Where should new creative ideas come from: customer research and reviews, competitor ads, or our own performance data?
Our answer
Find new concepts in customer research: reviews, comments, support tickets and customers' own words. Use competitor ads only to spot angles you're missing. Let your own results decide which concepts and personas get more versions.
- Confidence
- Low
- Strongest proof
- 1 Meta rule
- 1 dataset, 14 practitioners, 14 opinions or claims
- Sources
- 18 sources
- 30 statements
NTSL
MAFIAPG
New here? Take the 1-minute tour
Why this is the answer
3 reasons and our judgmentBrand-side creative leads start with customers. Harry's Daniel Rivera puts customer research first, and Everyday Dose's Travis Eubanks would start with reviews, competitors' included, and have AI analyze 200 to 300 (Motion panel). Wispr Flow's Eric Zaworski mines reviews, support tickets and ad comments.
Competitor ads are an input, not a template. Flighted, a paid media agency, reads ads running 30 days or more, 15 to 20 per competitor grouped by hook and offer, and says they don't replace your own testing.
A record of what worked keeps each brief from starting at zero. Manychat's Josh Levine calls his team's customer research the core of his campaigns and keeps a database of the messages and angles that worked.
Our judgment
We joined several sources' practice into one split: customer research for new concepts, competitor ads for gaps and your own results for more versions. Customer research comes first because the brand-side leads on Motion's panel, The Andrew Faris Podcast, Eric Zaworski, Josh Levine, Fraser Cottrell and Cody Plofker all start there. New Engen names reviews, interviews and social listening as where new creative hypotheses come from, and Meta asks for different reasons to care. We use competitor ads to find gaps, not as a template, joining Rivera's and Hubans's gap checks and Lorenzo Pravata's pre-concept breakdown with Flighted's view that the research is an input. We let your own results pick which concepts and personas get more versions, drawing on Levine's database, Zaworski's marketing OS, Eubanks's hit-rate picks, Cottrell's feedback loop and the persona, angle and hook naming from the Marketing Operators podcast.
Our judgment, built from Motion, Andrew Faris, Eric Zaworski, Josh Levine, Fraser Cottrell, Cody Plofker, New Engen, Meta, Lorenzo Pravata, Flighted and Marketing Operators podcast: our reading of these sources, not a finding from any one of them.
What the sources say
30 statements from 18 sourcesAndrew Faris says to completely ignore the ad library of brands funded very differently from yours.
“completely disregard their ad library. Like, completely. Don't even look at it. It's irrelevant to you.”
In a Meta customer story, Live Bearded used Meta's Muse to read 87,000 reviews and moved its ad wording toward customers' own words, which its marketer credits with better acquisition. No figures.
“Customers were plain and conversational, while the marketing had drifted into marketing-speak.”
Nick Theriot says reading one client's reviews revealed buyer groups with no ads; ads for one group unlocked another $100K a month in profitable spend. One client, no baseline; he sells this work.
“Read your reviews. For one client, we found groups of buyers they'd never made a single ad for.”
Alex Cooper says using AI to mine ad comments and reviews found a new angle (nose patches used on the chin), and ads built on it did really well.
“We started making ads around that. And that's completely new angle for us to hit. And the ads went on and go on to do really well.”
Nick Theriot says once creative formats are exhausted, testing different desires, awareness levels and avatars unlocks more spend. He cites 7 ads at $200,000 to $400,000 each over 3 years.
“this is where we start to go into the different desires, different awareness levels, different avatars”
Sarah Levinger says iterating on winning ads only squeezes small gains from a saturated audience, while inventing unlikely buyers can surface new angles. Her examples are hypothetical.
“When you iterate on existing winners, you're just optimizing for the local maximum. You're squeezing a 4% lift out of an audience that's already saturated.”
Nick Theriot, who sells agency services, says one brand grew from $186K to $444K in April revenue by finding new customer types. He tests one avatar a week with three new concepts. No control.
“Test one avatar per week, three fresh ad concepts each. After four weeks you'll know which one wins.”
Sarah Levinger says to analyze reviews from one star to five, not just positive ones, to map customers' emotional needs.
“It's better if you can get five star all the way down to one star, get the whole batch and throw that in there and then map it to your valence intensity zones.”
Meta says ads that are too similar may be treated as the same and shown to the same people, limiting reach and causing Creative limited or fatigue status. It suggests different angles such as price or ease of use.
