StyleMod,Styled By You
Concept mockup, placeholder data
Read the case studyCase study

Case study · Generative AI in marketing

Personalization drove results, but targeting without guardrails put customer trust at risk

StyleMod, a fashion retailer with about 4 million email subscribers, piloted generative AI to personalize its marketing. The numbers went up, and so did the complaints. Our group looked at where the AI created value, where it destroyed value, and how to keep the gains without the guesswork. The demo above is a working prototype of our recommendation: personalize with customers, not about them.

Format Group case presentation Deliverables Analysis, recommendation, interactive prototype Prototype HTML, CSS and JavaScript, runs in the browser

Value created

The pilot worked on results

2×Email opens
+40%Click-through rate
$2MIncremental revenue, about 10% growth

More relevant messaging

When each customer gets a message tailored to them, engagement goes up, and tailored messaging leads to higher retention.

Customer relationships protected

The model withheld discounts from people who would have bought anyway, so StyleMod wasn't giving away margin or training customers to wait for deals.

Lower costs

Training an internal model meant less effort spent writing copy by hand.

Threats to value

The same personalization broke trust

The complaints were few next to 4 million subscribers, but they went viral, and StyleMod's brand depends on trust and inclusivity. One message like these can do more damage than any gain in click-through.

01

Customer privacy

What happened
A male customer got stiletto promotions. The system combined his purchase history with third-party clickstream data and weighted recent browsing over what he actually bought.
Safeguard
Prioritize first-party purchase history and drop third-party data.
02

Trust and inclusivity

What happened
Maternity wear was emailed to a woman who had miscarried.
Safeguard
Block inferences about sensitive topics such as pregnancy, health and loss.

Recommendation

Keep the AI, add guardrails, and let customers say what they want

A better algorithm still guesses. Letting customers state their preferences removes the guessing. We'd keep the system with three guardrails, and add a short, optional style quiz backed by first-party behavior, so customers who skip it still get relevant messages.

01

No third-party data

Personalize from what customers buy and tell us, not from their browsing elsewhere.

02

Sensitive-topic block

A hard rule against inferring pregnancy, health or loss, whatever the model predicts.

03

Easy opt-out

Customers choose what to share, and can edit or delete it at any time.

The CFO's expansion plan

Scale the low-risk uses first

Expand

Product recommendations

Built on first-party data and low risk.

Expand

Media buying

Lower risk when the AI works within human-set spending limits and safety rules.

Pilot

Chatbot

Chatbots talk directly to customers, so start with narrow tasks like FAQs and order tracking, and route sensitive topics to a person.

Hold

Dynamic pricing

Personalized prices could make customers feel they're being treated unfairly.

From slides to prototype

How the demo puts the recommendation into practice

We built the demo to show the recommendation working rather than just describing it. A rules-based engine stands in for the model, and comments in the source mark where a real generative model would plug in.

  • Stated, not inferredFive questions drive every recommendation. Gender sets the section and dummy only because the customer chose it, and the quiz says so.
  • No third-party dataThe "How we personalize" panel shows exactly which answers are in use. Nothing comes from browsing on other sites.
  • Sensitive-topic blockThe quiz never asks about, or guesses at, life events or health.
  • Easy opt-outCustomers can turn personalization off on the results page, or clear all their data in one step. Everything lives in the browser tab and disappears when it closes.
  • Explain the pickEach outfit says why it was chosen, and every piece can be swapped, hidden or reshuffled.
  • Learn from feedback"What wasn't right?" adjusts the next picks for that session, so the customer corrects the system instead of the system guessing.

People and measurement

What changes, and how we'd know it's working

Copywriters

Their role shifts to training the AI, polishing high-profile campaigns and refining the brand voice, work that still needs human judgment.

Cost

The plan still saves on copywriting and scales the low-risk uses, while avoiding the cost of lost trust.

Net results

Track engagement and conversion alongside unsubscribes, complaints and repeat purchases, with a holdout group that gets no AI messaging to confirm the lift.

Q&A

Questions we prepared for

Is the pilot a success or a failure?

A success on results, with a real trust risk that needs fixing.

How do you know the AI caused the lift?

The results are consistent with the AI driving it, and we'd confirm with a holdout group that doesn't get AI messaging.

Why not just fix the algorithm?

A better algorithm still guesses. Letting customers state their preferences removes the guessing.

Would people actually take the style quiz?

Most won't, so we'd keep it short and optional, and first-party behavior would fill the gaps. That way low participation doesn't erase the gains.

Isn't "personalize with them" just collecting more data?

It's different because customers choose what to share and can edit or delete it.

Do customers care about privacy if they get a good deal?

Some don't, but the complaints show others do, and lost trust is hard to rebuild.

See it for yourself. Answer five questions and get three outfits built only from what you chose.

Try the demo

Clear my data?

This removes your answers, saved outfits and feedback from this session and returns you to the start.