What’s Wrong With You?
Defining what a problem is, and the art of Problem Framing
One of the biggest reasons AI projects fail is simple: teams build solutions to the wrong problem (or to no real problem at all).
Companies often pour resources into trendy AI solutions only to find they don't actually fix any real pain point.
This misalignment stems from a fundamental failure of problem definition.
This article explores what a problem really is and how to properly define and frame problems before jumping to solutions.
A problem well framed not only guides effective solutions, it also aligns teams and stakeholders on what truly needs fixing.
Let’s dive in.
What Is a Problem, Really?
At its core, a problem is the gap between where you are now and what the normal state and normal situation you want to be in.
Current reality ≠ Expected state = Problem.
In other words, a problem is the difference between reality and what is judged to be a nominal situation.
Easy, right?
But many teams still confuse symptoms with problems ("our sales are down") or leap straight to solutions ("we need a machine learning model") without clarifying the underlying issue.
A better approach: a problem exists when your normal way of doing things no longer achieves your desired results.
In other words, we need to define the problem in terms of what state should no longer exist in a quantified way, not as a wishful solution you would like to build.
Once the problem is solved, the current state should no longer be the same.
The Importance of Proper Problem Framing
A problem well-defined is a problem half-solved.
This adage highlights that investing time in understanding and articulating the problem pays huge dividends in the solution stage. Einstein reputedly said that if he had one hour to save the world, he'd spend 55 minutes defining the problem and only 5 minutes on the solution.
A clear problem statement aligns and focuses the team.
In design thinking, the problem statement acts as a north star for the project. Without it, teams can wander in different directions. Good problem framing means expressing the problem in negative terms (what is going wrong) rather than jumping to a solution description.
For example, instead of "We need a faster database" (which presupposes a solution), frame it as: "Our database response times are five times slower than acceptable, causing user frustration."
This framing highlights the gap and opens the door to multiple solution approaches.
The Data & AI Perspective
In AI projects, precise problem definition is especially critical.
These projects must clearly define the target metric or success criteria up front. Unlike traditional software, AI systems learn toward an objective function – so you need crystal clear, measurable success criteria (e.g. "reduce customer churn rate by 20% within 6 months").
Also, AI projects involve many stakeholders: business sponsors, domain experts, data engineers, data scientists, end users. A clear problem definition creates shared language among all parties. When everyone agrees on the problem statement, it's easier to get buy-in and collaboration.
Finally, AI projects also teach us about the iterative nature of understanding problems. You often start with an assumed problem, but as you explore the data and domain, you refine or redefine it.
For instance, a team might begin wanting to "predict customer churn," then realize they need to clarify what "churn" means – 30 days? 90 days? Voluntary vs. involuntary cancellations? Data exploration might also reveal that the business actually doesn’t need predictions, and help reframe the problem in the following way: “20% of current customers leave after only 6 months of subscription and we don’t know why”.
It is therefore crucial to be willing to revisit and sharpen the problem definition based on new insights.
Visualizing and Communicating Problems
Visual frameworks help everyone understand complex problems involving multiple factors and stakeholders.
Here are popular tools that can be useful for illustrating problems (note that they are also usually used to address the corresponding solution):
Double Diamond (Design Council): This framework separates problem space from solution space. The first diamond involves discovering and defining the right problem before moving to developing and delivering solutions. Teams should spend significant time discovering and defining before ideating solutions.
“Five Whys” Technique: This is a simple yet powerful tool to drill down to root causes by repeatedly asking "Why?" For example: "The machine stopped." Why? Fuse blew. Why? Circuit overloaded. Why? Motor bearing failed. Why? Insufficient lubrication. By the fifth "why," you've uncovered the root cause that wasn't obvious initially.
Empathy Maps: For socio-technical problems, map out people involved and their needs or pain points. This paints the landscape in which the problem sits and can highlight disconnects or overlooked issues.
Putting It into Practice: Tips for Effective Problem Framing
Defining a problem well is a skill you can develop.
Here are some practical steps and tips to try on your next project:
Articulate the gap: Clearly describe current state versus desired state. What outcome aren't you achieving? Make the gap obvious (e.g. "Our average response time is 10 seconds, but users expect 2 seconds").
Frame in negative terms, not solutions: Describe what is wrong or lacking, not a prescribed fix. Say "error rates have doubled causing customer complaints", not "we need a better algorithm".
Align with stakeholders: Draft a problem statement and share it with all key stakeholders. Do they agree this is the core problem? Refine until there's consensus.
Leverage visual aids: Try sketching problem diagrams or using frameworks like Problem Trees, Double Diamond, or Empathy Maps. Visualizing creates shared understanding.
Revisit and refine as needed: Treat problem definition as iterative. As you gather data or prototype, remain open to revising the problem statement. This isn't failure – it's learning.
Use "5 Whys" or other root-cause tools: Investigate why the problem exists before jumping to solutions. The first answer is often just a symptom. This step will help you address the right problem and start thinking about potential solutions to address it.
Your turn
The most successful projects start by asking: What's the real problem we're trying to solve?
Answer that clearly, and you're already halfway towards a solution.