“When creative is too similar, our system may not recognize it as different and could deliver it to the same sets of audiences.”
Ben Feys uses AI over the brand's stored reviews to pull a verbatim review that fits a given ad, saying it saves time.
“It can spit out a verbatim review”
Lorenzo Pravata, whose agency sells competitor breakdowns, says the Ad Library shows only which ads Meta shows most, not spend, CPA or ROAS per ad. No data.
“You don't see AG1's spend, CPA or ROAS per ad.”
Wispr Flow's Eric Zaworski says the team finds real customer questions in reviews, Reddit, support tickets and ad comments, builds ads to answer them, and shares lessons across paid, landing pages and email. No data.
“we find real questions in reviews, Reddit, social media, anecdotes, uh support tickets, even comments on Instagram from our ads”
Fraser Cottrell, from 10,000+ ads, scales static ads by mining reviews for one message, making versions look different rather than just read different, and expanding winners.
“Make it look different, not just read different”
Flighted, an agency, says competitor ad research shows which hooks survive past 30 days, and 30+ days of runtime is the strongest public proxy for performance. It advises turning it into 3 to 5 hook patterns. No data.
“compresses weeks of blind creative testing into a focused first batch of ads”
Karina Tukanova says comments on your own and competitors' ads are a rich source of hooks and user pain points.
“they are just a treasure trove of hooks”
Cody Plofker says for a new product launch, pick a top-three set of personas from intuition and research, test them, and adjust based on data.
“You got to take your top three and you're not going to get it right.”
A Marketing Operators host suggests putting persona, angle and hook in ad naming conventions so performance can be grouped and compared.
“You can set all those things up in naming conventions like persona, angle, hook.”
Phil Graham says repeated questions in ad comments are gold: answer them in the comments and make a new ad that answers them.
“if you're seeing repeated questions that are similar to each other, that is gold”
Matthew Gattozzi says review mining, including with AI, often misses new customer segments because reviews come from people you already market to.
“you're going to get reviews from the people you've already been marketing to”
New Engen, an agency, says untested needs found through customer interviews, reviews, social listening and performance data can each become a new creative hypothesis. No outcome data.
“Each can become a new creative hypothesis.”
Elise Zari advises new app advertisers to study the creative of top-grossing apps in their category and adapt their winning formats to start.
“I would look at what competitors do creatively in your vertical... and try to lift and shift their winning formats to your app to get started.”
Andrew Faris says to build Meta creative from reviews, comments and support tickets, mining customers' fears, failures and fantasies, and to write ads in their exact words. No data.
“write ads in customers' exact words”
Alex Cooper says to use customers' exact words in ad copy rather than paraphrasing, since that is how other customers speak.
“You don't want to paraphrase what people are saying because that is how your customers speak... your customers are your best copywriters.”
Motion's panel says a small team's priority is customer research: read your reviews and competitors' reviews, be active in comments, and compare your messaging with top competitors. Practice only, no outcome data.
“customer research, customer research, customer research”
Phil Graham says to study competitors in the Ad Library to find what they all do and then do the opposite, so your ads stand out instead of copying them.
“I look and see what they're doing so I can do the opposite so I can differentiate so I can stand out instead of copying them.”
Josh Levine of Manychat says product marketing researches customer pain points, and creative data goes into a database of working messages and angles so briefs do not start from zero. No outcome data.
“we're not starting from zero”
Harry Delmege argues manual research into customers' pain points and language beats AI-generated ad copy and research.
Sarah Levinger says to find concepts in customer research, comparing 5, 3 and 1 star reviews and mining Reddit, forums and comments, rather than in trends or competitor scans. No data.
“the differences between five star, three star and one star reviews is really interesting”
Adplexity's Gabe Ansel says his team studies competitors' ads every day to see what works on Meta and find ideas to test.
“ad spying is really good. We do it every day. It's like a mandatory on our team to like do some ad spying, figure out what's working in the space”
A host of an AI-focused ecommerce podcast says a modern creative strategy starts with a gap analysis comparing your ads, personas and angles with competitors in your vertical.
“the first thing is like a gap analysis and saying like you compare and juxtapose yourself to other folks in your vertical”
In the full answer
For readers who sign in- What to do this week3 steps
- The numbers behind it2 figures from cited sources
- Where each source standsA chart placing each source on the question
- The strongest disagreementThe best case against the answer, from a cited source
- How sure we areWhat its confidence rests on, and what would change the answer