What is your company’s “Decision Intelligence Quotient”?

Not too long ago, big data craze was all over. “Big Data is the new oil,” it was proclaimed. If you tame and mine big data, you will have a definite edge over the competition. Inspired by this collective wisdom of the time, many rushed to amass swaths of data, dumped data into a data lake, and even made it available to users by sticking self-service BI interfaces.

This newfound capability was rolled out with much fanfare and with heightened expectations of business benefits. Shortly after, but before tangible business benefits started to come, murmurs of tongue in cheek warnings began to float in the marketplace — it’s not the data that matters but what you do with the data. For that, one needs to have competency in AI/ML tools and technologies to complement the Big data competency.

To protect investments in big data and the strong desire to stay ahead competitively, companies went ahead to build AI/ML competency, acquired AI/ML platforms, and sourced data science teams.

Having acquired big data and AI/ML competencies, many, however, are still waiting to realize the promised business benefits. Sounds all too familiar! What went wrong?

Business benefits accrue by making the right decisions at the right time. When there is a struggle in realizing business benefits, most likely, the company is lacking in “Decision Intelligence” — a critical piece in the business value realization recipe. If that is so, the company must work toward raising their DIQ (Decision Intelligence Quotient) score.

You have a request to host a dinner party for 20 VIP guests. The expectation is to put the best show and leave a lasting impression on the guests. You brought in all ingredients for cooking, got all gadgets in the kitchen, and acquired all recipes. You have all the cooking intelligence but still quite apprehensive about the success of the event. That’s because you are not able to decide what to cook as there are many possibilities, and you have the slightest idea of who the guests are and what their individual preferences are. You have a fixed budget and limited time. It was evident that “Cooking Intelligence” is not enough, and it is not the same thing as “Putting-together-a-successful-dinner Intelligence.”

Defining Decision Intelligence

Like most enterprise-level capabilities, it is not easy to come up with a formal definition of “Decision Intelligence” that captures all its facets. It all depends on from which perspective one is coming from to have a grasp on this emerging category.

Strategic Perspective

Essential Questions: Is “Decision Intelligence” a new capability or merely a re-branding of an existing business capability? Why should an organization acquire Decision Intelligence capability?

From a strategic perspective, decision intelligence is the efficiency with which an organization can turn their experiential and empirical knowledge into general best practices for making effective business decisions.

As all organizations do make decisions at various levels, decision intelligence capability is intrinsic to all organizations, and thus, it’s not a new capability in that sense. What is new, however, is the need for enhanced maturity and sophistication in this capability. In the current environment demanding higher business agility, it is crucial to make optimal decisions on time consistently. Measures of decision timeliness and decision quality assume paramount importance as these directly impact overall business performance.

At the operational transaction level, where decisions are required continuously, and within seconds, decision making efficiency is only possible through complete automation of decision-making. Examples of such decisions include product recommendations during online shopping, and online approval of credit cards. Timely guidance of choice takes precedence over the goodness of the decision in most such cases.

As we move up higher in the decision-making hierarchy to operational management planning and control levels, the risk and reward associated with decisions tend to be much higher than operational transaction-level decisions. Furthermore, there is a higher degree of uncertainty and imperfect knowledge of the business context. Thus, the involvement of human decision-makers at this level is unavoidable, which comes with its limitations. Examples of such decisions include decisions on launching new products, deciding loan approval policies, selecting consumer segments for A/B testing, approving new parts suppliers.

Because of the pressure to achieve timeliness objective, the decision quality deteriorates as a result of a trade-off between the two goals. Decision quality can also suffer due to a slew of other reasons such as variations incompetence of human decision-makers and decision fatigue. In such cases, equipping decision-makers with automated decision intelligence support goes a long way in achieving both timeliness and decision quality objectives.

Technical Competency Perspective

Essential Questions: Is “Decision Intelligence,” a new technical discipline? How does it differ from Data Science? Can I retrain my business analysts and data scientists to acquire the “Decision Intelligence” capability?

As a technical discipline, one can think of decision intelligence as the augmentation of data science with behavioral and managerial sciences.

A data scientist’s expertise is in decoding the mechanisms underlying a system, people, or process behavior from the digital footprint of past behavior into an analytical model. This model is then used to predict future behavior.

