How Alibaba’s CRM Team Shifted from Waterfall to Agile with a One‑Team Approach

Background

In March 2017 the author joined Alibaba’s Intelligent Marketing Platform CRM team, which supports sales, agents, and advertising revenue. The CRM system handles opportunity management, customer development, contracts, risk control, account management, and financial settlement. Its quality and delivery speed directly affect sales efficiency and advertiser experience.

Initial analysis showed the project followed a waterfall model: two months of development, one month of testing, and a single release, causing long feedback cycles and information loss.

Problems with Waterfall

1. Information gaps due to hand‑off collaboration.

2. Inability to respond to change; requirements changed during development, requiring redesign.

3. Late user feedback; most issues were discovered just before release.

Improvement Plan

The team formed a cross‑functional “One Team” consisting of business, product, and development core members, later adding operations. One Team creates quarterly product roadmaps, aligns goals, and holds bi‑weekly meetings to review feedback, sync progress, and plan the next iteration.

They switched to a two‑week iteration cadence, emphasizing requirement splitting and a Definition of Ready/Done.

Requirement splitting uses user‑story‑mapping to deliver complete user journeys, starting with a Minimum Viable Product (MVP) and iterating.

Implementation Details

Bi‑weekly meetings cover: user feedback on released features, current iteration status, and upcoming core requirements.

Development and testing for the current iteration run in parallel with analysis for the next iteration, reducing waiting time.

Strict DoR and DoD criteria, automated testing, and CI/CD pipelines are introduced.

Results

One Team improved cross‑department trust and decision‑making. Development time dropped from 12 days to under 6 days, and delivery time stayed within 20 days. Quality remained high, with defect‑fix speed and test quality at group‑leading levels.

Frequent releases (every two weeks) enabled rapid hypothesis validation, faster value delivery, and better responsiveness to business needs, such as delivering prioritized data reports.

Challenges Ahead

With limited resources, the team must continue to focus on delivering high‑value features quickly, automate deployment to reduce release overhead, and deepen agile practices.

Conclusion

The One Team mechanism and short, rhythmic iterations successfully transformed the CRM team from waterfall to agile, achieving faster, higher‑quality delivery and strong business recognition.