For data-driven thinking to flourish in your organization, you need to give people easy access to ‘data products’ that will answer their pressing questions.
Easier said than done.
In fact, for most organizations, the collection of dashboards, reports, and analysis tools feels like a chaotic mess. When we worked for a global manufacturer, a survey of information workers revealed that the top problem was an inability to find data products that served their needs. Also a big concern: the quality and usefulness of those data products.
This is the challenge of creating a data product ecosystem. Creating a vibrant ecosystem for data products requires processes and tools. Processes set standards and ensure that the right priorities and qualities are built into every data product. Tools gather data, visualize the results, and distribute data products to users. Here are the six conditions (“the Six Ds” shown below) that are essential to building this ecosystem:
What are the most important areas that would benefit from the insights and guidance of better data?
There is nothing so useless as doing efficiently that which should not be done at all.
We begin with the end in mind. The consumers of data have needs. A healthy ecosystem will support those needs through the right data products. Discovering the information that will best serve the organization is the first step.
Understanding data consumer demand is not a one-time endeavor. It requires a process of continually mapping the important decisions made by the organization and evaluating whether and how data can improve those decisions.
One framework to use: Map the expressed needs of your data product consumers into a matrix to evaluate a) whether the data product will bring real value to the organization; b) whether a solution can truly drive better decision with data. This model will reveal data product concepts with the potential to deliver the greatest impact.
What processes and tools can help ensure the effective design of data products?
Less than 30 percent of the potential users of organizations’ standard business intelligence (BI) tools use the technology today. This low take-up is due to the fact that longstanding tools and approaches to BI are often too difficult to use, slow to respond or deliver content of limited relevance.
The three reasons cited by Gartner for this problem are:
1. Ease of use (“is hard to work with”)
2. Performance (“users are frustrated by delays”)
3. Relevance (“does not express content in line with their frame of reference”)
The first and last reasons link directly to issues of poor data product design.
In our role as dashboard and analytical application designers, this is an area that is close to home. We see it all the time: reports and dashboards that lack focus and a message that targets their audience, which is often undefined. We see poor choices in data visualization that distract from the important elements in the data and put the burden of deciphering meaning on the readers. We see data products that lack an obvious starting point and logical flow to conclusions.
Poor design is wasteful. It results in solutions that users don’t want to use, as noted by Gartner. It wastes the audience’s valuable time as it struggles to comprehend the data. And it wastes the development and distribution efforts necessary to deliver the data product.
What processes and tools support the efficient production of data products, including gathering multiple data sources, presenting this data, providing user customization, and delivering the information to data consumers?
Ideally, you want to have a small set of data tools that support the variety of types of data products your organization needs. A single solution is unlikely to offer the breadth of capabilities necessary. In our experience, four to five tools for data presentation are usually sufficient for most organizations.
There are many forms your data products may take. And for every form, there are many technology options. However, here are some common features that are worth evaluating in almost every case:
End-user customization—Some presentations may target a single audience. This is the exception to the rule. Most often, a data product goes out into the world alone and is used by many people, each of whom comes from a unique perspective. Whether it is their department, region, or products, all audience members will want to see data that is customized and scoped to their situation. Many interactive applications can support this ability to filter the relevant data.
Sharing support—Data should spur conversation. However, some solutions for data products create an isolating environment. The data product should make it easy to share, discuss, and capture insights— whether the discussion happens online, offline, on a desktop, or on mobile devices.
Quality visualization—It matters how data is visualized. Clean, clear charts can make it easy for readers to quickly understand the data. The default settings for data visualizations should adhere to the fundamentals shared by well-known data visualization authors like Stephen Few and Edward Tufte.
Fit workflows—Finally, it is important that data products integrate into how people do their jobs. If the consumer of data is constantly on the run, bombarded by information of all types, an effective data product will deliver simple, narrow content to this person. If the consumer wants to drill deeply into the data to understand underlying assumptions, the functionality should exist to allow for this need.
How can you help people discover the many data products in your organization and find the right information for their individual needs?
Data product discovery should mirror the capabilities of online content subscription services. Podcasts, blogs, or Twitter, all have established features for ensuring an audience can find and access the latest content. These include:
1. Searching of metadata about the content, including title, author, and description
2. Browsing of content sorted into categories and ranked by popularity or ratings
3. Surfacing of related content based on the consumer’s expressed areas of interest
4. Subscribing to allow consumers to sign up to receive updates to content
5. Automated pushing that allows consumers to receive updated content automatically rather than having to remember to return to the source
6. User permissions to control who has access to applications and content
What capabilities encourage data consumers to take the insights they find in the data and share these insights with others?
The best data ecosystems don’t simply assume discussions will occur. They encourage discussions through mechanisms for sharing, capturing, and saving information and insights. The discoveries found in the data are treated as precious assets—after all, they are the purpose of all the effort put into creating data products. Finally, the ecosystem should encourage people to take action when the discussion is complete.
Some organizations consider data products a one-way information broadcast. They implicitly assume that a dashboard is intended to deliver an information result, not drive action.
Discussions on data—like most of data fluency—are more a social and human problem than a technology problem. The technology approaches may be simple. For each data product, create a document or folder for capturing insights. The document may simply be screenshots of the relevant part of the content along with an annotation explaining why it is interesting. As a historical artifact, this document will reveal patterns of common issues and best practice approaches for responding to those issues. More important than a complex technology solution is an organizational culture that encourages dialogue and action after the insights are first found.
How do you filter out the irrelevant content and provide feedback to enhance those data products that remain?
The scourge of data in most organizations is the ever-growing collection of reports that get generated month after month. New reports are created but seldom killed. The mass of data products quickly becomes difficult to navigate and the right information is hard to find. Even for the data that has found a suitable audience, there is seldom a feedback loop. The direction and needs of the organization may change, yet the content does not change to fit evolving demands.
Data products should be living documents. They should improve over time or be removed if they are no longer relevant. It is a matter of survival of the fittest.
Data fluent organizations recognize that too much content—particularly data content—will clog up the channels of communication. The data products must be distilled to the essential information. You want to clean out distractions and emphasize the most useful remaining parts.
To filter and clean your data product ecosystem, you need processes in place to gather feedback from your users. The feedback needs to impact how data products are designed and produced. There are at least three ways to continuously distill the best data products:
Create a lightweight feedback mechanism, like a simple “star” system.
Track usage both on the volume of usage and the levels of engagement.
Conduct content reviews by gathering the audience for a product and having a focus group-style discussion.
Because knowledges are so specialized we need also a methodology, a discipline, a process to turn this potential into performance. Otherwise, most of the available knowledge will not become productive; it will remain mere information. To make knowledge productive, we will have to learn to connect.
—Peter Drucker, Pro-Capitalist Society
It is hard work to create an environment that enables the creation of high-quality data products, ensures those products get into the right hands, and has mechanisms for self-improvement. But without doing this work, all the data investments your organization makes will struggle to reach the important decision-makers and consumers of that data.