How do you apply the scientific method to your problem-solving?

The very entry point to all knowledge work is the observation by statistician George Box that “all models are wrong, but some are useful”. This heuristic tells us to make our models useful rather than truthful. Secondly, to find out what is importantly wrong with our models. This puts a bounding box on their domains of validity (their use-cases).

Defining the problem: 1. Understand why "the problem is a problem" 2. Understand how it became a "problem" 3. Understand how it remained a "problem" 4. For whom is this a problem? 5. How long has it been a problem for? These are some questions to ask yourself while defining the problem.

It is necessary to dig as deep as possible at this first step. Often the definition of the problem is too superficial, only sypmtoms which cover the real problem. To reveal the root cause I often use the 5 Why? technique: asking "why" questions after every answer help to uncover the hidden levels and the complexity of the topic. I do it until I find the Real Problem to solve which is often completely different from the original challenge.

When formulating my hypothesis I aim to ensure that cause and effect are measurable in a simple KPI. Hypothesis: "Our generative AI art application will increase the average daily user sessions to 3 sessions per day, resulting in higher user retention and in-app purchases." Components: Product/Service: Create personalized art for users. Key Metric: Increase in average daily user sessions to 3 sessions per day. Assumption: Personalized art generated by the application will encourage users to visit the app multiple times a day, increasing the average daily sessions. Expected Outcome: Increased average daily user sessions and higher user retention due to the appeal of personalized art. That's just an example though.

Scientific hypotheses need to be testable and falsifiable, so that can be subjected to scrutiny, and potentially proven false by empirical evidence. If the null hypothesis is true, any experimentally observed effect is due only to chance. Thus, proving this wrong establishes that a relationship exists, although further experimentation is needed to determine the nature of that relationship.

One of the main points that should be taken into consideration here is that the requirements of the experimental design are well defined. When experimental design requirements are not clear, unnecessary and inadequate measurement occurs. This makes the reliability of the research questionable.

There are some characteristics of the cientific method combined PDCA cycle which can bring considerable results. Combining the scientific method with the PDCA cycle provides a strategic and iterative approach. Specifically, the Check phase is where data analysis stands out. The scientific method contributes to a rigorous evaluation of the results obtained. The hypotheses formulated at the beginning are compared with the real data collected, allowing an objective interpretation of the results. If goals were not met or unexpected challenges arose, this data analysis guides identification of the underlying causes.

A comunica??o de descobertas deve ser realizada de maneira visualmente simples, evitando excesso de informa??es em cada etapa da apresenta??o, pra torná-la acessível a qualquer pessoa. Isso diminui a carga cognitiva e aumenta a compreens?o de quem terá acesso as descobertas. Alguns pontos a serem lembrados s?o: - Limitar a quantidade de informa??es por slide. - Utilizar gráficos com área de dados linear, como o gráfico de barras. - Desenvolver uma paleta de cores que harmonize entre si. - Construir um bom storytelling para relacionar os dados