Cameron Wasilewsky, Discovery Success Lead, with:
Jun (Blake) Im
Westpac is Australia’s oldest bank and company, one of four major banking organisations in Australia and one of the largest banks in New Zealand. We provide a broad range of banking and financial services in these markets, including consumer, business and institutional banking and wealth management services. We also have offices in key financial centres around the world including London, New York, and Singapore. Westpac Group's purpose is Helping Australians Succeed. It’s what we do, who we are and why we come to work every day. What's most important to us is understanding what success means to our customers and helping them get there.
As Australia’s oldest bank, and one of the major banking organizations in the country, Westpac encountered the typical data science challenges of companies in the financial services industry:
1. Complexity in processes
Being a large, highly regulated organisation, it can sometimes be a complex process to get approval on a business idea and bring it into production quickly. We set out to tackle this issue to improve innovation and the ability to solve complex business problems. We not only had to provide a pathway to move forward with more velocity, but also make it an understandable pathway through dedicated governance and documentation to fit within the industry’s security and regulatory framework.
2. Change in ways of working
Data practitioners at Westpac are comfortable with SQL, which shaped a certain understanding that data comes in tables and products are like reports. It is a hurdle to work through, given that they are comfortable with this way of working, and the key is not only to show them new possibilities, but also to communicate the value they will gain. Learning new tools and changing ways of working is extremely difficult, hence we had to make them see that we could build much bigger products and services, that data can come in any shape or form, and expand the very idea of what data can do – instead of just being an answer to a current question.
3. Technological change
The former tool structure was very simple, but we needed new tools to achieve bigger outcomes, which added complexity. But, relating to the challenges previously highlighted, people are naturally resistant to change and renewing the structure is a very cumbersome process – hence the need for a central data platform to bring everyone together on this journey.
Dataiku played a key role in this transformation, in several ways:
1. New organizational structure leveraging Dataiku as the central data platform
As we implemented Dataiku in September 2020, we created the Discovery Lab which, with only 5 full time and 5 part-time team members, supports 40 business labs and 120+ employees in their data science endeavors.
The Discovery Lab is structured into two teams:
Embodied through using Dataiku as the central data science platform, this new setup helped break down the barriers between business and tech to seamlessly work toward building cutting-edge data products and services.
2. New collaborative, self-service operating model to upskill and drive a closer alignment between the business and tech teams
We also developed a new operating model and new processes to ensure a strong alignment between the Lab and the business teams, while enabling them to broaden their understanding and gain new data skills throughout the project.
When a new team member joins, we plan an orientation session so that they understand how the team can leverage Dataiku. Then they go through the ‘Discovery Suitability Assessment’ to further define their use case, the objectives and expected impact, as well as how they visualize the outputs and outcomes of the initiative. This is key to understand the potential value that can be delivered to our customers through this use case. This enables us to assess where we can help, whether this aligns with our data strategy, and the potential resourcing gaps to cover in order to carry it through.
Adopting this high-level, end-to-end project view also enables us to identify the needs of the business in terms of data literacy, wrangling, and visualization, and train them to develop these new skills. With this practical approach, they immediately see the benefits they will gain in their day-to-day job. Dataiku’s visual interface also acts as an enabler since they understand the whole data workflow, while being able to learn more and dig deeper into the more technical work in one click.
3. Documentation & reporting made easy to efficiently carry out new data initiatives through production
Our main best practice is to document every request and action performed throughout the project for public knowledge – which not only helps with alignment and upskilling but has also enabled us to carry out more and bigger undertakings.
After the ‘Discovery Suitability Assessment’, the overall objectives, key metrics, and outputs are entered in a public Confluence page (example below), as well as the main developments, and any issues or bugs encountered. Conducting each project in this transparent manner helps with setting the right expectations for what the Lab can achieve - being clear on what we’re capable of doing, what we’re working toward, as well as what we’re building and when it will be delivered. It’s critical to build trust with our business counterparts.
Everyone can add and edit, and it’s everyone’s responsibility to keep it updated so that we move forward quickly and comfortably, while anticipating and mitigating any potential risks. Besides the Confluence page, all requests and actions are logged in Jira so that stakeholders are aligned on progress.
Dataiku’s tracking and monitoring capabilities are a critical piece of this equation. The built-in functionalities accurately reflect the work that is being done by the team and all actions performed on the project, in a visual and easily accessible manner. We have also created dedicated tracking projects to automate reporting and gather high-level risk metrics on the projects conducted by all our 40 labs (examples can be seen below) – this saves us precious time on formerly tedious reporting tasks and enables our team to focus on the bigger, more interesting undertaking that will transform the organization.
Dataiku dashboard tracking the number of weekly active users
Dataiku dashboard tracking our tickets and time to address them
We have also implemented a range of continued support for our labs in the form of weekly drop-in meetings, 1-on-1 sessions and Westpac specific video training material. Continued engagement through announcements and our recent showcase competition.
This new organizational structure implemented with Dataiku has brought tremendous benefits to drive data transformation within the organization.
The most striking aspect is the change in perspective, which is critical to enable change. The new collaborative, self-service model enables the tech team to serve the business, help them upskill, and work together to drive innovation at scale. It is altogether a different attitude to a lot of the former processes that were in place!
Breaking down the barrier between tech and the business has been our biggest success with Dataiku. Now we are able to tackle bigger data projects and get them in production state in a much quicker, yet realistic, time frame - whereas many used to be stuck in ideation due to misalignment and tooling segmentation.
As the central data platform, Dataiku enables us to gather all stakeholders around the table, regardless of their profile and data skills – which brings about a diverse mix of perspectives to the project, helps everyone upskill, and eventually leads to developing more innovative and transformative data products and services.
This structure and platform have enabled us to bring our 100+ users and 300+ analytics community members closer to the data, ensuring they are able to ideate and develop data driven insights with the appropriate governance, support, and mindset.