Decision Tree Technique
Decision Tree Technique
Decision tree analysis is a powerful decision-making tool which facilitates the evaluation and comparison of the various options, helps to
choose the most competitive alternative. Decision trees can be used for both qualitative and quantitative analysis and are particularly useful
for identifying the most important variables or factors that influence a decision.
Basically, it is a flowchart that starts with one main idea and then branches out based on the consequences of the decisions. It’s called a
“decision tree” as the model typically looks like a tree with branches.
To create an effective tree the one should, where possible, include quantitative data and numbers. The more data the decision tree has, the
easier it is to determine expected values and analyze solutions based on numbers.
It involves visual outlining of the potential costs, outcomes and consequences. Thus, one can use a decision tree to calculate the expected
value of each outcome based on the decisions and consequences that led to it and quickly assess the best course of action by comparing
the outcomes.
One of the applications of decision trees involves evaluating prospective growth opportunities for businesses based on historical data.
Historical data on sales can be used in decision trees that may lead to making radical changes in the strategy of a business to help aid
expansion and growth.
2. Finding prospective clients by using demographic data
Another application of decision trees lies in the use of demographic data (geographic region, purchasing power, age etc.) to find prospective
clients. They can help streamline a marketing budget and make informed decisions on the target market that the business is focused on.
Decision trees could also be used to assess the risks impact. Data from decision trees can also be utilized to build predictive models or to
analyze an expected value.
The use of a decision tree can help lenders predict the probability of a customer defaulting on a loan or evaluate a customer’s
creditworthiness to prevent losses.
Decision trees can also be used in operations research in planning logistics and strategic management. They can help in determining
appropriate strategies that will help a company achieve its intended goals.
Popular fields where decision trees can be applied include risk management, budget planning, operations, engineering, healthcare,
marketing, business analysis, data analytics & machine learning and of course finance & investment.
Root Node A top-level node that contains information about the ultimate objective or main explored
question.
Decision Node The node indicates a situation when a decision has to be made (also referred to as a
square leaf node).
Chance Node The node indicates possible outcome and is used in the cases where the outcome is
uncertain. Decision can lead to multiple outcomes, but mostly it is recommended to just
lead to two results at once.
End Node (Terminal These is a leaf node that represents the end of the decision diagram. No further
Node) branches are expected from an end node.
Every time a decision is made, it leads to different nodes. A branch would connect these
nodes and represent a situation or a result. Mostly, the result is written over the branch
Branch
in normal text.
Rejected option This is an optional symbol in a decision tree which is drawn after the entire diagram has
been created. We simply mark it over the branch of the option that we wish to no longer
continue.
Drawing a decision tree diagram as a rule starts from left to right but could also be expanded from top to bottom.
Important terms:
Splitting - the term to denote division of the available option (depicted by a node or sub-node) into multiple sub-nodes.
Pruning - denotes the process reverse to splitting, where the decision tree maker can eliminate one or more sub-nodes from a particular
decision node.
Expected Monetary Value (EMV) - the amount of monetary gain that one can expect from a particular decision. EMV can be calculated by
multiplication of the possible outcome (impact) by the possibility of its occurrence. In the case of having multiple risks, the EMV must be
calculated for each of them separately. The result can be either positive or negative. It is positive for opportunities (positive risks) and
negative for threats (negative risks).
Define and specify the problem area for which decision making is required. The first step is identifying the root node. This is the
main issue, problem that you want to solve, question, or idea you want to explore (e. g. decision on whether to launch a new product,
invest in a new market, technology to use etc.). Write your root node at the very top of the decision tree.
Gather and prepare data. Gather the data that is relevant to the problem/issue at hand. This includes both internal data, such as
sales and customer information, as well as external data, such as market trends and industry benchmarks. Identify the factors that
could affect the problem/issue you are trying to solve (e.g. market size, customer demographics, cost). Clean and format the data to
remove any missing or invalid values.
Evaluate possible outcomes for each decision. The next step is identifying & interpreting all possible solutions to the particular
problem/issue as well as their consequences. Continue adding chance or decision nodes after each decision to expand your tree
further.
Put in relevant data with probability values. Once you have organized all potential options, you have to pick every option and
evaluate every possible outcome or result you might come across if you choose the particular option. This step envisages presenting
the variables on a decision tree along with its respective probability values. These values could be monetary, such as revenue or profit,
or non-monetary, such as customer satisfaction or brand reputation. Not all your evaluations need to be accurate. Here you have to
rely on estimates, guesses, and predictions as per the analysis.
Calculate and evaluate the outcomes. Now you need to analyze every potential outcome and the risks and benefits that come with
the given outcome. First, calculate the value for each path beginning on the left-hand side with the first decision and cumulating the
values to the final branch tip on the right side as if each of the decisions was taken and each case occurred. Then, calculate the
Expected Monetary Value, if you are working with monetary value s of course.
The Expected Monetary Value (EMV) of each node will be calculated by multiplying Probability and Impact. Begin with the path values
at the far right-hand end of the tree and then proceed from the right to the left & calculate the value of each node. This calculation is
called “folding back” the tree.
Input data:
The client wants to hire sub-contractor to add a new feature to the platform, which cannot be covered by the current dev team and
has strict timeline to go live. The client also envisages the penalty of $5,000 per calendar day for every day sub-contractor
delivers late.
Sub-contractor 1 charges $250,000 for the development of a new feature. Based on the studied historic data the assumption is
that there is a 30% possibility of completing 60 days late.
Sub-contractor 2 charges $320,000 for the development of new feature. Based on the studied historic data the assumption is that
there is a 10% possibility of completing 20 days late.
PROS CONS
Transparent: decision trees provide a focused approach to decision Complex: decision trees can become complex if you add too many
making. When you parse out each decision and calculate their decisions to the tree. If your tree branches off in many directions, you
expected value, you have a clear idea about which decision makes may have a hard time keeping the tree under wraps and calculating
the most sense to move forward with. the expected values. The best way to use a decision tree is to keep it
simple so it doesn’t cause confusion or lose its benefits.
Efficient: decision trees require little time and few resources to Unstable: it’s important to keep the values within your decision tree
create and present information in such a straightforward way, they stable so that your equations stay accurate. Small changes in the
can be quickly analyzed and used to make crucial decisions. Other data can result in large changes in the tree structure.
decision-making tools like surveys, user testing, or prototypes can
take months and a lot of money to complete.
Flexible & adaptable: decision trees can be easily adapted to Risky: decision tree uses a probability algorithm, thus the calculated
accommodate new ideas and/or opportunities. You can add new expected value is an estimation, rather than an accurate prediction of
decision or other branches for possible outcomes into the tree with each outcome. If you don’t sufficiently weigh the probability and
little work. payoffs of your outcomes, you could take on a lot of risk with the
decision you choose.
Easy data interpretation and classification: decision trees ease Chances of classification errors: less experienced decision tree
out the process of segregation of the acquired data into different maker usually makes a mistake while putting the variables into
classes. different classes.
Considers both categorial and numerical data: takes into Expensive Process: collection of sufficient data, its classification
consideration the quantitative as well as the qualitative variables for and analysis demand high expense, being a resource-intensive
better results. process.