Create an analytics platform strategy in 5 stages

[#1]

Imagine you are responsible for the analytics platform in a bank. It has evolved over years and now supports hundreds of users performing analysis every day across a range of business functions – marketing, credit risk, fraud etc. The platform is delivering value but is struggling to cope with recent demands. Some business areas have been asking for new functionality which you have not been able to deliver. Users have complained the performance is poor and there have been some outages. You know the bank’s analytics platform is no longer fit for purpose and needs to be brought up to date. You need a strategy and fast. What do you do?

Having seen something like this a few times, I would suggest a five-stage plan. This is based on my own experience of working with several large organisations.

1. Assess current state
2. Align with other strategies
3. Create vision
4. Gain business support for use cases
5. Develop roadmap

Stage 1: Assess current state

An assessment can be a quick exercise involving no more than a week of workshops followed by a week or so to write everything up. Each workshop should be focused on one business function and involve up to 10 representatives. It’s perfectly feasible to do this online and it has the benefit of bringing together people who are geographically separated.

To get full coverage you need to consider all the business areas using the analytics platform. These could be mapped in a diagram. The following is a fictional example but based on a real bank:

Figure 1: Output of Assessment stage

In each workshop try to gather information on the following:

• Functions and capabilities: What does the business function do? How does it use analytics? What analytical platform capabilities does it need? Are there plans or a vision for analytics?
• Experience and issues: What’s going well? What’s not so good? Is there anything limiting the business function? If you could change something what would it be?
• People and process: Who uses analytics? What are their roles? How do they operate?
• Data: Where is it sourced? Is it reliable? How is it managed?
• Architecture: What software and tools are used? Is the documentation up to date?

Capture all this information in the simplest way possible. I have found PowerPoint slides with bullet points and verbatim quotes are quite adequate. If team members can share diagrams, these can be included in the final output.

You don’t need a full-blown analytics maturity assessment. These were popular a few years ago but I now rarely see them used. Apart from the time and effort, the concept of putting your organisation’s capabilities at a point on a scale can be overly simplistic. It could also be argued that a maturity model is inappropriate for data science.

One significant benefit of conducting an assessment is the engagement with the business functions. You are involving stakeholders in the process. These are the people with influence who will support your strategy and help drive through changes.

Stage 2: Align with other strategies

Analytics platform strategy needs to be seen in the context of other strategies.

Figure 2: Alignment of strategies

Looking at this diagram horizontally, each strategy is a way to get from the current state to a vision of the future state. In Stage 1 you assessed the current state of the analytics platform and in Stage 3 you will create a vision of the future state. In Stages 4 and 5 you will define the path to get there.

Now looking at it vertically, you will see that the analytics platform vision and strategy need to be aligned with similar strategies at different levels in the organisation.

Starting with business vision, look over any corporate presentations or reports you can find and note any statements of business goals. Understand where the organisation wants to be in the market and its future focus in terms of products, customers or geographies. How will it differentiate from the competition? Will the business model fundamentally change? The business strategy will describe how it will achieve the vision and objectives, including business principles, priorities and programmes. It may also explain the drivers – how the business environment is changing and why it needs to adapt.

Data and analytics strategy covers the wide range of organisational, process and technology capabilities required to utilise data effectively for analytics and support the business strategy. You may be able to note points in the following areas that are relevant to the analytics platform stategy:

• Operating model – Analytical Centre of Excellence, data science team, ModelOps
• Architectural principles - e.g. “data is a shared asset”
• Data governance – data quality, metadata, data lineage
• Technical – EDW, Hadoop, cloud

IT and cloud strategy could have huge implications for the analytics platform. With data moving to the cloud, analytics has to follow. Your analytics platform strategy must recognise the changing IT landscape and address issues of cloud services, data movement and application migration.

After reviewing these strategies you will have a long list of quotes and statements. Consider how the analytics platform is impacted by each of these and what changes or new capabilities are needed.

Stage 3: Create the Vision

In developing your analytics platform strategy, you will have examined current state and related strategies. It is now time to create a vision that brings it all together.

The vision will be in images and a common vocabulary that everyone can relate to, both business and technical. It needs to have just enough technical detail to be meaningful while still being high level and business-friendly.

Here is an example based on a presentation I gave to a financial services organisation:

Figure 3: Output of Vision stage

In this example:

• The analytics platform is positioned on top of the data platform
• An icon indicates which elements will be implemented on cloud or on premises
• SAS Grid runs a wide range of existing analytical processes which will be migrated from legacy servers.
• There are analytical business solutions specifically for fraud and AML
• A spectrum of analytical tools is available to suit a range of users from report designers/consumers through citizen data scientists to those comfortable with coding in open source.
• Model management handles the process of deploying and monitoring analytical models in operations.
• Customers interacting on digital channels will receive automated decisions (e.g. to a loan application) through APIs
• Models and rules can be triggered through an API to provide an automated decision, such as when a customer applies for a loan

Stage 4: Gain business support for use cases

By now you may have nods of agreement from smiling stakeholders. They like what they see but is it a realistic vision? Are they prepared to pay for it?

One way to get buy-in is to examine the use cases or business scenarios that the new analytical platform capabilities support. Focus on those that are prioritised in the business strategy or are needed for regulatory compliance. Others that may be categorised as ‘discretionary’ are harder to justify.

For a business case you need to work with business stakeholders to evaluate the financial benefit of the use case, the risk of not implementing (e.g. losses or fines) and any cost avoidance through decommissioning old systems.

Stage 5: Develop roadmap

The last part of the strategy is to describe how it is going to be achieved. It’s not going to happen all at once so you need to a define a roadmap of stages or transitional architectures. The order of these will depend on dependencies, such as data provisioning, and business priorities. If you are migrating existing analytical applications to the cloud then consider Gartner’s Five R’s: Rehost, Refactor, Revise, Rebuild, or Replace. The roadmap must deliver benefits as early as possible while building progressively towards the final state.

Did you succeed?

All the techniques describe above are tried and tested. I have used them myself.

If all has gone well, you will have created an analytics platform strategy that address current issues and supports the business in achieving its goals. You will have engaged all stakeholders and got their buy-in. This will give you a firm foundation from which you can scope out individual projects, secure funding and start implementing.