Find the right problems to solve
Many of the occasions, such as startup pitches, writing scientific papers, and product design processes, require a context of problem-solution. There’s even a concept named “Problem-Solution Fit” which is a premise of product market fit.
99% of problems are NOT worth solving
One of the counter-intuitive facts is that most of the problems are not worth solving. If someone gives you a problem, almost 99% of the cases, the problem was not worth solving at all. If you spend 100 hours to solve the problem, your expected value of “valuable hour spent” was only 1 hour, while 99 hours of your time are just a waste. The wasted time usually doesn’t even contribute to your growth that much because the problem is wrong. Unfortunately, most people are not considerate and intelligent enough to tell you the right problem; therefore there’s a high chance to waste your life.
For example, see the following math problem.
It’s easy to solve. The solution is as follows.
Then, how about this?
It looks difficult. Actually, it’s proven that quintic equations don’t have a general solution.
Let’s evaluate the worthiness of those two questions. In the real world, the solution has already been known for Q1, so we don’t need to solve this problem anymore. As for Q2, it’s not solvable, so it’s nonsense to spend time to try to solve. In short, both of the problems are not the ones worth solving.
Again, 99% of the problems are not worth solving. So many problems have small impacts, and so many others are not even solvable. That’s why the problem validation skill is critical. It’s even more critical in the research field. Identification of the right problems is the most significant part of core values as researchers.
Use words correctly
As a premise, for all the following steps, we must describe all the facts with NO ambiguity. We must be crazy about every single vocabulary choice. If we allow any of overgeneralization, exaggeration, ambiguous words or subjective words, all the logic will collapse.
Exaggeration
- All of our members said that … (maybe not all)
Overgeneralization
- Management team thinks that … (maybe each has a different view)
- Product quality is low (too many definitions of product quality?)
Ambiguous words
- His leadership is not enough (what’s the definition of leadership?)
Subjective words
- AI Accuracy is extremely low (what’s the border of low/high decision?)
Problem Solution Flow
The main message in the article is always to validate problems rather than solutions. However, of course, it’s nonsense only to say “this problem is not worth solving. That’s all!” because it doesn’t proceed anything forward.
I believe that the usual “problem-solving flow” could be like below.
Then, the actual flow should be like below.
In short, we should insert three steps to identify the right problems.
Step 0: Start from the initial problem setting.
We’re starting from initial problem setting, but this problem is most likely a wrong problem. The following example is a typical AI algorithm problem setting. Which sounds like worth solving, but don’t rush. We will make sure that this is the right problem to solve.
Step 1: Confirm Basic Facts
We can spend MAX 1 WEEK to collect as much information as possible to validate the problem. This step is to remove all the subjective feelings and gather facts. The first action is “ask ask ask!”
Understand the ultimate goal
Customer Satisfaction through projects
Gross Profit through the projects.
Don’t make the following inference. “Find a root cause” at the problem validation stage without finding the goal is merely evil.
Instead, do the “so what?” questions recurringly to finally reach to the main topic.
Understand the flow
Requirement Analysis -> Design -> Implement -> Test -> Delivery
Understand the structure
A project success stems from following three elements
(1) Quality
(2) Cost
(3) Delivery on time.
A gross profit = Revenue — Costs
Understand basic numbers
Revenue: $300k
Cost: $2k x 3 people x 3 months.
Accuracy: 70%
Understand how other people think
X said the client satisfaction was bad because the AI accuracy was low.
The important thing here is NOT to believe the reasonings. “AI Accuracy was bad” may not be a real problem. The fact here is only that “X said so.”
Step 2: List up the problem list
After understanding the structures of the business, we should write the issue tree like below. The top level would be the list up of elements. We should have it as MECE as possible.
In the following example, we can see that after breaking down customer satisfaction, AI Accuracy could be one of the problem candidates, but there are a lot more than that which also look like problems.
Step 3: Evaluate each problem
After listing up, we’ll try to figure out whether those problems are;
- Solvable
- Haven’t been solved yet
- Impactful after the solution.
Regarding (1), usually, we can see by seeing similar cases in other domains and competitors. For example, if Google couldn’t do it, it’s highly likely that we can’t solve it either. If a competitor has already solved the problem, it’s likely that we can also solve it.
Regarding (2), we must ensure that nobody else in the team has already worked on it. Ask, ask, ask!
Regarding (3), for example, if the increase of accuracy from 80% to 85% requires enormous efforts, but the impact of 5% is small, it’s not likely that we should spend that effort but find other problems to solve.
After evaluating those problems, if some of the problems satisfy all (1)-(3), it’s luckily the ones worth solving!
For example in the picture above, maybe “Are there any UXs for further AI quality inspection which may connect to customer satisfaction?” could be the right problem setting rather than AI accuracy improvement.
Quality of Problems
When we hear the story, I am most impressed by the problem statement rather than solutions. If the problem setting is right, usually solutions are rather straightforward.
One typical kind of high-quality problem statements is very counter-intuitive. For example, if everybody naturally feels that AI accuracy is a problem, but he/she focus on UI problem with good reasonability, this is usually a high-value problem because most of the counter-intuitive fact findings are very well thought out.
Conclusion
Think so well with the problems you want to solve.
To pursue it, we all must have a very wide range of thought and also it must be profound. The profoundness is often presented by its counter-intuitiveness.