A programmer wraps the model by an application code to create a prediction service. For many scenarios where prediction itself is the decision, and development of an analytical model is possible using AI/ML techniques, decision making can be fully automated. Prediction service can be stand-alone for direct use by human users, or it can be invoked from a task within a computerized business process, as in RPA (Robotics Process Automation).

For complex multi-criteria decision-making scenarios, one or more prediction services generate inputs for higher-level decision intelligence support. For example, a data scientist may develop a model for predicting sales lift if you target a specific marketing promotion to a particular consumer segment. Data scientist is not concerned with the actual decision of rolling out the promotional campaign. That decision is beyond the scope of the data scientist’s responsibility. Another business role owns this responsibility who maintains knowledge of the business context — marketing budgets, marketing channel and program options, marketing schedules, and is responsible for the overall effectiveness of marketing effort.

Traditional business analyst role acts as a bridge between business decision-makers and IT, by generating reports and analyzing historical data for the consumption of the decision manager. To provide higher-level decision intelligence support, business analysts will have to acquire additional expertise in data science as well as in decision science.

A decision scientist is concerned with the enumeration and evaluation of possible decision options and understanding the risks and rewards of trade-offs among multiple objectives and is well-versed in decision modeling and optimizing techniques.

As data scientists and decision scientists are hard to come by, by providing automated decision intelligence support to business analysts and business decision-makers, one can mitigate to a large extent the talent issue in creating an in house decision intelligence competency.

Decision Automation Perspective

Essential Questions: What automation support is possible for enhancing “decision intelligence” capability? How does decision intelligence automation differs from data science automation?

A data science automation platform provides support for the development and deployment of analytical models using Machine Learning and AI techniques. Decision intelligence automation goes beyond analytical models to provide support for decision modeling, optimization solvers, decision process orchestration, and business context management.

As outputs from prediction services serve as input to decision models, architecturally “decision intelligence” automation can be considered as a higher-level services layer on top of the “data science” automation layer. Decision Intelligence services are accessible to business analysts and business managers via a decision process orchestration interface.

Decision Intelligence Quotient (DIQ)

As “Intelligent Quotient” is a measure of an individual’s problem-solving capability in comparison to other peers, it is possible to assess the decision making capability of an organization by defining an analogous standard “Decision Intelligence Quotient (DIQ)”.

Decision Intelligence capability pertains to the ability of an organization to make timely and high-quality decisions consistently. The quality of a decision is high if acting upon it results in the best possible outcomes for the organization.

Decision Intelligence Quotient (DIQ) is a measure of an organization’s decision intelligence capability relative to an average baseline capability.

Improving Decision Intelligence Quotient Score

Improving the DIQ score is tantamount to improving business performance. Here are vital high-level steps for enhancing the DIQ score.

Step 1: Establish a baseline DIQ score

A baseline DIQ score must be established first, before one ventures out to start specific improvement actions. Determination of DIQ baseline requires administering an internal DIQ self-assessment test for the business process, business unit, or the entire organization within the scope of the DIQ improvement initiatives.

Pre-requisite for the DIQ assessment test is the historical record of past decisions and their quality. If the organization currently haven’t instituted the practice of collecting such data, it must be inferred from business performance data and by filling out surveys by operational decision-makers. For the future, a method must be put in place to record decisions regularly in a “decision intelligence” system repository.

Step 2: Identify decision areas for improvement

Few actions may cause an across the board improvement in all decision areas, but most improvement actions will be highly specific to individual decision areas. This step involves categorizing and ranking decision areas by performing a cost-benefit analysis of improvement.

Step 3: Identify Strategies for DIQ improvement

Here is a broad classification of available strategies for improving DIQ:

• Training of business analysts/decision-makers in best practices to mitigate factors that cause deterioration of decision quality.
• Improving decision-making processes to eliminate bureaucratic or data access delays
• Enhancing decision intelligence automation support by using decision intelligence platforms that allow the deployment of AI/ML/Optimization methods at scale

Step 4: Implement DIQ improvement actions

Implement DIQ improvement actions identified in step 3. Establish time frames and mechanisms for measuring decision performance and re-assessment of the DIQ score.