BrandyMelvilleA/BTest

The Data Analytics for Marketing module asked for an individual report built on primary research — collect your own data, test it with inferential statistics, and turn what you find into marketing recommendations. It was worth 75% of the module, with a limit of 3,600 words for a report that tested mediation.

I chose the brand, the question, the experiment and the analysis. I'm also in Brandy Melville's customer profile, so I knew the problem from the inside, and I thought the answer would interest the company as much as it interested me.

Brandy Melville built a loyal Gen Z following on minimalist style and Instagram. Its "one size fits all" policy, allegations of discriminatory hiring and a 2024 HBO documentary turned that following into a reputation problem.

My view was that the problem sits as much in the imagery as in the sizing. The site shows one narrow body type, so most shoppers never see themselves in it. If that's true, a change of imagery alone — with no change to the product — should move how people feel about the brand.

I started with that question and then looked for the theory to test it. Keller's Customer-Based Brand Equity model fitted, and it's a framework I'd enjoyed using in other modules. It says imagery shapes the associations people hold about a brand, and those associations drive what they think of it and whether they buy.

• H1 — Inclusive imagery leads to higher purchase intention than the current imagery.
• H2 — Inclusive imagery goes with more favourable attitudes towards Brandy Melville.
• H3 — Inclusive imagery and positive attitudes together predict higher purchase intention.

I ran a between-subjects A/B test in Google Forms. Each person was randomly sent to one of two identical surveys, and saw one image. Group A saw Brandy Melville's current product page. Group B saw the same page with models of different body types and ethnicities.

I made Group B's page myself. I searched Pinterest for photos shot the way Brandy Melville shoots its own, then rebuilt the page in Canva with those images in place of the originals. What "inclusive" should look like came from brands whose representation I wanted Brandy Melville to match.

I recruited 102 people aged 18 to 26 through QR codes on campus and a shared survey link, with snowball sampling to reach younger and older respondents. 52 landed in Group A and 50 in Group B. The sample was 95% women — mostly a matter of who I could reach, but also close to Brandy Melville's own target market.

The scales came from published research (Aaker et al., 2000; Spears and Singh, 2004; Dodds et al., 1991), as my professor encouraged, and a manipulation check confirmed the two images really did read differently.

All three hypotheses held, with large effects. People who saw the inclusive page rated the brand more favourably and said they were more likely to buy.

The bigger finding was why. In a regression, brand attitude and imagery together explained 69% of purchase intention — but once attitude was in the model, the image itself stopped mattering. A bootstrapped mediation test (5,000 samples) confirmed it. Inclusive imagery works entirely by changing how people feel about the brand.

That made sense to me. Most people's problem with Brandy Melville is its reputation. Take that away, and the clothes are still very appealing.

I didn't take the numbers at face value. Effects this large are unusual for a single change, and very high reliability scores can mean the scale items overlap. Most women I surveyed already thought Brandy Melville had an inclusivity problem, so response bias is possible, and I said so in the report.

• Bring in inclusive imagery gradually, starting with new product launches, so the brand changes without alienating the customers who like it as it is.
• Make brand attitude, not sales, the main marketing objective. It moved most, and it's the route by which imagery reaches purchase.
• Track brand attitude as a KPI, with pulse surveys before and after each campaign, so the team can tell real change from surface-level engagement.

The end goal is a relaunch with the new identity, phased — website imagery first, then social media, then the stores, including hiring people who embody the new look.

I declared my use of GenAI in full, as the module required. Microsoft Copilot suggested two survey questions that I kept after a pilot test. Claude helped me format the responses for jamovi and explain some of the output, which I checked against the lecture slides and with my professor.