Customer sharing purchase feedback in an interview

Customer Research for Ads: A Meta Ads Playbook for DTC

October 30, 2026

Customer research for ads is the work you do before anyone designs a creative: finding out who buys, what they were trying to fix, what almost stopped them, and the exact words they use to describe it. Then you turn those findings into hooks, angles and concepts you can test on Meta. I’m Stefano Mazzei, a performance marketer and Co-Founder of Cosma, and this is the process my team and I use with e-commerce and DTC brands before we spend a dollar on new creative.

Stefano Mazzei, Performance Marketing Expert

Turn Customer Insight Into Better Ads

Stefano Mazzei shares practical Meta Ads frameworks and customer research insights drawn from his work with e-commerce and DTC brands.

Explore Meta Ads insights

TL;DR

Customer research decides who you target and what you say. Ad testing checks the creative you built from it. Do the first before the second.
Research the people who already bought, the people who almost bought, and the people who bought once and left. Each group answers a different question.
Mine five sources: post-purchase surveys, customer interviews, reviews, support tickets and comments. Code what you find into themes, then keep the customer’s own words.
Turn each theme into an angle, then test angles before you test hooks, formats or edits.
Read results against the research. A losing test teaches you which belief was wrong.

What customer research for ads actually is
Why it matters more on Meta right now
Which customers to research
What to ask
Where to find customer language
Analyzing what you collect
Turning insights into hooks, angles and concepts
Testing research-informed ads
My perspective: research is the cheapest media you will ever buy
Where my team at Cosma can help
Common mistakes to avoid
FAQ
Sources

What customer research for ads actually is

Market research for advertising studies the audience, the competitors and the channel options before ad spend, so you know who to target and what to say. Ad testing comes later. It takes a draft or finished creative and measures how real people respond to it. QuestionPro frames the split well: research is the foundation for audience, message and placement, and testing is the quality check on the creative built from that foundation.

The timing matters because of money. Concept testing gives you evidence before the media commitment, while post-launch data only shows what happened after you paid for it. Customer research sits one step earlier still. It tells you what concepts are worth drafting at all.

For a DTC brand on Meta, that distinction is practical. Most accounts I audit have plenty of creative and very little reasoning behind it. Ads get made from a brand deck, a competitor’s feed or a founder’s hunch. Then the team wonders why a fifth round of “new hooks” performs like the first four. The ads were different on the surface and identical underneath, because they all rested on the same unexamined assumption about the buyer.

What customer research for ads actually is — overview diagram

Why it matters more on Meta right now

Meta’s targeting has narrowed over the last few years, and the creative has taken over most of the work of finding the right buyer. Research from economists working with Meta data, hosted by the FTC, estimated a median cost per incremental customer of $43.88 with offsite data, rising to $60.19 once the median loss in effectiveness was applied, a 37% increase. The same study reports that small advertisers and advertisers in CPG, retail and e-commerce were especially affected.

Study compares incremental customer costs before and after data loss

That is the environment most of our readers operate in. When the algorithm has less signal to find buyers for you, the ad itself has to do the qualifying. A message that speaks to a specific person with a specific problem lets Meta find more people like them. A generic message gives it nothing to work with.

The same logic is behind how we structure accounts, which I cover in the Meta Ads campaign structure guide. Good structure gives creative room to learn. Good research gives that creative something worth learning from.

Which customers to research

The common mistake is surveying “customers” as one blob. Your buyers are several groups, and each one explains a different part of your acquisition problem.

Recent first-time buyers

These are your best source for why people chose you. They bought in the last 30 to 60 days, so the decision is fresh. They can describe what they searched for, what else they considered and what finally tipped them. This is the group I interview first.

Repeat and high-value customers

People who reorder or buy multiple products show you what the product really delivers over time. Their language often differs from first-time buyers. They talk about outcomes and habits, not hesitations. Their stories make strong proof, and they point to the segments worth scaling toward.

Lapsed buyers and one-and-done customers

These people bought once and never returned. They tell you where the promise in the ad exceeded the experience in the box. That gap is expensive: if your ads attract buyers who churn, your acquisition cost looks fine on day one and bad on day 90.

Non-buyers who showed intent

Add-to-cart abandoners, repeat site visitors and people who clicked an ad and left. They are harder to reach, but a short exit prompt or an email to abandoned carts can surface the objections that never appear in reviews: price, shipping, sizing, trust, “I’m not sure it works for me”.

How many people is enough?

You do not need a statistically perfect sample for interviews. One qualitative study of active review users used semi-structured interviews and reported theoretical saturation after 19 interviews, meaning new interviews stopped surfacing new themes. That was a different topic, so treat it as a rough marker and not a rule. For most brands, 10 to 15 well-chosen interviews across the groups above will show the main patterns, and a survey adds the volume.

