Huawei Cloud Retail China Translated from Chinese

Dingdong Maicai

Marketing Personalization · Supply Chain Optimization

Dingdong Maicai needed more accurate demand forecasting, rider dispatch, and inventory optimization for time-sensitive fresh delivery, but its old big-data platform and algorithm environment were costly to run and hard to scale during peak periods.

Read the original story Explore all use cases

AI computing performance improved 3-5 times

average monthly orders per user increased by more than 30%

big-data job compute efficiency improved 20%

fastest delivery in 29 minutes

Reported impact

Business outcome

Dingdong improved prediction and dispatch quality while lifting large-data efficiency, lowering cost, and helping the business deliver faster service with higher order conversion and lower waste.

Solution

Dingdong Maicai rebuilt its analytics and AI foundation on Huawei Cloud so forecasting and dispatch models could keep up with fresh-grocery delivery. The architecture separates storage and compute by using Object Storage Service OBS as a shared data layer for transaction, browsing, supply-chain, warehouse, weather, and fulfillment data, while MapReduce Service MRS provides elastic large-data clusters for offline and real-time analysis. On top of that base, the company transformed its algorithm platform with CCE Turbo, heterogeneous X86/Kunpeng/Ascend resources, and high-performance GPU compute instances so training jobs and temporary workloads can scale automatically during lunch and dinner peaks. Models then use user, order-time, rider, route, warehouse, product, and weather signals to recommend products, forecast front-warehouse demand, and generate optimal fulfillment paths.

Data flow

Transaction, browsing, order, supply-chain, warehouse, rider, weather, and route data lands in OBS and is processed in MRS for offline and real-time analytics. The algorithm platform then trains and serves recommendation, demand, and dispatch models on CCE Turbo and GPU resources, returning predictions to warehouse stocking and delivery-routing workflows.

Solution architecture

6 components · 4 layers

Compute
- MapReduce Service MRS Runs the elastic offline and real-time analytics jobs that feed forecasting and fulfillment models.
- high-performance GPU compute instances Accelerate model training and AI inference for recommendation, prediction, and dispatch tasks.
Storage
- Object Storage Service OBS Stores the shared transaction, user, supply-chain, and fulfillment datasets used by analytics and AI workloads.
Serving
- algorithm platform Hosts the recommendation, demand-forecasting, and dispatch models used in daily operations.
Orchestration
- Cloud Container Engine CCE Provides the container platform on which the cloud-native algorithm environment operates.
- CCE Turbo Supplies second-level elastic scaling and heterogeneous resource scheduling for model workloads.

Architecture clues

CCE Turbo mixes X86, Kunpeng, and Ascend resources for the algorithm platform.
GPU resources are shared across multiple containers for temporary AI training tasks.
The design explicitly separates storage and compute with OBS and MRS.
The platform supports both real-time dispatch inference and offline demand forecasting.

Evidence from the source

"CCE Turbo supports heterogeneous X86, Kunpeng, and Ascend resources, and high-performance GPU instances lift AI computing performance by 3-5 times. (original: 基于华为云CCE Turbo支持X86、鲲鹏、昇腾等多种异构算力能力... 采用华为云高性能GPU计算实例... 在AI计算性能上比通用方案提升3-5倍以上。)"

"Dingdong and Huawei built a unified big-data foundation, generate best delivery paths in real time, and use cloud-native transformation to accelerate algorithm iteration. (original: 叮咚买菜与华为云合作，构建统一的大数据底座，实时生成配送最佳路径，并使用云原生改造算法平台，加速算法迭代。)"

"OBS provides unified data storage and MRS improves offline and real-time analytics by 25%, while cutting per-PB expansion cost by 30%. (original: 基于华为云OBS进行统一的数据存储... 相比之前大数据存储每PB扩容成本降低30%；基于华为云MRS... 提升离线及实时分析性能25%。)"

"The models analyze users, front warehouses, order time, rider capacity, weather, product type, and fulfillment cost to assign each order to the best rider and forecast warehouse demand. (original: 通过基于大数据和算法平台训练的模型进行实时推理，将用户、前置仓、订单时间、骑手能力、实时天气条件、履行成本、产品类型等因素进行关联分析... 同时针对... 订单、浏览信息等，构建预测模型。)"

"The program improved big-data job efficiency by 20%, cut overall cost 25%, and increased average monthly orders per user by more than 30%. (original: 大数据作业的计算效率提升20%，整体成本降低25%... 平均每个用户月度订单提升30%以上。)"

Transparency Evidence & methodology 100/100 reference value · 81% confidence · 1 source

Review how this reference was scored and trace each normalized claim back to its documented source.

Scoring methodology

How scores are calculated

Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.

Customer reference value100/100

Problem clarity20/20
Measures whether the customer's operational problem is specific enough to compare.
Implementation depth25/25
Rewards named components, data flow, integrations, and architecture detail.
Outcome specificity20/20
Measures whether the reported benefit is concrete rather than a generic claim.
Quantified impact25/25
8 metrics describe measurable impact.
Deployment scale10/10
Rewards evidence of a live or scaled deployment and meaningful integration scope.

Evidence confidence81%

Source authority20/25
The main evidence is an official vendor-published customer story.
Claim traceability21/25
10 of 12 key fields link to a named source.
Direct evidence20/20
5 direct source quotes retained.
Date provenance0/10
The publication date has no documented provenance.
Source transparency10/10
1 source reference document publisher, URL, authority, retrieval, and supported fields.
Implementation specificity10/10
Rewards source-backed components, data flow, quantified impact, and a specific outcome.

Traceability

Sources & provenance

Field coverage shows which facts each source supports. Dates are shown only when documented by the source.

Huawei Cloud
卓越运营客户成功故事-叮咚买菜-华为云

First-party
https://www.huaweicloud.com/cases/prime-insight-dingdongmaicai.html

Source type
Vendor-published customer story
Supports
- Customer
- Business problem
- Solution
- Products
- Solution components
- Data flow
- Integrations
- technical area
- use case category
- Business outcome
- outcome category
- Metrics
- Architecture
- evidence quotes

Not separately sourced Industry, Region, Company size. Treat these normalized attributes as lower-confidence until a dedicated source is added.

Similar use cases

See all like this

Google Cloud Modified · Oct 9, 2026 Purplle Retail India Purplle needed to support rapid 6x user growth, survive peak sale traffic, and automate search, CRM, pricing, and assortment decisions without increasing infrastructure costs or slowing down retail operations. 10x traffic spikes Marketing PersonalizationReal-time Analytics BigQuery · Cloud AutoML · Cloud Pub/Sub

Alibaba Cloud Aug 24, 2026 Zhiyi Technology (知衣科技) Retail China Zhiyi Technology needed to match millions of product images against millions of overseas social-media assets so merchants could use creator content and trend signals for cross-border product-selection decisions. The… 10 million product-side 512-dimensional vectors Marketing PersonalizationSupply Chain Optimization GPU cluster · MaxCompute · Self-developed CV model

Alibaba Cloud Archived · Oct 24, 2025 ZKH Retail United States As ZKH expanded, new challenges emerged, requiring swift solutions to maintain its competitive edge. To meet its high-resilience demands, the company needed a robust infrastructure capable of supporting a multi-region… By using AI to deconstruct and optimize 16 scenarios and 120 sub-scenarios, ZKH has vastly enhanced operational… Marketing PersonalizationRisk Management Certificate Management Service · Cloud Enterprise Network (CEN) · Container Service for Kubernetes