The scientific method is a proven route to successful, tested and verified improvement. Here’s how to combine it with business intelligence (BI). Good businesses are always looking for ways to improve. But sometimes they just stumble along trying things at random or maybe even following the “gut feeling” of a senior manager. Unfortunately, these methods are rarely a recipe for success. While there are no guarantees about what specific actions might improve the prospects of a business, there is a tried and true scientific process that’s likely to lead to better business outcomes: the scientific method.

What is the scientific method?

Humans have used the scientific method throughout the course of recorded history. Though that may sound like a “grand” statement, it’s far from an exaggeration. Through the scientific method, humanity has given itself a clear path to test assumptions against observations. Where early humans used the scientific method to identify simple phenomena such as “rubbing sticks (friction) makes fire,” it’s since become the cornerstone of nearly every modern technology and scientific advancement.

Of course, a complete history of the scientific method and its underlying scientific process is beyond the scope of this article. What does bear mentioning, though, is how the scientific method works, particularly how we can take advantage of it for BI.

Generally speaking, the scientific method involves six key steps:

1. Observation or question: What are you observing and what can you question about it? Observing and questioning some phenomenon (i.e., “why is X doing Y?”) is the first and most crucial step of the scientific method. For example, Isaac Newton may have questioned why objects fall to the ground after observing an apple fall from a tree.
2. Research topic area: Once you’ve observed something worth questioning, it’s time to research everything surrounding the topic. From the apple example, Newton might have looked to classical physics and other scientific knowledge and noted that all objects have mass.
3. Hypothesis: Now that you’ve observed something and researched the topic extensively, it’s time to link the two to “guess” why that certain something is occurring. Here, Newton might have hypothesized that an object’s mass had something to do with it falling to the ground.
4. Test with an experiment: Now it’s time to test your “guess” or hypothesis. Here you might set up several experiments to see if you can repeat or vary your observations as you change or control certain variables. As you might imagine, there are plenty of ways to conduct experiments — and how you conduct them will depend on your problem. For example, Newton might have tried measuring how different objects of different masses fall to the ground. Did objects with a heavier mass fall faster?
5. Analyze data: With your experimental data, you can try to identify clear relationships. Sometimes there won’t be any, but that doesn’t mean what you’ve observed isn’t worth observing. Here, scientists might rethink their experiments and try different scientific methods or maybe rethink their assumptions. When Newton observed that all objects reached the ground at the same time regardless of their mass, he probably went back and reformulated his hypothesis.
6. Report conclusions: Did you identify a clear relationship in your data? If so, it’s time to link why and/or how this relationship supports your hypothesis. Newton would later conclude that every particle is attracted to the other through a universal law of gravitation — or, alternately, the apple fell to the earth because the earth’s gravity pulled it to the ground. 

Note that this process is iterative. 

In other words, the scientific method is an ongoing process that often repeats itself with new conclusions, especially as they pose new observations, questions and hypotheses.

As we’ll see in the next section, the scientific method isn’t necessarily limited to “scientific” subjects such as figuring out gravity. Instead, it’s general enough to apply to many applications, including BI and scientific data analysis.

Applying scientific method data analysis to BI

The scientific method is hugely valuable for business intelligence analysts.

Let’s take a closer look at what the scientific method looks like when applied to a BI situation. In this example, you’ll be the owner of a sports team (congratulations!) trying to figure out how to maximize your income from seat ticket sales.

1. Make an observation and/or pose a question

First, you make an observation — in this case, say you notice something unusual or noteworthy about your ticket sales.

Suppose you’ve noticed that you sell more seats when your team plays at certain teams or during certain weather or certain days of the week. As your business goal is to maximize your income from ticket sales, these are interesting observations.

2. Perform background research

Next, you’ll do some background research into the processes, people, places, constraints and concerns surrounding what you’ve observed.

Here, the goal is to put your observations in context and develop a common context that can be used to discuss the issue or the problem. Statistical associations are helpful, but so is understanding the underlying mechanics of a question.

3. Form a hypothesis

Once we know the context for our observation, you might form a hypothesis such as “we can make more money if we raise ticket prices.”

To be useful, a hypothesis must be testable and measurable — or, more accurately, falsifiable. In other words, it must be possible to imagine a test result that could prove your hypothesis wrong. Any hypothesis (or supporting conclusion) that holds water should be able to withstand any attempt at being proved wrong.

A hypothesis also lets us make predictions. For example, if you raise prices everywhere in the stadium by $2 a seat, you’ll probably make more money. Predictions like these are testable by experimentation or taking some action and measuring the results. In the context of BI for a retail business, that might mean lowering the price of an item, changing store hours or running advertisements. 

