Our client, FORMEL Skin, is a subscription service providing medical advice and personalized cosmetic solutions for people with skin problems.
FORMEL Skin's Business Model
- Consultation Request: Customers initiate a consultation with a doctor via the website.
- Skin Concerns Addressed: Issues may include acne, rosacea, age spots, and more.
- Remote Consultation: Clients sign up for a video consultation, during which the doctor examines the skin, asks relevant questions, recommends tests if needed, and prescribes a tailored treatment plan.
- Monthly Treatment: FORMEL Skin provides ongoing care through monthly shipments of doctor-recommended cosmetics, specifically selected for each client.
The treatment is done on a monthly basis.
Every month FORMEL Skin sends the client a set of cosmetic products selected by the doctor especially for them.
FORMEL Skin Marketing Features
Significant Budgets: Monthly marketing expenses range from €20,000 to €50,000.
Large-Scale Campaigns: Hundreds of advertising campaigns are launched each month.
Extended Customer Lifecycle: Customer interactions can span up to two years.
Variability in LTV: Customer lifetime value (LTV) varies based on audience demographics and engagement channels. For instance, customers aged 18–25 with severe skin issues tend to remain in the system longer, contributing to higher revenue.
Delayed Purchases: Many campaigns utilize promo codes that customers may activate even after the campaign has concluded.
Large-Scale Campaigns: Hundreds of advertising campaigns are launched each month.
Extended Customer Lifecycle: Customer interactions can span up to two years.
Variability in LTV: Customer lifetime value (LTV) varies based on audience demographics and engagement channels. For instance, customers aged 18–25 with severe skin issues tend to remain in the system longer, contributing to higher revenue.
Delayed Purchases: Many campaigns utilize promo codes that customers may activate even after the campaign has concluded.
The main problem with subscription services is high advertising costs that take a long time to pay off.
A marketer's goal is to assess the effectiveness of a campaign promptly, without waiting several months.
Solution: Develop a user-friendly and insightful reporting tool for marketers.
The Process
Step 1: Build End-to-End Analytics for Advertising Campaigns.
We collect spend and click data from all advertising offices and link it to user purchases on the site.
We collect spend and click data from all advertising offices and link it to user purchases on the site.
Step 2: Develop a Basic Marketing Report
The report enables marketers to:
The report enables marketers to:
- View the costs of each campaign.
- Identify the number of paying users attracted by the campaign.
- Calculate the cost of acquiring one user (Cost Per User, CPU).
- Measure the average Lifetime Value (LTV) of a user—the revenue generated over their lifetime.
Step 3: Evaluate Advertising Campaign Effectiveness.
We consider an advertising campaign to be effective if the revenue from the attracted user exceeds the cost of acquiring them.
We consider an advertising campaign to be effective if the revenue from the attracted user exceeds the cost of acquiring them.
- If a user generates twice the cost of acquisition, the campaign is profitable.
- If revenue is less than the cost, the campaign is unprofitable.
Performance Evaluation Formula:
- LTV/CPU < 1: Campaign is unprofitable.
- LTV/CPU > 1: Campaign is effective.
- LTV/CPU > 2: Campaign is profitable.
Challenge:
The final LTV of a user becomes clear only after 1.5–2 years, as customers make regular payments. However, marketers need to evaluate a campaign's effectiveness as early as possible—ideally immediately after a user's first purchase.
Solution:
Machine learning predicts LTV early in the customer lifecycle.
The final LTV of a user becomes clear only after 1.5–2 years, as customers make regular payments. However, marketers need to evaluate a campaign's effectiveness as early as possible—ideally immediately after a user's first purchase.
Solution:
Machine learning predicts LTV early in the customer lifecycle.
Step 4: Train an ML Model to Predict LTV.
We use customer data to train the model, including:
We use actual LTV and lifetime data for each user to train the ML model. The model then analyzes the parameters of new users who make their first purchase and predicts their LTV.
We use customer data to train the model, including:
- Demographic Data:
- Date of birth, age, age group, gender.
- Pregnancy or breastfeeding status, pregnancy planning.
- Health and Treatment Data:
- Treatment goal, treatment status, presence of allergies, other medical symptoms.
- Diagnosis, complexity of diagnosis, treatment plan, and duration.
- Behavioral Data:
- First and last payment dates.
- Regular vs. one-time payments.
- NPS (Net Promoter Score).
- Marketing Data:
- UTM parameters (source channel, etc.).
We use actual LTV and lifetime data for each user to train the ML model. The model then analyzes the parameters of new users who make their first purchase and predicts their LTV.
How It Works:
- When a user acquired through an advertising campaign makes their first purchase, the system predicts their Lifetime Value (LTV). This helps determine if the acquisition cost will be recovered.
- With every additional purchase, the prediction is updated, providing marketers with increasingly accurate data to assess campaign effectiveness.
Results
From the First Purchase we can:
- Determine the profitability of a specific advertising campaign.
- Predict the customer’s Lifetime Value (LTV).
- Assess the effectiveness of each advertising channel.
