We thought it would be good to put some more light on a methodology we are using — Augmented Decision Making (ADM). What is it, and how does it change decision-making?
We need to begin with an annotation that ADM is a standardly defined field of business management. Practitioners describe it with several differences. But the main idea is the same: the use of computational methods to support decision-making.
The term draws from Augmented Reality (AR), phenomena that enhance interaction with the surrounding environment. The most recognized and commonly used example of AR is in visual applications. A couple of years back, Google presented Google Glass. This innovative product did not break into the mainstream, but the tech giant introduced the Google Lens app instead. The app connects the phone “vision” with the internet. You take a picture of a tree or animal, and Google tells you what the name is and connects to the right Wikipedia page. You get all the internet knowledge about whatever you capture through your camera. There are many similar AR solutions in vision, audio, or touch. You receive computer-generated information about anything that you experience in real life. Brilliant!
ADM does the same for decision-making. Imagine, that while solving an important business challenge, you can look at an option and get all the information about its outcomes. What will be the cost of implementation? Will the customers like it? How will your employees accept it? Etc. You want to look at a different option: no problem, click here. You like Option B, but you want to adjust one thing? You simply tune what you want. The ADM process does all the analysis in the back end and generates clear information about new outcomes. You can have the overall perspective on the key challenge and dive into nuances of a single detail and always get useful information that helps you make more accurate decisions.
How much can we augment?
There is a spectrum of how much “support” decision-making can get. Recent advancements in fields like machine learning or AI already allow to get rid of the human factor in certain business problems. In such cases, the software itself recognizes the problem, analyzes it, defines the solution, and even implements it. A step below is Automated Decision Making, ADM (or Automated Decision Support, ADS). A person defines a category the challenge falls into. Later the software does all the job and only needs the final approval.
On the other side of the spectrum, you can find the Excel Pivot Table that can be recognized as ADM for simple numerical problems. Between, there is an ocean of tools and methods that augment the decision-making process: data science, predictive analytics, process analysis and modeling, simulations, and many others.
What is augmented in ADM?
The decision-making, like any process, is a sequence of activities performed to produce the desired output, the final decision. Each activity in the path uses certain methods and tools to process (alter, analyze, redirect, or add value) information received from the previous step. There is always certain logic behind the order of the activities.
To answer the question “what can be augmented?” let us reformulate it to “what can go wrong?”. Process problems include:
- Logical errors — when the path itself is wrong. The order of activities is incorrect. Information is missing. Outputs from one step are not going to the right place. And other similar mistakes in designing the process path.
Consequently, the process stops in the middle because the requirements of a particular step are not fulfilled, or there is a defect in the final product. In decision-making, logical errors are the usual reasons for:
- Being stuck — no matter what additional information we analyze, we cannot move forward with the decision.
- Substitute decisions — solutions that are not answering the challenge the process started with. They either answer a completely different problem or solve only a fraction of the original challenge.
- Biased decisions — it happens when, because of various biases, we (usually unconsciously) design the process in such a way that it ignores key factors or information.
- Methods errors — when we choose the wrong tools and methods to process information in a particular step or commit a mistake while using them.
These are more difficult to spot than logical errors. The solution is produced and seems logical, but it is inaccurate, its uncertainty is high, or both.
ADM is a methodology that not only fixes these errors but also enhances the process logic and methods used. It starts by creating a model of the business environment. The model maps and recreates all the driving forces behind the goal and factors that influence the decision. To describe them, the necessary data and information is defined and associated with proper collection and analytical tools. Where data is absent or incomplete, probability scenarios model uncertainty. As a result, we have a logical path that tells us what and how to analyze to produce workable solutions. And the model tells us how to evaluate them and compare, so they achieve the business goal.
ADM in practice
In practice, ADM influences every stage of decision-making.
1. Problem Framing
The first and most crucial step. Traditionally it is used to define what is the objective of the particular decision problem. In ADM, because we need to create a model, we dig deeper by asking more about why and what. Why do we have this decision problem? Because we want to achieve some higher business goals. Ok, so what are those goals? What are the internal and external factors that influence them? Which of these we know and what is uncertain? ADM extracts the following:
- Business goals — a set of values that will facilitate the evaluation of future solutions. When we have competing alternatives, it is good to have a big picture — select the one that not only answers the challenge better but also aligns more with our business values.
- Decision constraints — what are the objectives of the decision and what limits the portfolio of possible solutions, what are the musts, and don’ts that we have to obey.
- Decision variables — these are all the factors that we must take into consideration while making the decision. After creating the initial model, we have a general idea of how they affect our decision, but we still have not measured them. Some variables are known, and we will use data to analyze them. Others are unknown or uncertain, and we will use probability scenarios to define them.
2. Definition of the Process
All the above points and information about decision stakeholders, available data and budget, expected length of the process, and the desired precision, are considered to define proper logic and tools used in the process.
3. Information gathering and analysis
The hard part is over. The main framework of the model is defined. Now we just need to complete. The first step is data gathering. This includes details about internal processes, external environment, variables, and uncertainty mentioned in the first step. At this step, due to numerical analysis and statistical methods, we always learn that in fact there is less uncertainty than anticipated in the very beginning. In the end, all the known is measured, analyzed, and the existing connections are mapped into functions and processes creating the business model. The uncertainty is broken down to single factors with assigned probability distributions. This gives us the uncertainty model that will be used to generate probability scenarios.
4. Identification of solution alternatives and evaluation
Since we already have the business model, decision objectives, variables, and finally, the uncertainty model, we generate possible solutions. To do it, we combine all possible values of variables that are within previously defined constraints and shuffle it with probability scenarios. As a result, we get possible solutions with the defined likelihood of occurrence.
Evaluation and comparison of formulated alternatives are what takes the most time in the traditional decision-making process and frequently are a reason for conflicts between decision stakeholders. ADM speeds up and structures this stage. Because the solution alternatives were generated through a business model, they carry the information not only about how they answer the decision challenge but also how they affect the overall business goals. The model scores the outcomes and informs us about each scenario’s performance in a form of a ranking.
6. Decision making and execution
At this point, we have a list of scenarios with known likelihood and business outcomes. Still, if we want to change something in the assumptions, available budget, or whatever in the model, we can do it, and the list of possible solutions gets updates. At this stage, we usually choose the solution with the highest likelihood and best performance, but there may be different preferences. Do we need to inform someone before making the final decision? Or should there be a certain communication about the decision? The stakeholder’s analysis from the first step answers these questions.
In real life execution of a decision is a whole new process and may face various internal or external obstacles. Major decisions usually bring a change — new processes, new systems, new work organization. Something will be done differently, and the new way needs to be implemented. Change management should include information, that we already defined in the business model (mapped processes, values, and stakeholders) in the implementation plans to assure smooth and successful execution.
The model will also help us define evaluation tools and milestones that the implemented solution should achieve. Previously we analyzed variables and influencing factors by decomposing them into the smallest features. Therefore, we know exactly which processes contribute to achieving the goal. To evaluate the achievement of the goal we just need to assign control measures to those processes.
If you are interested in solving your business problem using ADM, leave us a message and we will reply.