Angle-first creative testing

Creative should never be the reason you can't scale spend.

Anyone can generate 50 variants. Knowing which 3 are worth testing is the part nobody automates — for DTC, consumer app and SaaS teams, at a pace (12 tested hooks a month) none of them look mass-produced.

A buyer problem is the opening scene.Then the product earns attention.

Video is the output.

The actual work is finding the product truth worth testing, then turning it into creative people can understand before they scroll.

Built for
DTC, consumer app and SaaS teams without an in-house creative bench.
Formats
Paid social hooks, explainers, UGC-style, product film, personalized ABM.
Judgment
Every hook tests a different hypothesis, not just a different edit — the pick is the work, the volume is proof it happened.
Working
Remote, on overlapping US and UK hours.

Selected independent concepts

Same discipline.
Three different markets.

Concept pieces built from real product context. Physical product, consumer app and software buy for different reasons, so each gets its own vocabulary instead of one template with the logo swapped.

Physical product

Nobody buys a candle or a skincare syringe off a feature list. Desire is the job here, not comprehension, so the camera stops narrating and lets the object and the room do the argument.

Uvola / supplements

The pitch livesin the cabinet.

Not in the product. A shelf already crowded with bottles nobody finished, so the hook isn't "try this." It's recognition: you own this exact cabinet. Shipped as a set, because a market this crowded needs more than one cut to test against: the same angle in English and Spanish, plus a written note on which line is doing the work and why, so the next version isn't guesswork.

01Recognition hook, UGC-style: no presenter selling
02Two languages off one angle, not a translation job
03A creative note, so the test can be repeated
Cut
30s / English
Open PDF ↗
Creative note / page 1
On demand

Avatar-led DR cut / Uvola

The case above is mine end to end: angle, script, output. This is the other way to work with me. The brand supplied script and footage; the avatar lip-sync, the motion, the transitions and the cut are mine. Avatar-led cuts are also where volume comes from, cheap enough to carry a test so twelve hooks a month is possible, and the winner gets re-shot with a real founder.

2:16 / six problem-solution beats, price anchor, full offer close

Studio

Theillon

The applicator is shaped like a syringe, and the brand's own instructions spend real effort making sure you don't mistake it for one. The whole film's job is disambiguation before persuasion: massage, wait, seal. A topical routine, not an injection. You can't sell a system nobody understands.

18s / skincare, studio

Product film

JF Fantasy Jewelry

JF's Spring Night Garden Stroll has too much story for a catalogue spin. The film starts close enough to read the twin serpents, flowers and Rainbow Moonstone, then lets those details open into a rain-dark garden. It returns to the ring in clean light at the end. The atmosphere earns attention; the final product view keeps the fantasy tied to the piece.

24s / jewelry, cinematic product film

Ancaria

A skincare brand built in partnership, not a spec concept and not a client brief. Packaging, material and light get decided the same way an angle does: pick the one true thing, then remove everything that competes with it.

Ancaria acne treatment serum in dark blue glass with a brushed silver lid, lit on a stone plinth
Ancaria cream jar with a concentric machined silver lid on clear glass
Ancaria bottle material study in glass, brass and black gloss

Packaging & product direction / brand partnership

Consumer app

The buyer and the user are the same person, scrolling on their own time. No procurement process, just a reason to stop and a reason to trust. Consideration collapses into seconds, and the whole job is the first three.

HAVIT / metabolic health app

The constraintbecame the concept.

HAVIT replied to a cold pitch and confirmed the angle: day 1 vs. day 30, tracked with real in-app data. Then set a boundary: no exaggerated before/after on the body. Their team came out of JUVIS and Noom, so the caution was clinical instinct, not fear of a rule. Two treatments off that same line: TikTok-native for the feed, then a slower ritual cut once there was a real brief to build from.

01TikTok organic: recognition over demonstration
02The Ritual: same morning routine, day 1 vs. day 30, zero body claims
33 seconds / TikTok organic
18 seconds / The Ritual

B2B software

The person who watches rarely signs. Each cut does one job: make the problem legible, or make the mechanism believable. Never both.

Cranston / accounting AI

One product.Two useful truths.

The explainer sells capacity, not a new hire. Henderson's 2025 return: 14 source documents, 12 minutes 40 seconds, signed off before the reviewer's coffee goes cold. Every figure links back to its source document, so the demonstration is the ad. For the feed, the hook opens on the week's most expensive hour: the one spent retyping what the software could have read.

