A complex problem is made up of many interrelated elements that can behave adaptively. Each element can be clearly identified and understood with regards to their own characteristics and local rules but when considering the relationship between them, there is a high level of dynamic, non-linear interactivity. This means that you cannot predict the overall outcome of the system by looking at the interaction between small parts. The system is more than the sum of its parts as the main patterns within the problem can only be identified and analysed after the fact. This dynamic nature fundamentally alters how you approach problem solving in real-world scenarios.
To fully understand what a complex problem is, it helps to examine it alongside two adjacent categories formalized in systems thinking: a complicated problem (as delineated in Dave Snowden’s Cynefin framework) and a wicked problem (first formulated by design theorists Horst Rittel and Melvin Webber in 1973).
- Complicated Problem : A problem that has a number of elements (sub-tasks and components). The relationships between the elements can be clearly defined and everything can be predicted. This can be broken down using domain knowledge to find a definitive solution.
- Complex Problem : A system of adaptive elements that can interact with each other. Each element has clear local rules but high levels of non-linear interaction. This makes the overall outcome unpredictable. However, the final definition of a solution is clear.
- Wicked Problem : A systemic social issue that is greatly affected by human subjectivity, competing desires and conflicting values. It becomes impossible to even agree on an objective definition of the problem, meaning a stopping point is impossible to identify and no standard solution exists.
In Simple Terms
If you found the formal definitions confusing, then here is a simple version:
- A complicated problem is a large puzzle with a set of rules. There could be thousands of elements, but they all work in the same way.
- A complex problem is a living system. Each element can be understood individually but when they interact with each other there are chain reactions that cannot be predicted using any formula.
- A wicked problem is a human conflict. The issue isn’t the mechanics of the problem itself, the problem is the people involved want different things.
How to Classify Your Problem in 30 Seconds
Before investing time and engineering resources, you can diagnose what kind of problem you are dealing with using three practical questions:
- Can the outcome be reliably predicted using blueprints or standard rules? If yes, you are dealing with a Complicated Problem. Break it down into modular sub-tasks or consult domain specialists.
- Does altering one component trigger unpredictable, systemic ripple effects? If yes, you are facing a Complex Problem. Stop relying on rigid linear planning; build simulations and run iterative feedback probes.
- Do key stakeholders fundamentally disagree on what counts as a successful outcome? If yes, you have a Wicked Problem. Technical analysis alone will fail because the friction stems from competing human values, jurisdiction boundaries, and conflicting trade-offs.
Structured Breakdown: Complicated vs. Complex vs. Wicked
| Breakdown Dimension | Complicated Problem | Complex Problem | Wicked Problem |
|---|---|---|---|
| Elements & Rules | Predictable: High density of sub-elements and steps, but relationships are clear and fixed. | Dynamic & Non-Linear: Local rules are clear, but dense interactions make overall behavior impossible to predict in advance. | Value-Driven: Intertwined with human will, competing jurisdictions, and conflicting interests. |
| Splitting It Up | Whole = Sum of Parts: Can be sliced top-down into independent sub-problems and reassembled. | Whole > Sum of Parts: Cannot be split into independent pieces; cutting it apart changes the problem itself. | No Shared Consensus: Causes and authorities are easy to spot, but stakeholders cannot agree on the trade-offs. |
| How You Solve It | Blueprints & Expertise: Driven by domain knowledge and past experience; manuals make it much easier for beginners. | Simulation & Probing: Build a test environment, feed in elements, watch the feedback, and adjust on the fly. | Competing Solutions: Solutions and trade-offs are clear, but action stalls because nobody agrees to compromise. |
| Completion Boundary | Clear & Definitive: Objective finish line with zero subjective value trade-offs. | Clear & Definitive: Even though the path is unpredictable, what counts as “solved” has a clear boundary. | No Stopping Rule: No standard answer; clear solutions exist, but they are rejected by opposing sides. |
How I Understand the Three Problems
I don’t rely on theory when understanding how these problems play out in practice. Below is how I understand each:
1. Complicated Problems
When faced with a complicated problem you can break down a large problem into individual sub-problems. This means you have a ‘whole is equal to sum’ situation. The end goal is very clear and there’s no compromise in value judgement.