What to ask

Good questions are about the past, not the future. People are poor at predicting what would persuade them and good at describing what already did. Here is the set we lean on.

For recent buyers (survey or interview):

What were you trying to solve or achieve when you started looking?
What had you tried before this, and what went wrong with it?
How did you first hear about us?
What else did you consider, and why did you choose us instead?
What nearly stopped you from buying?
How would you describe this product to a friend, in your own words?
What result did you expect, and what did you actually get?

For repeat buyers:

What made you come back?
What has changed for you since you started using it?
Who else have you recommended it to, and what did you say?

For lapsed buyers:

What did you expect that did not happen?
What are you using instead?

For non-buyers:

What is the one thing that kept you from buying today?

A few rules keep the answers honest. Use open text, not multiple choice, for the questions that matter. Ask “how did you hear about us” in a post-purchase survey, but do not treat it as proof of attribution. It is a directional check next to platform data. And never ask a leading question like “Did you love our fast shipping?” You will get a polite yes and no insight.

Where to find customer language

Interviews are the richest source, but they are slow. Combine them with data you already have.

Post-purchase surveys. One or two open-ended questions on the thank-you page produce a steady stream of verbatims at almost no cost.
Customer interviews. Twenty to thirty minutes each, recorded and transcribed with permission. Look for stories, not opinions.
Your own reviews and competitor reviews. Reviews show the problem before, the result after and the objections in between. Read the three-star reviews first. They are the most specific.
Support tickets and live chat logs. Every repeated question is an objection your ads and landing pages have not answered.
Ad comments and DMs. Comments under your ads are unfiltered reactions from the cold audience, including doubts, jokes and questions.

Do not skip the question of whether the reviews themselves are believable to shoppers. A 2021 Journal of Business Research study based on 27 in-depth interviews found that consumers judge the trustworthiness of online word of mouth through networked, multidimensional judgments shaped by competing persuasion agents and personal experience. In practice, that means a quote from a review lands better in an ad when it sounds like a real person with a specific situation. For the wider effect of reviews on how shoppers find and trust a store, this overview of online reviews and ecommerce SEO is worth reading.

It also helps to remember that shoppers do not rely on one place. A 2020 Journal of Marketing study that combined clickstream and survey data across 20 product categories and 40 retailers found that information-channel use differs by product purpose and purchase stage, spanning social media, third-party reviews, search engines, deal sites and retailer pages. So ask buyers where they checked you out, not only where they first saw you. A customer who saw a Meta ad, then read Reddit and Trustpilot, then bought through a branded search tells you the ad had to survive a trust check. Your creative can anticipate that check.

Analyzing what you collect

Raw answers are not insight. The process is simple and a bit tedious, which is why most brands skip it.

Set the objective first. Write down the question: “Why do first-time buyers choose us over the alternative?” or “Why do buyers churn after one order?” Separate what your campaign data already answers from what needs new research.
Clean the data. Remove junk responses, duplicates and one-word answers that say nothing.
Code the responses. Thematic analysis, as Dovetail describes it, means reading through transcripts, tagging meaningful excerpts, grouping tags into themes, refining them and reporting each theme with supporting examples and direct customer language. A spreadsheet works fine. One column for the verbatim, one for the tag.
Count and rank. How many people mentioned each theme? Which themes came from your highest-value customers?
Validate against the original objective. Analyze against the question you set at the start, not against whatever happens to be interesting.

You should end up with a one-page summary: the top five problems buyers were solving, the top five objections, the most vivid phrases they used, and the three or four segments that look most valuable.

A worked example

Say a skincare brand collects 120 post-purchase answers to “What were you trying to solve?” After coding, the tally reads like this:

Theme	Mentions	Sample verbatim
Sensitive skin reacts to most products	38	“Everything else made my face red.”
Wants a shorter routine	29	“I don’t have time for ten steps.”
Previous product stopped working	22	“It worked for a month and then nothing.”
Gift or recommendation	17	“My sister wouldn’t stop talking about it.”
Other	14	n/a

This is an illustration of the method, not data from a real brand. But the reading is typical: the brand’s current ads may lead with ingredients, while the largest group of buyers is really buying relief from irritation. That is an angle, and it is not the one the brand deck would have picked.

Turning insights into hooks, angles and concepts

The step most teams blur is the one between a finding and an ad. I keep three layers separate.

Angle: the core reason someone should care. “Finally a product that does not irritate sensitive skin.”
Hook: the first line or first three seconds that opens that angle. A customer quote, a before-and-after situation or a blunt question.
Format: how it appears. Static, short video, UGC, founder talking to camera, carousel.

