The Emily in Paris playbook: Two personalized models that drive engagement

Incorporating personalization into customer experiences has become essential for brands aiming to meet and exceed revenue targets. According to recent research by Forrester and Deloitte Digital, brands that effectively use personalization strategies report a significant boost in loyalty and revenue performance.

Forrester found that brands prioritizing personalization throughout the entire customer lifecycle achieve greater alignment with customer expectations, leading to stronger loyalty and increased sales.

Deloitte Digital’s 2024 report similarly highlights that companies emphasizing tailored, data-driven experiences are 48% more likely to exceed revenue targets, illustrating the value of personalization when it’s built as a core part of the strategy.

In today’s competitive market, I hope to demonstrate the undeniable value of personalization. However, I’d rather illustrate my points through a concrete example instead of delving further into theoretical aspects.

Pop culture inspo

I’m a big fan of drawing inspiration from popular culture, especially TV series, to craft relatable scenarios. So, for this article, let’s consider characters from the hit Netflix show Emily in Paris and imagine developing a subscription-based beauty box service tailored specifically for them—akin to the popular UK-based Glossybox.

Let’s dive into this fictional exercise to outline how such a service could be strategically developed for each unique persona in the show.

Personalized subscription services are on the rise, particularly in the beauty sector. A pop-culture example, like the characters from Emily in Paris, offers an intriguing opportunity to craft unique, engaging, and profitable subscription boxes.

By leveraging AI algorithms and data, businesses can deliver highly personalized and memorable experiences. This article explores how to build a subscription model for beauty boxes using activation, retention strategies, and personalized offers. Additionally, it discusses creating CX personas to better engage customers and drive performance.

Problem

I suppose that problem of personalized subscription services - engagement and retention. Subscription services often struggle to retain users when their offerings lack personalization. Addressing the unique needs of different customer segments is essential when working on activation, churn reduction, or increasing purchase frequency.

When we started developing our beauty box subscription, the concept seemed straightforward: take a proven idea and tailor it to our brand. For our MVP, we mirrored strategies from competitors, using broad promotions and generic product offerings. However, the initial results fell far short of expectations.

Baseline challenges

It quickly became clear that copying competitors’ methods blindly wasn’t a path to growth. Instead, it highlighted some significant issues:

Low conversions: Generic offers failed to attract new users. Despite our marketing efforts, the conversion rate barely reached 6–8%.
High churn rates: Customers would try one box but find little value, leading to a 30% dropout rate after the first subscription cycle.
Low customer satisfaction (CSI): Customer satisfaction ratings were only 65%, with complaints about the lack of personalization and product diversity.

It was a wake-up call: our one-size-fits-all approach wasn’t resonating with customers or driving business results. We needed a fundamental change.

"In our approach, we focus on several goals: activating new users, reducing churn, engaging frequent and infrequent shoppers, and replacing mass promotions with personalized offers. With over 140 filters and advanced ML models—uplift, upsell, cross-sell, and more—we can precisely predict which offer or channel will work best for each segment."

Step 1: Creating CX personas inspired by Emily in Paris

Developing CX personas is a great way to segment the audience and personalize their experience. To implement the CX persona method, we conducted in-depth interviews and applied the Jobs-to-Be-Done (JTBD) framework.

We also identified emotional behavior patterns through deep analytics combined with this approach, focusing on the concept of Emotions to Be Fulfilled. This allowed us to uncover not only functional needs but also the emotional drivers that influence customer decisions, such as the desire for excitement, reassurance, or a sense of exclusivity.

Here’s how personas for the show’s characters might look:

Emily Cooper (New to the city): A young, vibrant woman who values trends and loves trying new things. Her beauty box would feature samples of popular new products. Metrics: Activation of new users, conversion rates. Triggers: Interest in luxury brands, one purchase per month, high email open rate on Monday mornings.
Mindy (Expat with a bold image): She loves experimenting with makeup and perfume. Her beauty box could focus on unique, niche items. Metrics: Increase in average order value and purchase frequency. Triggers: Interest in niche brands, frequent purchases in "makeup" and "accessories."
Sylvie (Sophisticated professional): Loyal to classic and premium brands, her box would feature full-sized skincare products. Metrics: Reducing deep churn. Triggers: High cart value, infrequent purchases (once every three months), interest in anti-aging cosmetics.

Step 2: Leveraging ML models for activation and retention

AI-powered insights:

With ML models, you can calculate:

Uplift: The likelihood of purchase after receiving a specific offer.
Upsell: Strategies to increase cart value.
Cross-sell: Adding complementary products.

Optimal channels: Identifying the best communication method (email, push, SMS). We also incorporated data from the CX persona modeling, including character behavior patterns, habits, and emotional experiences with similar services—and beyond—into our machine learning model.

We train our ML model using interaction data from multiple sources, such as customer views, purchases, add-to-bag actions, and items saved for later. This interaction data is passed through a neural collaborative filtering algorithm, which transforms it into embeddings for each product and customer.

