Your next winning ad is already in your data.
A practical guide for D2C performance teams on turning six data sources into a steady pipeline of creative ideas. Every idea comes with a reason to exist, a hypothesis, and a way to measure it.
Creative is your biggest lever. Most teams still pick it by gut.
On Meta today, targeting is mostly automated. Advantage+ audiences and broad targeting mean the algorithm decides who sees your ad. What you still control is what the ad says and how it says it. The creative has become the targeting.
Yet most creative calendars are built from a brainstorm, a trend someone saw on Reels, or what the agency made last month. When a batch fails, nobody can say why, so the next batch is another guess.
The fix is to treat every new creative as a hypothesis drawn from a signal. Some signals tell you what already works and should be pushed further. Some show you gaps you have never tested. Some come straight from what customers say and search for. Combine them and you get a backlog that is deep, varied, and explainable.
What already works
Mine winning creatives for repeatable patterns, then map the combinations you have never tried.
Sources 01, 02Why people bought
Buyers tell you the trigger, the doubt they overcame, and the words they use. That is ready-made copy.
Source 03What people want
Competitor ads, Amazon queries and Google searches reveal demand and messaging you are not addressing.
Sources 04, 05, 06The creative loop
Every source in this guide feeds the same five-step loop. Run it every week or two and the learning compounds.
Collect
Pull signals from all six sources into one place.
Find patterns
Group signals into themes: a hook, an angle, a customer need.
Write briefs
One brief per hypothesis, each linked to the signal behind it.
Test
Launch in a controlled test with comparable budget and audience.
Read results
Judge on true performance, tag the result, and update the matrix.
โป Results from step 5 become the input for source 01 in the next cycle.
Iterate on creative patterns that are already winning
The fastest win is usually not a new idea. It is a better version of an idea you already know works.
A winning ad is rarely winning for one reason. It is a combination of a hook, an angle, a format, a person on screen and a product. If you can name those parts, you can keep the ones doing the work and change one at a time. That is how a single winner becomes ten.
- Top creatives by true ROAS and CM2, not platform ROAS. Platform numbers over-credit retargeting and under-credit prospecting.
- Hook rate (3-second views รท impressions) and hold rate (ThruPlays รท 3-second views) to separate a strong opening from a strong body.
- Creative tags: hook type, angle, format, length, talent (founder, UGC creator, influencer), product shown.
- Fatigue curves: how fast CTR and CPM decline as frequency climbs.
- New Reach share, to check whether a winner is finding new people or recycling warm audiences.
- Rank by a fair metric. Prospecting ads on 30-day first-touch ROAS, retargeting on 3-day last-touch. Mixing them crowns the wrong winners.
- Group winners by tag, not by ad. Ask which hook types and angles appear again and again across winners.
- Diagnose the strength. High hook rate, low hold: keep the opening, rewrite the body. Low hook rate, high CVR: the message sells but the opening loses people.
- Change one variable per variant. New first 3 seconds, new creator, new format, or new offer. Never all four.
- Refresh before fatigue. When a winner's CTR drops sharply as frequency rises, the next iteration should already be live.
| Variation | What changes | What stays | Use when |
|---|---|---|---|
| Hook swap | First 3 seconds: text overlay, opening line, first visual | Body, CTA, creator | Hold rate is good, hook rate is falling |
| Talent swap | A different creator delivering the same script | Script, structure | The winner is fatiguing at high frequency |
| Format swap | Video to static, carousel, or a 9:16 cut | Message, angle | You need coverage across placements |
| Audience-of-one rewrite | Same angle, aimed at a different persona | Product, angle | Survey shows a second buyer group (see 03) |
| Angle shift | New reason to buy, same hook style | Hook type, format | The matrix shows an open neighbouring cell (see 02) |
Fill the hook ร angle cells you have never tested
Your account has blind spots. A simple grid makes them visible.
A hook is how an ad grabs attention in the first seconds: a question, a bold claim, a demo, a founder speaking. An angle is the reason to buy it argues for: price, results, ingredients, convenience, identity. Any hook can carry any angle. Put hooks on one axis and angles on the other and every cell is a creative concept.
Tag each live and past ad into its cell, then colour the cell by result. Most brands find they have tested a handful of cells heavily and left most of the grid empty. Click any cell below to see how to read it.
Expert POV ร Price / value
Never testedHigh-priority gap: the expert pov hook already wins with another angle. Test 3 ads here with a fixed budget. Check survey answers and search terms for language to use.
- Fix your two lists. 6 to 8 hooks and 5 to 7 angles is enough. Keep the labels stable so tagging stays consistent over months.
- Tag every ad. Creative Intelligence in BooleanMaths tags hook, angle and format automatically. Review a sample by hand the first time.
- Colour by true performance, using the same funnel-stage metric for every cell.
- Set a minimum test size (for example 3 ads and a fixed spend per cell) before calling a cell a loser.
- Neighbours of winners. An open cell in the same row as a winning hook, or the same column as a winning angle, has the best odds.
