| name | creative-research |
| description | Run structured creative research across three realities — customer (generational context, micro-moments, emotional mapping), brand (production capacity, cost, and what the business is actually optimizing for), and presentation (which format the brain in each journey phase will accept). Use whenever the user is researching customers or audiences for ads, building personas or avatars, hunting for ad angles or insights, mining reviews and Reddit for customer language, asking what format or creative mix to run, planning a creative brief, or feeling buried in research without knowing what to do with it. Also trigger on TEEP, trigger/exploration/evaluation/purchase, micro-moments, emotional mapping, valence and intensity, generational marketing, brand alignment, creative capacity, production ceiling, 'what should our ads say', 'how do I research my customer', 'I have tons of research and no idea what to build', or 'we don't agree on what we're optimizing for.' |
Creative Research
Sarah Levinger's research system. Three realities, researched in order, on a fixed cadence.
Adjacent skills: creative-diagnosis decides what to work on. performance-creative runs the test loop. static-ads builds the artifact. This one feeds all three.
The rule that gates everything
Before opening any research tool, ask:
> What decision will this research help me make?
If you can't answer, close the browser. You aren't ready to research.
Every piece of research should change something: what you build, how you frame it, who you target, why you're doing it at all. If it won't change those, you don't need it yet.
Research without a decision to serve is busy work — it feels productive for two hours and leads nowhere. Having a lot of information is not the same as knowing what to do with it. Most strategists can tell you exactly what every competitor is running and cannot tell you why any of it works or what to make next.
The failure isn't talent or headcount. It's a systems problem.
Fun facts vs. insights
The distinction that governs all customer research.
A fun fact is a statement. "She wants to lose weight." Every competitor already knows it.
An insight is a situation. She's getting dressed for lunch with a friend who just lost 40 pounds. Her favorite jeans don't fit the way they used to. She's anxious before she even leaves the house.
The insight tells you which creator to source, what tone to use, what emotional truth she needs to hear, and what brief to write. The fun fact tells you nothing.
Stop researching for fun facts.
The three realities
Research them in this order. Each answers a question the others can't.
| Reality | Question | Failure if skipped |
|---|---|---|
| Customer | How does this person actually experience the problem? | Perfect ads talking to the wrong person at the wrong moment |
| Brand | What does the business actually need from creative right now? | Ads everyone approved that get killed anyway |
| Presentation | Which format will this brain, in this state, accept? | Great message wrapped in a format the brain rejects |
You can have emotionally precise customer research and still build the wrong creative, because customer truth doesn't tell you what the business needs done with it.
Full detail: references/[customer-reality.md](http://customer-reality.md), references/[brand-reality.md](http://brand-reality.md), references/[presentation-reality.md](http://presentation-reality.md).
Reality 1 — Customer
Three layers, bottom to top. Order matters.
Generational insights — the world that shaped this person before your product existed. Four questions: their childhood informational environment, what shaped their relationship with money, what their actual (not aspirational) life is right now, and how they evaluate risk and decide whom to trust.
The brain is a pattern-matcher. It maps everything new onto everything it has already experienced. Write to who someone is today and you're ignoring the map their brain is actually using.
Critically: you are not looking for angles here. You're building a library in your own head — becoming a registered expert on the chosen human. Absorb. Later, when you're drafting or prompting, you'll be able to say this person wouldn't love that headline; they'd love this one.
Micro-moments — a good strategist finds pain points. A great strategist finds out when those pain points show up, and why. The exact moment in the year, the week, the day. Real-world situations, not marketing situations. Hyper-specificity gives you somewhere to put the camera, a creator to pull, a story to tell.
Emotional mapping — for each micro-moment, three questions:
- Which emotion is present, most strongly?
- Is it private or social?
- Is it acute (sudden) or chronic (years)?
These change the message, not just the angle. Hope + private + chronic makes "join others just like you" too hype-y; "you don't have to suffer alone anymore" lands. Nearly identical messages; entirely different customers.
The Valence + Intensity model (positive/negative × how intense) organizes which emotion to target and how. Credit belongs to psychologist James A. Russell (1980); Levinger applied it to ads.
Reality 2 — Brand
Open with a question almost nobody asks:
> What is this brand actually optimizing for right now — and does everyone agree?
Usually they don't. The media buyer optimizes for scale, the founder for margin, the creative team for brand equity. Every ad you ship tries to serve all three, satisfies none, and the blame lands on the creative. The ads never had a chance.
Part A — Health. Cost is not media spend. It's development time, team labor, attention, bandwidth, operational load.
The scenario that recurs: a plan needs 50 ads a month; the team can make 20. You launch anyway. The plan runs far below what it needs, the data gets muddy because there isn't enough creative to test properly, and everyone concludes the strategy failed.
> The strategy did not fail. It was never tried.
Nobody asked whether it could be built. Know the ceiling before you build toward it. And ask: what happens if it works? Great creative plus broken fulfillment is a pile of refund requests — you paid to create angry customers.
Part B — Alignment. At the start of every month, ask every person who touches an ad:
> If you could only optimize for one thing right now, what would it be?
Don't summarize their answer. Write it down. Then compare. You are looking for disagreement — disagreement isn't a problem, it's information. This is what happens when alignment is assumed instead of established.
Homework, always: one sentence the team agrees on and commits to for a defined window. For the next 60 days we are optimizing for X.
Reality 3 — Presentation
> You're not making ads for an audience. You're making ads for a specific type of brain.
