Data Science Project : Approach
Organisations globally are filled with Project Managers, and they come equipped with several Certifications, however the traditional project management relies on requirements and careful planning. Key Ingredients are :
- Scope
- Cost
- Schedule,
Till now, Project management has been very successful in many organizations driving down cost, adhere to schedule and keep to Scope. But will this approach work for an exploratory Data Science Project : The Data science team are looking for new insights ..there is no point wasting time looking for existing insights.
New Solution requires an empirical approach , leverage Design thinking and build an agile solution in Sprint mode ! One of my recent Solution, a prototype was developed in a joint workshop and we used design thinking to sharpen the vision.
Data science is an empirical process : While implementing a data science project in South Africa for a bank, while we were focused on getting an outcome about customer behavior while doing mobile banking in big cities, we realised that people in rural/ countryside were making more mobile payments, accessing bank accounts on mobile. New insight allowed us to collect digital documents through mobile from customer in far fetched areas, so if I would have boxed the team to think in a particular way then we would have not reacted to the data as we did. You need to expect the unexpected if you want to gain new insights. If you knew exactly what you'd find, then you wouldn't be gaining any new knowledge. In general, data science looks for new opportunities or tries to address current assumptions. It focuses on research and delivering insights. Project management handles ideas that are already understood and wants to deliver a product. Think about the things that you do in your life that are more exploratory and empirical.
Project management discipline have been very beneficial to many organizations, however Unfortunately Project management discourages uncertainty and hence will not be beneficial to Data Science team. It forces the data science team to only try and verify what's already known. If they find anything unexpected, then it's seen as a bug and not a feature. When you create milestones and deliverables, you're telling the team that they have a set time to verify what's already known.
The language of most organizations is driven by key metrics : What is mission, objectives, and it is always focused on outcomes. It's difficult to step back and imagine a team of pure exploration. For most organizations, working with a data science team will be a difficult transition, so let's look at a typical project and compare it to that of a data science team. Then let's imagine what would happen if you started applying planning and objectives. Let's start with a typical software project.
Last year involved in developing a robotics led automation project for wealth Management customer. The project charter was to create an interactive persoanlised experience which allowed customers to decide on key investments before they speak to Relationship Managers. The project had set budget, outcome was number of hits, savings from eliminating People in local branches and ROI was great. The plan lays out all the features in a requirements document. There's an estimate of the development schedule and all the costs are laid out. All these are outlined in the project plan.
The project manager will own and update the plan throughout the project. They will also help balance out any changes....Sounds Familiar. Now lets switch to a data science team. It's a small team of 3-4 people led by a research lead, 1 or 2 data analysts, and a Project Leaders. The key task is explore, research the customers' needs and behaviors.
The goal is to make data science very actionable, and increase value or revenue. The research lead starts by asking a few questions. What do we know about our customer? What do we assume about our customer? Why does our customer shop with us instead of our competitors? What might make our customers shop with us even more? She subsequently works with the data analysts to break these down into reports. Maybe they create a report of the customers' income.
They measure their success by ignoring insights and focusing on what they already know. That's the opposite of what you want for your data science team. You don't want to think of data science as a project. Instead, try to focus on the discovery. The more insights you find, then the easier it will be to create real value
