Besides the basic information of the decision, it is worth understanding the characteristics of each decision. Important characteristics include things like volume, timeliness, consistency, time to value, value range, and degrees of freedom.
While all the decisions being considered for Digital Decisioning systems are repeatable, there is a significant difference between a decision to assign a particular team to a call center shift for the week and a decision to alert a network operator that a piece of equipment is at risk of failing. The shift assignment decision may be made every week or 52 times a year, but the alert/don’t alert decision could be made every second for thousands of pieces of equipment—amounting to tens of billions of decisions each year. The volume of a decision constrains how a Digital Decisioning system automates how a decision should behave. Volume also plays a key role in determining whether any human can be involved at all.
The volume of a decision is paired with its timeliness. Generally, the more often a decision is made, the less time can be spent making the decision. It is worth noting the timeliness requirement of a decision separately, however, as some low-volume decisions can have short time windows in which they can be made. Responding to an emergency signal might not be a decision that happens very often (most years, it may not happen at all), but when it does, the decision as to which equipment to shut down must be made nearly instantly. Timeliness will act as a major constraint on the behavior of a Digital Decisioning system.
Consistency Over Time
Some decisions remain remarkably consistent over time, with only slow incremental change in the decision making approach being required. Others are in a constant state of flux, with new regulations, policies, or analysis being incorporated all the time. Even within apparently very similar decisions, there can be significant variation. Pricing decisions, for instance, range from product pricing guidelines that are updated once a year to pricing in dynamic markets, where prices are updated every day based on competitive situations or deals with suppliers. You may be wrong about how consistent a decision is going to be in the future, but an initial assessment of the extent to which the decision-making approach will remain consistent over time will help you design an appropriate Digital Decisioning system.
The difference between the value of a good decision—of the best decision possible—and a bad one varies tremendously. Some decisions offer only small opportunities for improvement. A bad up-sell decision is of zero value, while a good one increases the profit on a transaction by a small amount. A bad choice in subject lines means an email won’t be opened, while a good choice means it will be opened. For these decisions, the value range is small. Other decisions have a much larger value range. The difference in value between a good loan origination decision that might be worth a few tens of dollars in fees and interest over and above the intra-bank lending rate, and a bad decision where the whole loan has to be written off is thousands of dollars. Multiply that by thousands or tens of thousands of loan origination decisions, and the total value runs into millions. The value range for a decision makes a big difference to the kind of Digital Decisioning system you end up building, in particular to the degree to which predictive analytics can and should be part of the system.
Time to Value
The impact or the value of some decisions is felt immediately. The decision of which ad to display to a website visitor either works (they click on it) or fails immediately. In contrast, a loan origination decision won’t show a positive or negative outcome for some time—not until you see if the borrower will make payments reliably. When the time to value of a decision is short, you get immediate feedback on how well you are making the decision. This allows for rapid experimentation and highly adaptive Digital Decisioning systems. In contrast, a decision with a long time to value must be monitored for an extended period, and any experiments are going to have to run for a while so you can gather the data you need to evaluate them. How quickly your decisions will have an impact on your business is worth knowing for all the decisions in your inventory.
Degrees of Freedom
Some decisions are highly constrained by the policies and regulations that must be followed. Once all the rules for eligibility for a benefit are followed, there may be no choice remaining—either the person applying is eligible or they are not. For such a decision, there will be little or no analysis and no reason to experiment. Other decisions have a large amount of freedom. There may also be uncertainty in judging the success or failure of a decision. Judgment, analysis, and experimentation all play a role in determining how these decisions should be made now and in the future. Understanding the amount of freedom and the nature of that freedom is helpful when designing Digital Decisioning systems that automate the decision.