The same is true for our sports stadium: could you make more money by lowering ticket prices, hosting games at different times (such as on weekends instead of weekdays) and/or promoting upcoming games around town?

4. Design an experiment

You can now design a series of experiments that test the effects of certain actions as they support (or disprove) your hypothesis.

In most cases, you’ll maintain not only an experimental group but also a control group on which no action is taken. Doing so allows for a better comparison by minimizing uncertainty. To zero in on which factors have the most influence, it’s best to change a single factor at a time and measure the results. However, that’s not always possible in the real world.

A practical example of experimentation commonly used by web companies is the A/B test, in which users are randomly assigned a control or experimental experience so that they can be compared head-to-head.

For the stadium, you might perform the following experiments:

• Raising and lowering ticket prices: All other things made equal, how does ticket price affect turnout when a game is hosted at the same time and same day every week? For example, you might charge the normal price one Saturday afternoon, a higher price the next Saturday, a lower price the next and so on, all the while keeping your advertising and promotions unchanged. Here, the normal price is your control group, while the higher and lower prices are your experimental groups.
• Hosting games on different dates: Are your fans weekend warriors or would they rather break up a mundane work week with a night at the stadium? The only way to know is to try both and see what happens.

As you might imagine, there are lots of different ways to design an experiment — in fact, you’ll likely have more possibilities than you can practically experiment with. Instead of trying them all, stick to only a handful of “major” variables such as ticket prices and game times.

5. Analyze the data from your experiment

As the experiment proceeds (or when it’s finished) you’ll need to analyze the data you collected. 

In our sports example, we may confirm that enough people are willing to pay more for a seat at games for revenue to increase — or we might find just the opposite, that when ticket prices get past a certain point, people will attend fewer games and, as a result, stadium revenue drops.

Though our stadium experiment is pretty easy to measure (ticket prices vs. attendees), many real-world problems are far more complicated. As a result, analyzing the data is the obvious place for using BI tools, which we’ll cover later on.

6. Report conclusions

In scientific research, the next step would be to publish the results to give other researchers a chance to reproduce and confirm them.

While that’s not the goal of the scientific method as it applies to business, it’s still worthwhile for businesses to share the results of their experiments internally so other managers, departments and divisions can use them as the basis of their experiments.

For example, the stadium’s marketing team might highlight the “best” ticket price in a series of new promotions.

Rinse and repeat?

As we mentioned, the scientific method is an ongoing, iterative process.

Forming a hypothesis, performing an experiment, drawing conclusions and then using those conclusions informs the next hypothesis. In this way, we set up a continuous cycle of improvement.

Suppose our price experiment showed that raising prices by $2 a seat brings in more money. We might run the experiment again and raise the prices by another $2 the next year. At a certain point, we’ll have maxed out the revenue we can get from broad, general ticket price raises and we’ll have to make new observations, generate new hypotheses and conduct new experiments.

Conversely, the results may be more nuanced, suggesting the need for a range of ticket prices for different kinds of seats or seats purchased in advance and those purchased on the day of a game.

In any case, all of the observations we make on our experiment become fodder for future experiments.

Tools for scientific method data analysis

Many factors might pertain to each hypothesis. Since this can get rather complicated, many businesses use data tools to support their BI experiments.

For example, a marketing writer might experiment with different kinds of content (blog posts, case studies, ebooks) of different lengths, different publishing frequencies, different modes or media, promoting content or depending on organic site visits, promoting it on different sites (LinkedIn, Facebook, Twitter) and so on. 

Have you encountered similar scenarios in the areas you’re responsible for?

To analyze scenarios with multiple (and potentially complex) factors, you need an analytics infrastructure, business intelligence software and statistical analysis techniques that can draw data from multiple sources, collate it, filter it, sort it and ultimately present it to decision-makers for review.

• Business intelligence tools (such as Tableau, Microsoft Power BI or Qlik) to analyze your data and extract insights.
• Data warehouses (such as Amazon Redshift, Microsoft Azure Synapse, Google BigQuery or Snowflake) to store your data.
• ETL data pipelines (such as Fivetran) to provide a seamless connection between data sources, warehouses and analytics tools.

These are all crucial components for using the scientific method to unlock more insights within your business’s data.

How Fivetran empowers scientific method data analysis

Our hypothesis? Your business will be more profitable if you combine the scientific method and BI tools to improve your business processes. We also suggest using the scientific method to build your data infrastructure. With Fivetran, you can seamlessly integrate data from a wide range of disparate sources, extract and load them to your data warehouse and transform them for any number of applications.

‍