Each research theme becomes an angle. Each angle can carry several hooks in several formats. This is why research changes your hit rate: you stop swapping hooks over the same angle and start testing genuinely different reasons to buy. I go deeper on the structure in the post on Facebook ad angles for ecommerce, where we test angles in a 3×2 grid before changing hooks.

Practical translation rules:

Use the customer’s words, not yours. If buyers say “my face stopped stinging,” do not write “reduces sensitivity.”
Lead with the problem or the objection the biggest theme names, not the product feature.
Turn objections into proof. If “does it really work for me” shows up in 20 answers, build a concept around a skeptical buyer’s before-and-after.
Match the format to the stage of awareness. A cold audience who does not know they have a problem needs a story. A shopper comparing options needs a direct comparison or proof.

Some formats carry research better than others. The breakdown in four creative-first DTC Facebook ad formats that scale is a good place to match insight to execution.

Testing research-informed ads

Research gives you hypotheses. Meta gives you the cheapest large-scale test environment you will find. Academic work on this by Braun and de Langhe covers four primary methods for collecting field data through digital ad platforms, and notes that any study should define an objective, budget, audience and creative variations up front. They also explain that a multiple-ad study without a holdout splits consumers into groups eligible for different ads, which corresponds to what the industry calls A/B testing.

That matches how we run tests. One research theme per ad set, the same audience, comparable creative quality, and a budget that gets enough conversion events to read the result. Our 50 event rule for Facebook ad testing explains how we judge when a test has enough data, and the 14-day Meta campaign optimization playbook shows how we handle the first two weeks of a new launch.

Testing also does not have to wait for launch. Qualtrics recommends testing before, during and after launch: pre-testing can help avoid costly pitfalls, and in-flight monitoring tracks the performance arc of the campaign. For smaller brands without a research panel, a lightweight pre-test can be five customers reading your three concepts and telling you which one they would stop scrolling for, and why.

Reading results back against the research

When an angle wins, look at which theme it came from and expand it: new hooks, new formats, new proof. When it loses, ask which belief was wrong. Maybe the problem is real but not urgent. Maybe buyers describe it differently than you wrote it. Maybe the objection you tried to answer was not the real one. Feed the answer back into the next interview round. This loop is what makes a creative program compound, instead of resetting every month.

My perspective: research is the cheapest media you will ever buy

We see the same pattern across e-commerce accounts. Brands treat creative volume as the lever and customer research as a nice-to-have. In practice it is the other way around. Ten ads built on one shallow idea will teach you one shallow thing ten times. Three ads built on three distinct, well-researched reasons to buy will tell you which customer to chase.

Our approach joins paid media strategy and creative testing with customer insight and performance data. We trust what comes out of the accounts, and we use research to explain why the numbers moved, not to replace them. A survey will not tell you your cost per acquisition, and a dashboard will not tell you why someone hesitated. You need both.

The most common thing we find when we do this properly is a gap between what the brand thinks it sells and what customers think they bought. Closing that gap usually does more for efficiency than another round of audience tweaking.

— Stefano Mazzei

Where my team at Cosma can help

If you would rather not build this loop alone, my team at Cosma combines media buying, creative strategy, customer research and performance creative for e-commerce and DTC brands. You can see how that plays out in the Cosma case studies, look at creative examples in the portfolio, or book a call with Cosma to talk through your account.

Common mistakes to avoid

Researching only happy customers. Reviews and testimonials skew positive. Interview churned buyers too.
Asking people what ad they would like. Ask about their past decisions instead.
Collecting everything, analyzing nothing. A hundred unread survey responses are worth less than ten coded ones.
Treating the findings as permanent. Customers change, competitors change, and the platform changes. Re-run a light version every quarter.
Changing everything at once. If you launch six new angles, new landing pages and a new offer together, you will not know what worked.

FAQ

How is customer research different from ad testing?

Customer research happens first and decides who you target and what you say, by studying buyers, competitors and channels. Ad testing measures how a draft or finished creative performs with real people. Research gives you the ideas worth testing, and testing tells you which of them works.

How much does customer research for ads cost?

The basic version is close to free. A post-purchase survey, a review mining session and ten interviews cost mainly time and perhaps a small gift card incentive. Paid panels and full research agencies cost much more, and most DTC brands do not need them to find their first strong angles.

How often should I run it?

Run a full round when you launch or reposition a product, and a light round every quarter. Keep the post-purchase survey running continuously, so the verbatims keep coming in and you spot shifts early.

Can I do customer research without a big customer base?

Yes. With few buyers, lean on interviews, competitor and category reviews, and comments on your early ads. Even five to ten honest conversations will usually reveal repeating themes worth testing.