- Angles backed by other signals. If survey answers and search terms keep mentioning an angle, that column deserves tests.
- Cells competitors own. If rivals run one combination heavily and you have nothing there, that is a gap (see 04).
- Skip implausible cells. Some combinations make no sense for your product. Mark them closed so they stop showing up as gaps.
Write ads in your customers' own words, from post-purchase surveys
Right after checkout is the one moment a buyer will happily tell you why they bought.
Most brands use the post-purchase survey only to ask "where did you hear about us?". That is useful for attribution, but the open-ended answers are where the creative gold sits. They show the trigger moment, the doubt the buyer had to get past, and the exact phrases real people use. Phrases from customers almost always beat phrases written in a brainstorm.
- "What made you finally decide to buy today?" Gives you the trigger and the urgency angle.
- "What almost stopped you from buying?" Gives you objections to answer head-on in an ad.
- "What were you using before this?" Gives you comparison and switching angles.
- "Who is this for?" Reveals gifting, family and second-persona buyers.
- "Where did you first hear about us?" Shows which creators and channels actually sparked the purchase.
- Export open-text answers for the last 60 to 90 days, with order value and first-purchase vs repeat marked.
- Cluster them into themes. An LLM can do the first pass in minutes. Ask it to group by trigger, objection, outcome and persona, and to count each theme.
- Rank themes by frequency and value. A theme that is common among high-AOV first-time buyers is worth more than one common among discount repeat buyers.
- Lift the exact phrasing into hooks, on-screen text and UGC scripts. Do not polish it into brand language.
- Turn every frequent objection into an ad that names it and answers it in the first 5 seconds.
Study competitor creatives for hooks you have not used
Competitors are running tests with their own money. Their longest-running ads are their results.
The Meta Ad Library shows every active ad a brand runs, along with when it started. An ad that has been live for 60 days or more is almost always profitable, because nobody keeps paying for a loser that long. That makes ad longevity the best public proxy for a competitor's winners.
- 5 to 10 competitors: direct rivals, adjacent categories, and one or two global brands that set the creative bar in your space.
- Long-running ads (live 30+ days) and ads with many near-identical variants. Heavy variation means the brand is scaling that concept.
- Hook, angle and format for each ad, using the same labels as your own matrix.
- Offer and landing page: bundle, free gift, subscription, quiz.
- Comment sections on their organic posts, for objections their customers raise.
- Tag competitor ads into your matrix as an overlay. Cells they fill heavily and you leave empty are your first targets.
- Borrow the structure, not the content. Take the format and pacing that works, then fill it with your proof, your product and your customers' words.
- Look for angles nobody owns. If every competitor shouts price, a results or transparency angle stands out by default.
- Re-check monthly. When a long-running competitor ad disappears, it has probably fatigued. That tells you how long the concept lasts.
| Competitor signal | What it likely means | Your move |
|---|---|---|
| Same ad live for 90+ days | A proven, profitable concept | Test the hook ร angle cell it occupies with your own proof |
| 10+ variants of one concept | They are scaling it and fighting fatigue | Expect the concept to saturate. Enter with a fresh format |
| New angle across several brands | A category-wide message shift | Test fast before it becomes noise |
| An angle nobody uses | An open lane, or a concept that failed quietly | Run a small test and check survey data for support |
Mine Amazon searches for buying language
Amazon search is closer to the wallet than anywhere else online. People type what they are ready to buy.
Amazon search terms tell you the attributes that matter at the moment of purchase: size, ingredient, use case, who it is for, price band. If you sell on Amazon, your Sponsored Products search term report and Brand Analytics are a direct read on demand. Even if you do not, Amazon's own search suggestions and top reviews are public.
- Search term report: queries that converted for your ads, with orders and ACOS. Connect Amazon to BooleanMaths to see these next to your D2C data.
- Brand Analytics top search terms in your category, with click and conversion share.
- Modifier words people add to the core term: "for men", "without fragrance", "travel size", "under 500".
- 1 to 3 star reviews on top competitor listings. Complaints are angles you can win on.
- "Customers also bought" pairings, which suggest bundles and routines.
- Group search terms by modifier. Each modifier group (persona, ingredient, use case, price) is a possible angle.
- Weight by conversions, not volume. A term with modest volume and high conversion share points to strong intent.
- Put the winning modifier in the hook. If "for men" converts, run an ad that opens with a man using the product.
- Answer competitor complaints. If rival reviews say "sticky", make "non-sticky in 30 seconds" a demo hook.
| Search term (example) | Modifier | Angle it suggests | Hook idea |
|---|---|---|---|
| niacinamide serum for oily skin men | Persona | Identity: men's skincare | "Guys, your oily T-zone is not a personality trait." |
| vitamin c serum without fragrance | Ingredient exclusion | Transparency, sensitive skin | Label close-up: "Nothing on this list you can't pronounce." |
| sunscreen no white cast | Outcome | Results, demo | Split screen on darker skin tones, 10-second rub-in |
| face wash combo pack | Format | Value, routine | "The 3-step routine that costs less than one cafรฉ visit a week." |
Read customer intent from Google searches
Google search shows the questions people ask before they are ready to buy. Every question is a hook.
