Using DIKW models for intelligent continuous improvement in software delivery

Register now

Generating business insights and value from data is the latest currency of business interest. By implementing effective systems to turn software delivery data into business intelligence, organizations can gain significant competitive advantage.

Yet reaching this level of maturity isn’t easy. In a survey of 64 of c-level technology and business executives representing large corporations such as American Express, Ford, General Electric, General Motors and Johnson & Johnson, 69% said they have not created a data-driven organization. The DIKW Model, however, can help them wade through the sea of data to glean the wisdom that can impact real change.

This quest typically starts with a data lake project, whereby all structured and unstructured data is collected in one big repository (or “lake”). The aim is to then to apply analytical techniques to obtain insights to optimize specific systems or processes. Extracting the data from the many tool repositories that plan, build and deliver software into a single data store is often the first step, but the volume of raw data that gets extracted can quickly become an overwhelming abyss, and the promise of making this data useful has rarely come to fruition. That’s where the DIKW Pyramid comes in.

The DIKW Pyramid (sometimes called the DIKW Hierarchy) is a knowledge management model that describes how raw data can be processed and transformed into information, knowledge and wisdom. What makes data so challenging and complex is that, by itself, it doesn't do anything; it is the base of the pyramid with the least amount of perceived value.

Information has a higher value than data, knowledge has a higher value than information, and wisdom has the highest perceived value of all.

Data is simply a meaningless set of entries, logs or records that has yet to be interpreted. It is raw, unorganized and unprocessed and without context. Without context, data doesn't answer any questions or allow us to draw any conclusions. Companies that are using a data lake as a starting point are now struggling because, without any organization or structure, the data just continues to pile up and ultimately is never used. Data requires a system to process it to move up the value chain to become information.

Information is data with meaning; it takes us higher in the DIKW Pyramid. You get information when you start to make data useful. To do so, organizations need to apply systems to collect the relevant data, organize and process it. Such systems need to extract and classify the data and apply different modeling techniques to visualize the relationships in the data. Some examples of information might be the units of value (features, defects, risk, debt) that are flowing through the value stream at certain speeds and efficiencies, or the number of items being actively worked.

Knowledge involves learning how to turn information into a form that can improve decision-making. Interrogation is carried out by asking “how” we can apply the information to achieve our goals. If faster time to market is a business objective, it would be useful to understand your actual time-to-market. Having this knowledge would allow for experiments that uncover what is slowing the system down.

So, taking the information and running experiments to get feedback gives us knowledge about areas such as how a change impacted our system, did we address the right constraint or bottleneck, and so on. Trending data helps to analyze the information further in a manner that produces knowledge, so it’s important to collect, combine and store historical data from the systems that help plan, build and deliver software.

Many SaaS tools, for instance, purge their data after a year, so it is crucial organizations have a means to continuously extract data into a centralized point. Trend analysis techniques can explore the cause-and-effect relationships underlying a problem. Flow graphs formulate and plot trending information on a dashboard, allowing you to visualize trends that are not necessarily explicitly stated as information.

Flow metrics dashboards are an advanced way to enrich the information. For example Flow velocity, a measure of productivity that tells you how many items were completed over a given period of time (week or month over month), provides information into whether value delivery is increasing. This information may prompt an investigation into the root cause of a noticeable change in velocity -- for example, a new process, change in workflow or staffing change.

Wisdom is the topmost level in the DIKW pyramid and can be thought of as the process by which people can use knowledge to act; for example, to implement an improvement decision or fix a problem. Wisdom enables us to move beyond speculation and assumptions to:

  • Analyze system performance
  • Reveal bottlenecks
  • Perform data-driven continuous improvement
  • Apply “what if?” analysis and gain feedback

Wisdom answers the questions related to "Why," such as why our system efficiency is trending down. Excessive flow load -- the number of flow items being actively worked on in a value stream, denoting the amount of work in progress (WIP) -- is correlated to inefficiency. By analyzing how flow load and flow time (if your time-value is getting longer) interact with one another, you can identify the ideal WIP limits for a value stream and act to balance demand vs. capacity correctly to improve efficiency.

Another similar approach would be to measure waste time and delays: We know that the lower the flow efficiency (active work vs. wait states), the longer work is stagnating in a wait state. This points to the existence of bottlenecks, inefficient processes, dependencies or lack of resources. We can act to address the constraints and monitor efficiency over time to see whether we improve.

Data-driven cultures can realize higher business returns. Organizations can ascend the DIKW Pyramid by treating data as a critical aspect to drive continuous improvement and accomplish extraordinary things. By capturing objective data from your integrated IT toolchains across teams, tools and departments, you can generate these flow metrics to pinpoint what is slowing you down and impacting your time-to-market.

For reprint and licensing requests for this article, click here.