A key feature of these embeddings is that similarity within the [embedding space](<https://medium.com/asos-techblog/getting-personal-at-asos-bc1599e0c2a9>) (based on a similarity function) represents the similarity between products or the preference relationship between users and products.

In our case, this data not only enhances the personalization of offers for each customer but also optimizes the experience of interacting with the beauty box. It allows us to accurately predict which products and product combinations will interest each user, thus improving engagement, loyalty, and conversions.

Example for Emily:

"Emily responds well to push notifications in the mornings, particularly early in the week, improving her open rate by 15%."

Example for Sylvie:

A high-risk churn prediction could prompt a personalized gift offer, such as premium skincare with her next purchase.

Step 3: Creating data-driven segments

Segmentation can be refined using over 140 attributes, including:

Demographics: Age, gender.
Purchase behavior: Product categories, frequency, cart value.
Engagement history: Response to offers, preferred channels.

Example:

For infrequent shoppers like Sylvie, a segment titled "High-Value, Low-Frequency Buyers" might target her with exclusive gifts tied to premium purchases.

Step 4: Hypothesis testing and campaign launch

Every campaign should undergo A/B testing:

Offers: Discount vs. loyalty points.
Channels: Push notifications vs. email.

Example:

"For Mindy’s segment, a niche brand promotion resulted in a 20% increase in conversion rates."

Step 5: Tracking metrics and refining the model

Key performance indicators include:

Revenue Turnover (RTO): Measures revenue growth.
Ebitda and Gross Margin (GM): Evaluates profitability.
Conversion Rate: A critical metric for activating new users.
Customer Satisfaction Index (CSI): Assesses user satisfaction.

Results

By adopting this approach, businesses can achieve:

Activation: Attracting new users through tailored offers.
Retention: Reducing churn with personalized communications.
Increased frequency and spend: Driving repeat purchases and higher cart values through precise targeting.

Two approaches, one goal

The combination of CX personas and AI-powered personalization highlights two complementary strategies for crafting subscription services. CX personas help map emotional and behavioral connections with users, while AI drives precision through data and predictive analytics. Businesses can balance these approaches based on their goals and resources.

While some professionals may lean on personas for their human-centered insights, others might prioritize AI for scalability and real-time adaptability. Each approach, when implemented thoughtfully, has the potential to transform subscription models into memorable, revenue-generating experiences.

Subscription models with CX personas and AI

In the world of subscription services, personalization has become a necessity rather than an option. But how do you strike the right balance between human insight and data-driven precision? Combining CX personas with AI-powered personalization offers a powerful framework for creating experiences that feel both empathetic and hyper-relevant.

CX personas provide a foundational understanding of customer motivations, while AI scales this knowledge by predicting behaviors and delivering real-time, personalized interactions.

By integrating these approaches, subscription services can craft unique, memorable offerings that drive retention and loyalty. Inspired by the vibrant characters of Emily in Paris, let’s explore how to create subscription experiences that captivate every persona while leveraging the cutting-edge capabilities of AI.

Let’s see the challenge 😎

For any subscription service, retention is the backbone of profitability. Yet too many offerings lack relevance, treating diverse customer needs as one-size-fits-all.

Common retention challenges:

New user activation: How do we get the “Emilys” of our audience to convert and stay engaged?
Churn prevention: How do we retain high-value but at-risk users like Sylvie?
Increasing frequency: How do we entice frequent shoppers like Mindy to purchase more often or expand their basket?

Data from CVM teams supports this:

"Our focus is to move beyond mass promotions by targeting segments based on behavior, demographics, and triggers. AI models like uplift, upsell, and cross-sell predict what offer works best for whom, in which channel, and when."

These same principles can transform subscription services, ensuring every product and offer is hyper-relevant to the customer.

At the core of subscription success is the ability to personalize at scale. With AI, you can move from reactive personalization (static personas) to proactive, data-driven strategies that adapt dynamically to user behavior.

The AI personalization framework

1. Data collection:

Aggregate all available touchpoints: demographics, preferences, purchase history, triggers (e.g., abandoned carts, browsing patterns).
Use interaction data (e.g., time of day users engage, preferred channels).Example: Emily responds to push notifications about trending products early in the week.

2. Behavioral segmentation:

Leverage CX personas to guide high-level strategy but allow AI to refine groups based on micro-trends.

Emily: A curious new user who wants sample-sized versions of popular products.
Sylvie: A lapsed but high-value customer drawn to premium skincare.
Mindy: A frequent buyer seeking bold, experimental products.

3. AI-driven offer optimization:

Train machine learning models to deliver:

Uplift modeling: Predict who’s likely to buy with the right nudge.
Cross-sell opportunities: Identify product pairings to increase basket size.
Dynamic discounts or rewards: Adjust offers based on customer elasticity.

4. Communication strategy:

Match content to channels and timing preferences.

Push notifications for Emily, with colorful and trendy visuals.
Email for Sylvie, with a premium tone and exclusive gift offers.
Social media for Mindy, emphasizing niche and creative products.