Where Amazon shows the language of buying, Google shows the language of problems and research. Someone searching "why is my skin oily even after washing" has not picked a product. That is exactly the person a prospecting ad needs to stop. Map queries to the stage of the journey and you know which ad to make for whom.
- Google Ads search terms report: actual queries that triggered your Search and Shopping ads, with conversions.
- Search Console: queries that bring organic visitors to your site and blog.
- Google Trends: seasonality and rising queries, filtered to India or by state.
- "People also ask" and autocomplete: the follow-up questions around a topic.
- Keyword Planner: volume, to size which intents are worth a creative.
- Collect 100 to 300 queries in your category and label each with a funnel stage.
- Pull out the questions. Queries starting with why, how, does, can, is and vs are ready-made hooks for problem-aware audiences.
- Use rising queries for timing. If "sunscreen for monsoon" climbs every June, brief that creative in May.
- Watch brand-plus queries. Rising "[brand] side effects" or "[brand] vs X" tells you which doubts or comparisons to address in retargeting.
- Use state-level trends to localise: language, climate and festival timing differ a lot across Indian states.
More signals worth mining
The same method works for any source where customers reveal what they want, fear, or say. Add these once the core six are running.
Product reviews
Your own 5-star reviews give you proof lines. 3-star reviews show what nearly disappointed people.
Support and WhatsApp chats
Pre-purchase questions are objections, asked in plain language. Repeated ones deserve an ad.
RTO and return reasons
If "not as expected" drives returns, an honest demo ad may cut RTO and wasted spend together.
SKU and bundle sales
Products rising without ad support show organic demand. Give them a creative before a competitor does.
Quick-commerce search
What sells on Blinkit, Zepto and Instamart shows impulse and replenishment angles by city.
Organic comments and DMs
Comments on your top Reels show what made people stop, and what they argued about.
Reddit and Quora threads
Long, honest discussions of problems in your category. Good for myth-busting hooks.
Geo performance
States or cities that outperform suggest regional language, climate or festival creatives.
Cohort and LTV data
Which first product leads to the best repeat customers. Lead your prospecting creative with it.
Prioritise the backlog
Six sources will give you more ideas than you can produce. Score each one so the team works on the best bets first.
| Factor | Score 1 | Score 3 | Score 5 |
|---|---|---|---|
| Evidence | One source, weak signal | One strong source | Confirmed by two or more sources |
| Size | A niche persona or rare query | A clear segment of buyers | Core buyer, high volume or high AOV |
| Novelty | Close copy of a live ad | Variation on a known cell | An untested cell or new angle |
| Effort | Static or text swap | New edit of existing footage | New shoot or new creator |
Ideas that show up in more than one source are your strongest bets. If survey buyers mention purging, Amazon reviews complain about breakouts, and Google shows rising "does X cause purging" searches, that one concept has three independent votes. Brief it first.
A healthy weekly mix for a brand spending โน10 to 50 lakh a month on Meta is roughly 60% iterations on winners (source 01), 30% new cells and new angles (02 to 06), and 10% wild cards nobody is sure about.
The brief template
Every creative brief should trace back to a signal. If you cannot fill the first two rows, the idea is not ready.
| Field | What to write | Example |
|---|---|---|
| Source signal | The data point and where it came from | Survey: 1 in 5 objections mention purging |
| Hypothesis | "We believe X will do Y because Z" | Naming the fear upfront will lift CVR among fence-sitters |
| Matrix cell | Hook type ร angle | UGC testimonial ร Fear / risk |
| Audience stage | Prospecting, middle of funnel, or retargeting | Middle of funnel |
| Opening 3 seconds | Exact line, on-screen text, first visual | "I was terrified this would make me purge." |
| Proof | What makes the claim believable | Day 1 to day 14 diary, no filter |
| Format and length | Video, static, carousel; duration; ratio | 9:16 video, 25 to 35 seconds |
| Success metric | The metric and attribution window you will judge on | 14-day linear ROAS, CVR |
| Kill or scale rule | Decided before launch, not after | Below account median after โน30k spend: kill |
Judge results fairly
A good creative pipeline falls apart if the scoring is wrong. Three rules keep it honest.
Use the right window for the stage
Prospecting creative on 30-day first-touch ROAS. Middle of funnel on 14-day linear. Retargeting on 3-day last-touch. A prospecting ad judged on last-touch will almost always look like a loser.
Do not trust platform ROAS
Meta and Google each claim credit for the same orders. Judge creatives on measured performance and contribution margin (CM2), so a winner is a winner in your P&L, not just in Ads Manager.
Write the result back
Tag every test as won, average or lost in its matrix cell, with the reason you think it happened. Next cycle, source 01 starts from a richer base.
Run this playbook on your own data.
BooleanMaths brings creative tags, true ROAS, survey answers, Amazon and Google data into one place, so you can go from signal to brief in an afternoon.