatives added.
4Shopping is split by margin tier, so the bid target reflects profit, not just revenue.
5Performance Max runs feed-only with brand excluded and is reported on its own, so it can't take credit for brand demand.
6Video prospecting is judged on blended lift and branded-search growth, not on platform-reported conversions alone.

Illustrative account · names generalized · not a specific client's results

Ads Manager Subscription app · United States Last 30 days

Amount spent$2.19M

Purchases49.8K

Cost per purchase$43.96

Purchase ROAS1.80

Meta ad sets in ASC | Prospecting | Purchase · Value optimization, Last 30 days
Ad set	Delivery	Budget	Results	Cost per result	Amount spent	CPM	Hook rate	Purchase ROAS
PROS \| Broad \| Concept test \| Wk 40¹Highest value · Testing	Learning	$6,000.00Daily	3,480Purchases	$49.14Per purchase	$171,000.00	$10.75	32.0%	1.58
PROS \| Broad \| Scale \| Winners²Highest value · Scaling	Active	$38,000.00Daily	26,100Purchases	$43.22Per purchase	$1,128,000.00	$11.46	34.0%	1.86
PROS \| Advantage+ audience \| ScaleHighest value · Active	Active	$22,000.00Daily	13,700Purchases	$46.79Per purchase	$641,000.00	$10.47	32.0%	1.69
PROS \| LAL 1% payers \| Holdout control³Highest value · Active	Active	$3,000.00Daily	1,610Purchases	$54.97Per purchase	$88,500.00	$12.83	30.0%	1.48
RT \| Engaged 30d \| Capped at 8% of spend⁴Highest value · Capped	Active	$5,500.00Daily	4,920Purchases	$32.72Per purchase	$161,000.00	$16.43	28.0%	2.28
Results from 5 ad sets		$74,500.00	49,810	$43.96	$2,189,500.00	$11.39	32.8%	1.80

Why it's built this way

1New concepts enter a dedicated test ad set with its own budget, so testing never resets the learning in the scaling campaigns.
2Winners graduate as concepts, not single ads, with fresh hooks queued before the current ones fatigue.
3A small lookalike cell runs as a control to check whether broad delivery is actually finding better customers.
4Retargeting is capped as a share of spend, so it can't inflate results by claiming people who would have bought anyway.

Illustrative account · names generalized · not a specific client's results

Tracker CPL and paying-user offers · all traffic sources Last 30 days

Visits900.0K

Conversions16.5K

Profit$127.7K

ROI33.5%

Tracker report by traffic source and lander, Last 30 days
Traffic source / lander	Visits	CTR	Conv.	CR	Revenue	Cost	Profit	ROI	EPV
Meta · PROS \| Broad	412,300	31.99%	9,870	2.39%	$296,100.00	$214,800.00	$81,300.00	37.8%	$0.718
LP-03 · Quiz → calculator²	208,400	38.00%	6,110	2.93%	$183,300.00	$108,600.00	$74,700.00	68.8%	$0.880
LP-01 · Advertorial	203,900	25.85%	3,760	1.84%	$112,800.00	$106,200.00	$6,600.00	6.2%	$0.553
Google · SRCH \| Non-brand	96,500	43.01%	3,420	3.54%	$116,300.00	$71,900.00	$44,400.00	61.8%	$1.205
Native · Content rec.	233,000	20.00%	2,010	0.86%	$60,300.00	$52,400.00	$7,900.00	15.1%	$0.259
Push · New source test³	158,200	19.03%	1,190	0.75%	$35,700.00	$41,600.00	−$5,900.00	-14.2%	$0.226
Total	900,000	27.79%	16,490	1.83%	$508,400.00	$380,700.00	$127,700.00	33.5%	$0.565

Why it's built this way

1The tracker, not the ad platforms, is the source of truth for revenue: conversions arrive by server-side postback and are matched to the click.
2Landers are compared on the same traffic. The quiz lander converts better, so it takes the budget and the advertorial gets a new angle.
3A new source runs as a capped test with a decision date. This one is under water, so it's fixed or cut, not scaled on hope.

Illustrative account · names generalized · not a specific client's results

Reporting Consumer brand · all paid channels Last 30 days

Paid media spend$2.86M

Revenue (business)$4.91M

MER (blended)1.72

New-customer CAC$104.76

Platform-reported conversions vs. orders the business recorded

Platforms claim 51,400

Business recorded 38,200

Google AdsMetaOther paid

35% over-claim. Budgets set on platform totals would fund conversions that happened once but were counted twice.

Illustrative account · names generalized · not a specific client's results

About this page

What you're looking at

Client work is shown with permission and credited to the business it was made for.

Product campaigns are real performance campaign creative our team produced to promote the products shown. The product brands didn't commission them, and they haven't endorsed MSG.

Presenters and characters in these videos are AI-generated, and each piece is labeled. They aren't real customers, and we don't present them as such.

The account views are illustrative: they show how we structure, test and read accounts, with names generalized. They aren't a specific client's results.

Want ads built to be tested?

Share your goals, your channels and roughly what you spend. We'll reply with an honest view of where we'd start, or tell you if we're not the right fit.

Discuss your growth See our approach