Based on my experience, a complicated problem often depends on your knowledge in the particular domain and if you’ve had experience solving this type of problem previously. If you have you’ll find it much easier, however if there’s a guide or playbook you can follow, even someone with no knowledge will be able to solve the problem with ease compared to someone who hasn’t.
This is similar to having a large Lego set to build with. With a guide you can build up individual parts and then snap them together. Or even build the whole thing at once. As each part can be broken down into individual sub-problems, you can give individuals different pieces to work on to speed up the process. This reduces the complexity. For example in manufacturing, the car won’t be built from scratch by the car manufacturer. They outsource certain parts to specialist factories and then assemble the car. This reduces R&D costs massively. In software engineering, this is identical to building a standard CRUD pipeline or migrating a database schema against a known specification—high cognitive load, but entirely solvable with expert playbooks.
2. Complex Problems
When dealing with a complex problem you cannot break down a large problem into individual sub-problems.
The only way to solve a complex problem is to simulate a scenario and keep testing elements to run mini-experiments. Once you have a simulated system, you start to add elements to test how it reacts. You keep altering your approach until you find a solution. This makes solving them quite complex due to the amount of elements and links between them. However there is a definite end goal.
This is like trying to untangle a messy ball of yarn. If you pull one loose piece of yarn on the left you may cause the yarn on the right to become a tight ball. You can’t just force a solution and have to try gently pulling on a piece of yarn. The solution is very clear: either you untangle all the yarn or knit it into a jumper.
You can’t break down a complex problem into smaller pieces. For example if you took a pair of scissors and cut the yarn ball in half to make it easier to untangle. Whilst you could untangle each half, the yarn ball you have just solved is no longer the yarn ball you started with. The length of yarn has changed so you have altered the sub-elements. In modern software and business, this mirrors debugging cascading failovers across distributed microservices or navigating a two-sided marketplace cold-start—interconnected feedback loops mean you cannot isolate a single element without altering overall system behavior.
When facing this kind of complex problem where every element pulls on the whole system, you cannot rely on gathering all data upfront or dissecting it with a static plan. Instead, you need a dynamic hypothesis-driven approach—using iterative, small-step probes and constraint testing to gradually understand how the interrelated elements actually behave together.
3. Wicked Problems
A wicked problem often starts with a visible symptom, yet stakeholders fundamentally disagree on what the core problem actually is. Because there are many overlapping jurisdictions, competing interests, and conflicting values involved, each group frames the causes and boundaries completely differently. Multiple competing solutions exist, but each comes with heavy trade-offs that opposing sides refuse to accept.
Because of this, there can never be unanimous agreement among the different parties.
This seems to me like a group of people trying to pull on a large object. As the object is so large they can’t see the whole thing. Everyone is using their full strength to pull the object towards them. The large object doesn’t move at all. The solutions are very clear but no-one will agree to them. In organizational strategy, this manifests in defining platform content moderation rules, structural incentive designs, or corporate return-to-office mandates—every definitive technical fix directly harms the interests of at least one vocal stakeholder group.
Frequently Asked Questions
Can a complex problem be broken down into smaller parts?
No. Unlike a complicated problem where the whole equals the sum of its parts, cutting a complex problem into isolated pieces alters the underlying sub-elements and severes their dynamic interactions—meaning you end up solving a completely different problem than the one you started with. Instead of mechanical decomposition, you must use contextual problem formulation to clarify system elements before taking action.
What is the best way to approach a complex problem?
Because you cannot predict the outcome of a complex problem using a static blueprint, you must rely on simulation and iterative probing—testing small elements within a controlled environment, observing system feedback, and adapting your approach. To systematically structure your iterative probes before taking action, you can map out your testing hypotheses using a logic tree without isolating the underlying elements.