End-To-End Analytics Use Case Delivery: The Big Picture
Part Two of the Delivering Data Analytics series by OCTAVE will focus on the Planning Phase of a development environment. Part One of this series is available here.
Production Ready Development Lifecycle
For any Advanced Analytics (AA) project to successfully deliver potential business value, it needs to run through a rigorous iterative process with multiple testing cycles during the inception, piloting, roll out, and maintenance processes.
The following flowchart illustrates the stages of a successful AA delivery process.
Key Stages of the Planning Phase
1. Release Planning
Timeline: Start of the Advanced Analytics project
Duration: A day
· Start with epic planning based on business priorities
· Identify the assumptions, risks, and dependencies
· Prioritize the stories allocated to sprints
2. Sprint Planning
Timeline: Beginning of each sprint
Duration: Timeboxed (2–3 hours)
· Prioritize based on busines needs and the team’s ability and capacity
· Manage cross team dependencies such as the use case owners of the business units and the data stewards
· Agree and define the acceptancy criteria
· Define the testing criteria
· Make a commitment based on the team velocity
3. Daily Scrum
Timeline: Beginning of each day
Duration: Timeboxed (15 -30 minutes)
· Identify impediments that could affect the development, such as data issues, platform issues, algorithm issues, parameter issues, etc.
· Communicate with the team and decide on problem solving sessions
· Start tracking the deliverables for the AA project
4. Sprint review
Timeline: Last day of the sprint or the next working day
Duration: Timeboxed (1–2 hours)
· Make sure to complete the sprint planning before the show and tell
· Make sure all tools are updated and the scrum team agrees with the current AA project status
· If the AA project can be demonstrated, conduct a show and tell with the business. In some AA projects every sprint deliverable cannot be a show and tell, hence make it a practice to align with the release planning to identify the acceptable criteria for the show and tell
5. Retrospective
Timeline: Last day of the sprint or the next working day
Duration: Timeboxed (1–2 hours)
· Make sure to get every team member to participate, as every idea will help to groom the team to the best they can be!
· Identify action items and ETAs
· Review the feedback submitted by the Business Unit or SME
· Agree on a process to adopt improvements from next sprint onwards for better deliverables
Although a typical AA use case delivery can be executed in iterative phases, special attention should be directed at end objectives during planning session. Ideally, releases should be organized in a manner that supports the execution of a post inception stage piloting, followed by a thorough retrospective review to assess whether the AA project can deliver its value. Thus, it will be guaranteed that it is technically and financially fit to continuously deliver at the same rate with minimum modifications in future.
To this end, it is important to set evaluation criteria and KPIs beforehand, and track them vigilantly. The product owner of a typical agile project may only engage in a backlog grooming session, but when it comes to AA project delivery, all key stake holders should be in contact during all phases of the project. Make sure to conduct focused and object driven problem solving sessions to maximize the expected output.
Before the planning stage is completed for each iteration, both the Data Engineering and Data Science teams should identify the key business deliverables and meet all required dependencies. Gamification dashboards are essential for tracking, and tools such as Azure Boards/Jira should be used throughout to manage the project delivery and governance.
The next article will be a deep dive into some of the best practices for the AA “CODE” stage, where we will discuss how an Advanced Analytics Continuous Integration Process works.
Written by Duminda Jayathilake, Analytics Delivery Lead.