That brain is inside an ecosystem purpose-built for mindless scrolling. It does not want to leave. It is conserving energy. Choosing your creative stack based on what's trending, what feels fresh, or what the founder likes ignores the only variable that matters: the psychological state of the person processing the ad.
The TEEP framework — Trigger, Exploration, Evaluation, Purchase. Drawn from Google research on digital purchasing. People morph in and out of exploration and evaluation for a long time before purchase.
Most brands run creative toward roughly two of the four phases. Find out which two you're missing.
| Phase | State | Your job | Formats |
|---|---|---|---|
| Trigger | Pre-aware. Doesn't know a solution exists. | Name the moment. Make them feel seen, not sold to. | Micro-moment content, "this is so me" |
| Exploration | Problem-aware, low commitment, high novelty appetite | Earn the right to say more. Talk about the industry. | Short-form video, visual hooks, novelty, memes |
| Evaluation | Comparing options. Skeptical, wants evidence. | Talk about us. Prove it. | Testimonials, before/after, comparison tables, case studies |
| Purchase | Credit card in hand | Let them buy. | Offers, free gift, long-form, founder ads |
The purchase-phase failure is real and expensive: someone is ready, and you keep educating them, and they walk away.
Optimization goal determines format weighting:
- Top of funnel → weight Trigger. Consider link-click/traffic optimization rather than conversion. Wide net, pattern interrupt, novelty.
- Sales growth → weight Evaluation and Purchase. Depth, proof, specificity, trust. Not a place for short hooks.
- Cost reduction → whatever your team can make at extremely high quality. Quality is a cost reducer. Relatable, well-told content draws customers who identify with the brand, and that's a cheaper sale.
The cadence
A system you don't run is a nice afternoon.
| When | What | Output |
|---|---|---|
| First Monday of the month | Customer research (all three layers) | Updated customer profiles |
| First Monday of the quarter, and first of the month | Brand research | Brand optimization statement + tracker |
| Every Monday, 30–60 min | Presentation research (mostly organic scrolling) | Updated TEEP inspiration library |
| Then | → the strategic creative brief | |
Tag every idea you source by TEEP phase. This builds an ad library that is browsable — by customer reality and by brand reality. When a concept dies, you pull the right format for the right customer at the right optimization level immediately, instead of starting a search.
If the team is small
Prioritize in this order, and say so plainly:
- Micro-moments and emotions (customer)
- Presentation (formats)
- Brand — a small team usually already knows what the business is doing
Skip the generational layer. Levinger loves it and says to skip it first. Take her at her word.
Positions to hold
Avatars are a crutch. "There is no avatar that will help you scale to the moon. What actually helps you scale is having a distinct point of view for the brand" — one you can apply to any avatar. Build 25 personas if you like, but if they don't help you communicate the point of view, it's busy work. Her own avatars are increasingly emotional rather than demographic: people who believe everyone should be included and who love skateboarding.
> Note the live disagreement. Denney treats persona as structural, because she reads Andromeda as a persona-based algorithm that ranks and retrieves by persona. Levinger reads Andromeda as interest-based — "not necessarily just an angle or customer-based" — and demands hyper-specific situations rather than richer avatars. They converge on the prescription (get hyper-specific, map creative deliberately) and disagree on the mechanism. When a user asks, present the fork rather than resolving it.
Volume means consistency, not a number. "At quality" is the metric. Aim for 25 ads at a QA standard; if you hit 7, seven is your benchmark. Then upgrade the system, not the target. Nobody knows where "500" came from.
Top of funnel belongs on organic where you can manage it — cheaper, and it directly improves your paid performance. She'll die on this hill.
Longevity is a signal, not proof. 30+ days running is worth noting, but check the story. Some brands park spend on a hundred ads specifically to confuse competitors. Ask whether the ad is actually well-produced, well-told, well-cast. If not, ignore the runtime.
Click-through rate lies. A click is not a customer. High CTR with low conversion means the message is pulling the wrong brain.
On AI. She used to have an LLM run large-scale research and stopped, for a specific reason: it mirrors what you love. Prompt it and it returns more of your own bias — "Sarah loves psychology, here's more psychology" — while silently dropping everything else you asked for. Now she uses agents to scrape real conversations by keyword across platforms, compares what's said now against 18 months, 5 years, 10 years ago, then hands it over with "what do you see in here?" rather than letting it assume. Still uses AI heavily for condensing.
It transfers. B2B, SaaS, services, nonprofits, events — the research is nearly identical, because you're always working with humans. Only the output changes.
How to run this with someone
Ask the gate question first. What decision will this research help you make? If they can't answer, that's the work. Don't start scraping Reddit.
Then locate them:
- Drowning in research, no direction → the gate question, then the three realities in order
- Has customer insight, ads keep getting killed → they skipped brand reality
- Great message, no traction → presentation reality; check TEEP coverage
- "What format should we run?" → what are you optimizing for? Then weight accordingly
- Building personas → check whether the personas serve a point of view, or are decoration
Push for the situation, not the statement. When a user offers "our customer wants X," ask when. Where. What just happened. Who else was in the room. Keep pushing until it's a scene.
Reference files
references/[customer-reality.md](http://customer-reality.md) — the three layers, four generational questions, micro-moment hunting, emotional mapping, and where to find eachreferences/[brand-reality.md](http://brand-reality.md) — the real cost of an ad, capacity ceilings, the optimization question, alignment statement, and sourcesreferences/[presentation-reality.md](http://presentation-reality.md) — TEEP in depth, format selection by phase and by goal, the tagging system, and research sources