01Capacity angle for a decision-maker
02Cost angle for a scrolling peer
38 seconds / Explainer
12 seconds / Social hook

Maia / medical coding AI

The fear tax.

Not because it's wrong. Because a lower code almost never gets denied. Their own site admits the loss: coders down-code to dodge a denial, and nobody questions a code that's too low. AutoCode reads the same operative note and shows why it picked that code: not a black box, a citation.

01Behavioural insight over feature claim
02Real UI, fictional patient data
26 seconds / Landscape explainer

Wassist / inbound commerce

The sales associatethat waits to be invited.

Every inbound DM is a lead someone still answers by hand. The reframe: not a chatbot bolted onto checkout, but the associate that was always supposed to be there, the one who never chases and only waits to be let in.

01Assistant framing, not automation framing
02Customer-initiated, not an outbound push
24 seconds / Landscape explainer
ABM concept

Filed

The bet isn't "AI does your taxes." It's that the review process shouldn't live in two people's heads. Every figure traces to its source, and the card ends on Filed's own published number: 94% still needs professional review. Not hiding the limit. Using it as the trust signal.

20s / personalized B2B

Explainer

Structured AI

Construction QA, sold on the cost of finding the defect after the concrete has set. The software arrives as relief, not as the subject.

36s / high-stakes B2B

Where the method travels

Not a fourth market. The same angle-first process, pointed at a category that sells a decision instead of a product. It's here as the stress test: if the method only works on software, it's a template, not a method.

UniRoadmap / admissions consulting

A dream-school listis not a strategy.

For GCC parents evaluating US/UK admissions support. Most competitors argue about which schools they can get a student into. This reframes the argument before that fight starts: career direction has to come first, or the dream-school list is wishful thinking dressed as a plan.

26s / category reframe

26 seconds / Category reframe

Strategy first.
Then the pixels.

Find the useful tension

Start with a buyer problem, objection or costly habit. A feature list cannot carry a first three seconds.

Give the product a role

Translate product truth into one clear promise, then make that promise visible through script, sound and motion.

Build the next test, not just the asset

Each cut should teach the team what to make next. That is how creative compounds instead of being posted once and forgotten.

Portrait of Le Dinh Tai, known as Travis

About

Le Dinh Tai.
Travis, in most rooms.

I came to creative from the revenue side, not from film. Every job before this one was some version of the same question: why did that person buy, and why did the next one walk.

2023–24

Selling in person

B2B furniture sales in Hanoi. Objections to your face, no landing page to hide behind. This is where the instinct for what a buyer actually resists came from.

2 years

Direct-response copywriting

Self-taught, deliberately: awareness stages, market sophistication, offer structure. Long-form advertorial and VSL work, the discipline of earning the next line, then the next.

2026

Messaging & conversion lead

Retained by Scale Agency as messaging & conversion lead: VSL, landing page, email ecosystem. Behind those, the content system that fed them: five pillars mapped to funnel stage, five to six reels and two story sequences a week, winners re-tested one variable at a time.

Now

Creative strategist

The production stack changed. The questions didn't. Testing six angles now costs what polishing one used to, which only matters if you know which six are worth testing.

Working

Based in Vietnam, on your clock

Remote, with hours that overlap the US and UK working day. Async by default, on a call when a decision needs one.

Proof beyond video

The paper trail behind the judgment.

Written work and funnel architecture from the years above: the receipts for why a brief here starts with buyer logic instead of a shot list.

1MD Nutrition / long-form advertorial

Direct-response advertorial for a brain-health supplement, including headline test sets, body copy and cited research support.

Read work ↗

Scale Agency / lead-magnet VSL and funnel

Training-video script, opt-in path and the follow-up architecture behind it. The part worth reading is the no branch: what happens to a lead who does not book is designed, not left to decay.

Funnel flow diagram: YouTube, Instagram and Meta ads feeding a lead magnet, then VSL, qualification call and both the yes and no paths
Three content sources into one lead magnet. Every exit routes somewhere: newsletter, follow-up sequence, or back into the call.
Read script ↗

Scale Agency / founder-led Instagram system

A weekly operating system with defined top, middle and bottom funnel roles, plus a rule for reposting formats that already earned attention.

Instagram reel from the content system showing 61.2K views
A founder-led reel produced out of that system. The angle was the whole bet: no product, no pitch, just a question the audience was already asking.10K views in 24 hours / 61.2K within the week
View plan ↗

Bring me a product
and a real constraint.

If it needs to be understood, remembered or bought, we can find the creative angle worth putting in front of people.