Taobao
Real-time product recommendation with deep learning, served across hundreds of scenarios in the app
On Taobao, real-time product recommendation with neural networks covers more than 80% of mobile users and 1000+ scenarios; productized as Alibaba Cloud AIRec, it delivered a 9% conversion gain for a sportswear retailer.
Key points
- Real-time product recommendation with deep learning, the second navigation mode after search.
- Recall-then-ranking chain on Alibaba Cloud AI OS, productized as AIRec.
- More than 80% of Taobao Mobile users covered, 1000+ scenarios; 9% conversion gain for an AIRec client.
- Evidence level B, living status confirmed.
Objective
Make recommendation the app's second navigation mode after search, so that a vast catalog stays browsable and converts.
The deployment
On Taobao, Alibaba's consumer e-commerce app, personalized recommendation has become the second navigation mode after search. The system captures user behavior (views, clicks, purchases) and serves relevant products within milliseconds, on the home page, in the You May Like feed, on search pages, and during major events like 11.11. The architecture relies on deep neural networks and a recall-then-ranking chain, with model update cycles reduced from hours to minutes. According to Alibaba Cloud, the system covers more than 80% of Taobao Mobile users and powers more than 1000 personalized scenarios across the group. The same technology, productized under the name Alibaba Cloud AIRec, is sold to third-party merchants: in one sportswear retailer case, it raised the transaction conversion rate by 9% and reduced the empty-results rate by 80%.
Results Proof B
Figures published by Alibaba Cloud (official technical blogs), therefore an interested source not audited by a third party. Two concordant articles describe the scale on Taobao; the conversion figures come from a client case of the productized AIRec solution.
How it works
Documented architectureThe stack in detail
- plateforme Alibaba Cloud AI OS Alibaba's internal platform that unifies recall, ranking, and real-time serving of recommendations across Taobao surfaces.
- llm Reseaux de neurones profonds recall + ranking In-house models for candidate selection then ranking, with update cycles reduced from hours to minutes.
- plateforme Alibaba Cloud AIRec Productized version of Taobao's recommendation engine, sold to third-party merchants.
- infra Flux comportemental temps reel Continuous capture of views, clicks, and purchases that updates the user profile live.
How it runs, concretely
For ops teams-
1Capture behavior AI / platform
Every interaction in the app is reported and updates the user profile live.
-
2Recall candidates AI / platform
A first layer (recall) selects a subset of relevant items from a vast catalog.
-
3Rank and serve AI / platform
The neural networks rank the candidates and deliver the list within milliseconds on the relevant surface.
-
4Refresh the models Data / infra team
The update cycles, reduced from hours to minutes, quickly incorporate new signals, useful during peaks like 11.11.
In-session behavior (views, clicks, purchases) captured continuously. Without this real-time stream, recommendations fall back to generic lists and the feed loses its relevance.
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
- Real-time behavioral stream at audience scale
- Broad product catalog with usable attributes
- Low-latency serving infrastructure
Org prerequisites
- ML and infra team able to run a real-time service
- App surfaces instrumented per scenario
Possible stack
- Alibaba Cloud AIRec or an equivalent managed recommendation engine (AWS Personalize, Google Recommendations AI)
- Feature store plus a ranking model for a custom version
The plan, step by step
- Step 1Instrument the behavioral stream (views, clicks, purchases) and structure the catalog with stable identifiers.Deliverable: Reliable real-time event pipeline connected to the catalog
- Step 2Launch a first managed recommendation (AIRec, AWS Personalize, or equivalent) on a single surface, the home page.Deliverable: A for-you-style feed in beta on a share of traffic
- Step 3Run an A/B test against navigation without recommendation and read CTR, conversion, and navigation share.Deliverable: Controlled test result and rollout decision
- Step 4Extend to the other surfaces (search, product pages) and shorten the model update cycle.Deliverable: Recommendation served across multiple scenarios with freshness monitoring
First step: Add a for-you feed on the app's home page, powered by session behavior, and measure its share of navigation and conversion against search alone.
The same pattern, proven in other industries
Function of Beauty
personalized product formulation by algorithm from a quiz
Henkel
genAI content supply chain for one-to-one personalization at scale
Hims & Hers
AI-driven personalized treatment routing on intake data (MedMatch)
Sources
- S1 The Secret Behind Taobao's AI-Powered Personalized Recommendations - Alibaba Cloud Community Interested party archive pending
- S2 How Taobao Applies AI to Personalized Recommendations - Alibaba Cloud Community Interested party 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.