Release Success Prediction 360, referred to as R360, is an asset for Scrum masters, Product and Project Managers who are managing large teams and want an insight into what is happening with their Deliverables/Releases. - And my first ever product [Time to ship - 1 year, June 2021 - May 2022]

1. The idea

R360 is part of Accenture’s proprietary SaaS platform MyWizard. MyWizard has an history to infuse intelligence into business processes across the organizational systems. For R360, the key users being targeted now are Scrum Masters, project and product managers managining large releases of huge scale in question. R360 was conceived to be eyes, ears and brain for the users so that they can manage they can focus on issues that are cruicial.

2. The Journey: Building the Product

By the time I joined the team, the Visual design was laid out and the product was about to be built. Started with the ground level with writing user stories in a way to effectively comminucate the features. Given my AI background, slowly started getting into figuring out solutions to problems on how to implement some of the key features. For example, I was able to help solution build Story viability predictions - To predict whether a particular story is going to be delivered in the release. There were high number of use cases enabling connecting and stiching across the entire Software Development Life Cycle. I got introduced to the Agile practices, concept of Continuous Integration and deployments, managing backlog, features prioritization etc.

3. Lessons Learned

  • Iterations are essential: Early designs and features were over-complicated. Simplifying the initial designs, and iterating through design helped gain quick wins.
  • Not all use cases need AI: The team initially was attempting to build complex AI models for simple use cases. Simplifying the models helped not only with the use case success but also made go to market possible soon.
  • Data issues Data for building projects was literally non-existant. So generating synthetic data and building models on that was
  • Stakeholder Buy-In Matters: Engaging with scrum masters and release managers throughout development helped tailor R360 to real-world needs.

4. Some interesting features

  • Success Probability Score: A predictive score for each release, based on AI analysis of past and current sprint data.
  • Root Cause Analysis: Insights into key risk factors affecting the release success.
  • Team-Specific Recommendations: Tailored suggestions to mitigate risks and improve future sprints.

5. What I’d Do Differently

If I were to build R360 again, I’d:

  • Focus more on real-time insights from the beginning rather than starting with batch processing.
  • Spend additional time refining the user interface for better usability.
  • Invest earlier in tools to automate model validation and monitoring.

Snapshot of the Experience

  • Inspiration: Simplify release management for large-scale scrum teams.
  • Timeline: 12 months from ideation to deployment.
  • Biggest Challenge: Ensuring model transparency and user trust.
  • Proudest Moment: Seeing R360 reduce release failures by 20% in pilot teams.
  • Future Plans: Expand features to predict long-term project success and integrate sentiment analysis.