PS 2nd Week

• Feature Engineering | External Data

• Code Review (Clean, commented, etc.)

• Train and test ML Model

• End-to-end Pipeline

• what is the performance of my model? E.g. confusion matrix

• task of the data translators

• translate data into comprehensive insights

Questions to prepare the pilot:

• What do you want to know? What are my goals with this pilot?

• Who and how many are the end users? What’s the best way to provide them with the information?

• How to best define the pilot groups?

• How long should the pilot be?

• what is the impact of the solution rollout

• how to fully integrate s in operations?

Stakeholders influences by rollout:

• Comunications

• People

• Processes

• Technology

• Client/ Other Stakeholders

• from raw data to impact

problem -> data -> analysis -> solution -> actions -> impact

• We perform analysis, using the data, to inform the actions that will allow us to achieve the goals, thus having an impact in our problem (!!!)

Model monitoring: monitor …

• … incoming data

• users feedback

• evaluate need for training

• model redeployment

• keep version control

1. Clear problem definition and measurable goals (what is being solved, why it matters, how success is measured).

2. Strong stakeholder engagement and committed project champions.

3. Early validation of data availability and quality.

4. Iterative prototyping with user feedback and pilots.

5. Scalability in mind — solutions should serve multiple stakeholders with flexible visualizations.

6. Trust and adoption — the best model is useless if decision-makers don’t use it.

7. Sustainability — plan for model monitoring, retraining, and long-term improvement.

Problem: MMR vaccination rates have beendeclining dramatically

Goal: Predict which children are at risk of not receiving the MMR vaccine so that healthcare providers can proactively

intervene to promote vaccination

• Data Lifecycle and Ownership

• Better understand the data quality and

  specific features, to better plan the data curation phase (e.g., Data may be in a different language)

• Only “lock” the scope after auditing the

  data/ Problem definition

• Change Management: Product/ prototype

  showcase for Doctors buy-in

• Internal buy-in: The importance of

  strongly committed Partners/ Project Champions

• Scale the solution: Try to serve as many stakeholders as possible with the same solution and different visualizations

1. What is the challenge you are trying to solve? What are the business processes impacted?

2. What is the opportunity we are trying to capture?

3. What are the main numbers about that challenge (e.g., How big is the problem?)?

4. Why is it a problem?

5. Has the partner tried any other solution before? If so, what failed? What were the lessons learned?

6. If there is already a solution in place, how good is it?

7. What are the consequences/ negative impacts of not solving the problem?

• Data science solution

• Goal of the model/ system

• What/ who is being targeted

• Complement with industry/ area

• Add relevant details

Governance

• Resources allocated only to projects within their silos

• No view of analytics activities or priorities outside their function or business unit

Location

• Analytics are scattered across the organization in different functions and business units

Project Management

• Little to no coordination

Pros (+):

• Early-stage Data Science

• More domain focus

Cons (–):

• Difficult to hire / define career path

Governance

• Stronger ownership and management of resource allocation

• Project prioritization within a central pool

Location

• Analytics reside in a central group

• Serve a variety of functions and business units

• Work on diverse projects

Project Management

• Coordination by central analytic unit

Pros (+):

• Encourages career growth

• Cohesive group

Cons (–):

• Chance to disconnect from business lines

Governance

• Better alignment of analytics initiatives and resource allocation to enterprise priorities

• No direct operational involvement

Location

• Analysts are allocated to units throughout the organization

• Their activities are coordinated by a central entity

Project Management

• Flexible model with balance of centralized and distributed coordination

Pros (+):

• Balanced

• Highly coordinated

Cons (–):

• Demands more flexibility / open mind to implement a “non-traditional” model

• Ensure the alignment of activities with the objectives of the specific program/ area

• Guarantee that the project is aligned with the respective company stakeholders

• Define processes and best practices in terms of modelling

• Challenge modelling results and evaluation

• Validate projects in terms of fairness and bias

• Monitor squads regularly on the most critical issues

• Define processes and data architecture

• Monitor squads regularly on the most critical issues

• Promote the agile methodology and the respective best practices in the squads

• Support the application of the methodology in each project

• Unlock methodology-related topics