5. Continuous A/B testing:

Test variants of messaging, channels, and offers across user cohorts to optimize ROI.

6. Measuring success:

Align personalization efforts to business KPIs:

Conversion rate: Higher acquisition of new users like Emily.
Churn reduction: Improved retention for Sylvie.
Revenue lift: Increased average order value for Mindy.

**Bringing the Emily in Paris personas to life**

Here’s how AI personalization translates into actionable strategies for different customer types:

Emily (New User)
- Objective: To activate Emily by capturing her attention and encouraging initial engagement.
Solution: Introduce a trial beauty box featuring trendy sample-sized products tailored to her preferences. Push notifications, timed to peak engagement moments (e.g., Monday mornings), deliver the offer in an engaging and timely manner, leveraging data on her behavior and communication preferences.
Expected outcome: By aligning the offer with Emily’s interests and delivering it via the optimal channel, the trial aims to convert her casual curiosity into sustained loyalty, increasing the likelihood of repeated interactions and eventual subscription.

Sylvie (lapsed customer)
Objective: to prevent Sylvie from churning by re-engaging her with targeted offers.
Solution: Launch an email campaign offering an exclusive skincare product paired with a premium gift. The messaging highlights luxury and sophistication, aligning with her preference for premium brands and products.
Expected outcome: High-value incentives tailored to Sylvie’s tastes are designed to rekindle her interest and bring her back into the fold, increasing the likelihood of future high-value purchases

Mindy (frequent buyer)
- Objective: to increase Mindy’s basket size by encouraging her to explore complementary products.
Solution: Implement a cross-sell strategy featuring bold lipsticks paired with niche perfumes. Highlight these combinations through visually appealing Instagram campaigns, leveraging her interest in experimental and unique beauty products.
Expected outcome: The curated product pairings aim to boost Mindy’s average order value (AOV) while enhancing product discovery, encouraging her to explore and purchase more items per transaction.

Results: AI personalization in action

When AI-powered personalization is implemented effectively, the results are measurable:

+20% increase in conversions through tailored first-touch offers.
15% reduction in churn by targeting users with at-risk behaviors.
Higher customer satisfaction (CSI) due to relevant and engaging interactions.

Conclusion: Dual methodology AI personalization & CX personas

As product managers and marketers, we often balance two approaches to personalization:

Use CX personas for strategy: They’re excellent for framing product positioning and segmentation during early stages or when launching a new feature or line.
Rely on AI for execution: Machine learning allows for real-time adjustments, uncovering micro-segments and delivering highly individualized experiences.
Integrate for best results: CX personas give your team clarity and empathy, while AI ensures precision and scalability. Together, they create a robust personalization engine that’s both human-centric and data-driven.

This dual methodology strikes a balance between creative intuition and data-informed decision-making, allowing businesses to create stronger, more relevant connections with their audience. The flexibility of this approach is particularly noteworthy.

Brands can lean more heavily on personas when crafting emotionally resonant campaigns or rely on AI’s predictive capabilities to achieve efficiency at scale. Whether the focus is immediate revenue growth, customer retention, or long-term loyalty building, this synergy ensures adaptability to varied business goals.

BONUS My top recommendations 🎁

Mine data for precision insights

Drill down into customer lifecycle data to uncover actionable insights. This involves segmenting audiences, diving into micro-segments, and mapping behavioural, transactional, and engagement patterns. This groundwork enables us to set laser-focused personalization KPIs that directly enhance customer value.

Explain the importance of data collection to users

It’s crucial to communicate to customers why sharing their preferences and data is beneficial for both parties. Implementing simple, short surveys during onboarding can ensure users understand how their input helps create personalized experiences, saving them time and providing more relevant recommendations.

For example, online clothing store ASOS asks customers to fill out their size and style preferences, which helps tailor the shopping experience. Personalized features such as the “You Might Also Like” carousel on product pages, curated “For You” sections on the homepage, and customized search results are possible because of this data. You can create a more personalized and user-friendly experience.

Deploy real-time activation at scale

Leverage AI and predictive analytics to serve customers with context-aware, timely interactions. By setting up rapid activation engines—such as propensity models and next-best-action algorithms—we’re equipped to respond in real-time, ensuring the right content reaches the right customer at the perfect touchpoint in their journey.

Engineer a targeted Martech stack

Build and optimize a fit-for-purpose martech ecosystem calibrated to drive specific customer outcomes. We work backwards from key use cases and desired outcomes to assemble data capabilities and technology enablers, aligning the stack to support scalable personalisation with pinpoint precision.

Adopt an agile, cross-functional operating model

Organize cross-disciplinary teams that run on an agile framework, using a hub-and-spoke model. Each hub owns a piece of the personalization journey, while spokes develop distinct use cases, allowing us to scale up experimentation. This approach maximizes test velocity and lets us adapt rapidly based on insights.

Take part in The State of Product Ops 2025 Survey and give exclusive insights into the latest trends, challenges, and opportunities shaping this critical function.