Olay
AI skin diagnostic on a selfie leading to a product recommendation
In 2016, Olay (P&G) launched Skin Advisor, a deep-learning skin diagnostic on a selfie that doubled the conversion rate, cut bounce by a third, raised the average basket by 40% in China, and engaged more than 4 million consumers.
Objective
Reduce friction and uncertainty when buying skincare online by replacing the catalog with a personalized diagnostic, to convert better and raise the basket.
The deployment
Olay Skin Advisor asks for a selfie and a few answers, then estimates a skin age and points to the areas to treat with deep learning. From there, the tool suggests a routine and tailored Olay products. Launched in 2016 after two years of development and the analysis of millions of selfies, it first ran in beta in North America before a global rollout, including China. P&G reports that the platform doubled the sales conversion rate, cut the online bounce rate by a third, and raised the average basket. In China, the average basket rose by 40 percent. The tool has engaged more than 4 million consumers and serves recommendations to tens of thousands of users each week, 94 percent of whom receive personalized suggestions. The recommendation engine was built with Nara Logics, with the image analysis relying on NVIDIA's GPU platform.
Results Proof C
Figures attributed to P&G and reproduced by several press outlets (VentureBeat, Marketing Dive, The Drum) naming Olay; no separate line in the group's financial results, hence C rather than A.
How it works
Documented architectureThe stack in detail
- llm Modele de vision deep learning P&G Image-analysis model developed for Olay, trained on millions of selfies: estimating skin age and detecting the areas to treat.
- outil Nara Logics AI recommendation engine that matches the skin profile and questionnaire answers to Olay products and routines.
- infra NVIDIA (plateforme GPU / deep learning) GPU platform used for training and inference of the image-analysis models.
- outil Skin Advisor (web app Olay) Consumer web app that carries the selfie, questionnaire, diagnostic, and recommendation journey, linked to Olay e-commerce.
How it runs, concretely
For ops teams-
1Selfie capture and analysis AI
Estimating skin age and detecting the areas to treat with vision and deep learning.
-
2Follow-up questionnaire user
A few questions about concerns and habits refine the profile.
-
3Routine recommendation AI (Nara Logics engine)
The engine matches the profile to Olay products and proposes a routine.
-
4Move to purchase e-commerce
The recommended products are pushed to the e-commerce cart, measured on conversion and basket.
The quality of the selfie and the consistency between the skin score and the recommendation. Poor framing or lighting degrades the diagnostic and undermines confidence in the recommendation.
How your customers perceive this type of use
Sourced studiesLe paradoxe est documente des deux cotes : 71% des consommateurs attendent des interactions personnalisees et 76% sont frustres quand elles manquent (McKinsey, 2021), mais 75% declarent ne pas acheter aupres d'organisations auxquelles ils ne confient pas leurs donnees (Cisco, 2024). La « creepy line » est localisee : messages recus quelques secondes apres une recherche et suivi de localisation sont les pratiques qui mettent le plus mal a l'aise (Periscope by McKinsey, 2019).
Acceptance conditions
- La confiance dans le traitement des donnees precede l'achat : 75% ne achetent pas sans elle (Cisco 2024)
- Un cadre legal protecteur rassure : 59% des consommateurs disent que des lois fortes sur la vie privee les rendent plus a l'aise pour partager des informations dans des applications IA (Cisco 2024)
- La personnalisation elle-meme est attendue quand elle est consentie : environ la moitie des consommateurs (US 55%, UK 52%) disent s'inscrire souvent ou parfois a des services personnalises (Periscope by McKinsey 2019)
Red lines
- Le message declenche quelques secondes apres une recherche ou un achat : deuxieme ou troisieme cause de malaise selon les pays (Periscope by McKinsey 2019)
- Le suivi de localisation percu comme de la surveillance : 40% de malaise en Allemagne et au Royaume-Uni (Periscope by McKinsey 2019)
- Le mesusage des donnees personnelles par l'IA, devenu la premiere inquietude des consommateurs, a 53% et en hausse (Qualtrics 2025)
Sources: McKinsey & Company 2021 · Periscope by McKinsey 2019 · Cisco 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- corpus of labeled skin images to train or calibrate the model
- structured product catalog with skin attributes
- GDPR-compliant biometric consent
Org prerequisites
- data science team or a vision partner
- governance on the non-medical nature of the output
Possible stack
- third-party skin-analysis API (such as Haut.AI, Perfect Corp, ModiFace)
- product recommendation engine
- integration with the CMS / e-commerce
The plan, step by step
- Step 1Pick a category with high purchase hesitation and frame the diagnostic: which skin attributes lead to which products in the catalog.Deliverable: Skin-profile-to-catalog mapping, validated by the product teams
- Step 2Integrate a market skin-analysis API (Haut.AI, Perfect Corp, ModiFace) with the follow-up questionnaire, white-labeled on the site.Deliverable: End-to-end working diagnostic prototype
- Step 3Secure the biometric aspect: explicit consent, minimization, clear notice that the output is indicative and not medical.Deliverable: Legally validated consent journey
- Step 4Launch an A/B test of diagnostic vs standard catalog page and measure conversion, basket, and bounce.Deliverable: Test read with a go/no-go decision
- Step 5Industrialize: CRM integration, a full multi-product routine, and a usage-feedback loop to refine recommendations.Deliverable: Feature in production with continuous monitoring
First step: Test a skin diagnostic on a high-hesitation category and measure conversion and basket against the standard catalog page.
The same pattern, proven in other industries
Taobao
Real-time product recommendation with deep learning, served across hundreds of scenarios in the app
Amazon
Item-to-item collaborative filtering product recommendation engine, served across every touchpoint
ASOS
catalog-scale deep learning product recommendation, plus a conversational style assistant
Sources
- S1 How Olay used AI to double its conversion rate Established press archive pending
- S2 VentureBeat: Olay doubles conversion rates with AI-powered skincare advisor Established press archive pending
- S3 How P&G harnessed AI to create 'first of its kind' Olay Skin Advisor Established press archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.