Visual analytics is a terrific software category that is essential to delivering business benefits from big data investments. Exciting business opportunities demand that projects to deliver visual analytics applications be successful.
Here’s a discussion of the common landmines that project managers must defuse on the road to visual analytics success.
1. Vendor pre-sales demo risk
The vendor pre-sales demo generates stratospheric expectations by ratcheting management and end-user enthusiasm through the roof. The demo shows visually stunning results. Supposedly these can be created by anyone (not true) to solve real business problems after an incredibly short development time measured in hours.
This completely unrealistic demo creates a serious risk of expectation mismatch for the project. No one remembers that the demo ignores data quality issues, the widespread inability of capable end-users to state requirements, as well as project cost and schedule realities.
Ideally you, as project manager, can avoid holding the demo altogether and rely on the project team demoing visually appealing results from its own prototype development. Failing that, you can stand up at the beginning of the vendor demo and outline the substantial differences between what will be shown and the work that lies ahead for the project team.
2. Technologists imposing scope
Sometimes the data warehouse team wants to jam a bunch of deferred upgrades and some ETL refactoring into your visual analytics project because you obviously have approved funds.
Sometimes the data scientists want to develop arcane statistical analysis techniques that they can barely explain and that your end-users don’t understand.
As project manager, pointedly ask the technologists to link the need for their proposed work to the business goal of the visual analytics project. If you’re not satisfied, say no. As defender of the project scope, you hold the ultimate decision-making authority.
3. Data management neglect
Visual analytics, as you might expect, is critically dependent on adequate data. Inaccuracies and gaps in the data will make your visualizations look bizarre even to untrained eyes. Stakeholders tend to see that horrible result as a project failing rather than as a data management shortcoming unless you, as project manager, explain otherwise.
At many organizations, the data management processes are insufficient to deliver the data quality your project depends on. This neglected state of affairs exists because applications owners are often measured by low operating cost and not the quality of the data they manage.
As project manager, to ensure your visual analytics project success, you will need to encourage the organization to fix at least some of these data management processes.
4. Cadillac DBMS
Sometimes the IS department wants to acquire more Database Management System (DBMS) toys at project expense without a really solid benefit case. In the trade press and at conferences, there’s a raging debate about using Hadoop or a data warehouse DBMS for large analytics projects.
For most visual analytics projects, Hadoop or its competitors are over kill. The in-place DBMS software is sufficient to achieve project goals.
As project manager, pointedly ask the IS department to link the need for a new DBMS to the business goal of the visual analytics project. Only agree if the added cost, schedule and risk are essential to your project.
5. To move or not to move data
Typically, some of the data your visual analytics project needs to access is stored in operational datastores associated with applications. The rest of the data will be stored in a data warehouse that is populated from various other operational datastores.
Inevitably, the architectural purist will want to talk you, as project manager, into an elegant technical solution that expands the data warehouse by adding in the rest of data from operational datastores.
The associated development work and operational complexity will add considerable cost and schedule to your project while only improving visualization capability and query performance slightly. By contrast, accessing data where it sits costs your project almost nothing.
Can you share any experiences from running a visual analytics project to deliver business benefits?