Introduction to hypothesis thinking for marketers: The basics and practice for using it in practice and getting results

"We have data, but we can't use it properly," or "Even after analyzing it, we can't see what we should do next." As a result, we end up getting lost in the data and suffering from analysis fatigue. I'm sure many of you can relate to this situation.

One of the keys to overcoming this hurdle that many marketers face at some point is "hypothesis thinking." In this article, we will clearly explain how to create hypotheses, which are essential for honing your data analysis skills, and the process of verifying them. By not just analyzing, but consistently proceeding from planning to implementing measures and reviewing them based on "hypothesis thinking," you can increase the speed and accuracy of your decision-making.

This articleClick here if you would like to download the document version.

Diagnostic chart for whether hypothesis thinking is rooted

First, check how much hypothesis thinking is ingrained in you and your organization as a whole. If few items apply to you, it may be that hypothesis thinking is not being fully implemented. Use this as an opportunity to find areas where there is room for improvement.

What is hypothesis thinking and why is it necessary for data analysis?

What is hypothesis thinking?

Hypothetical thinking is a way of thinking in which you first draw your own answer (a tentative guess) for something you don't understand, and then proceed while checking whether it's correct. It may sound difficult to explain in words, but we often unconsciously take actions that are the "entrance" to this way of thinking in our daily lives.

For example, when choosing a seat at a cafe, we may unconsciously make predictions such as "I think I can relax by the window" or "It might be noisy near the entrance" and then decide what to do. This is the starting point of thinking in which you make a choice based on a hypothesis, but if you actually sit down and confirm it through experience by saying "It's quiet" or "It's less relaxing than I thought," then it becomes a proper hypothesis thinking process.

Similarly, in marketing, by forming a hypothesis and verifying it, you will be able to interpret data with a purpose and use it to make decisions, rather than just doing a vague analysis. In other words, you will be able to proceed with your analysis with a clear direction, asking yourself "which numbers to look at, what decisions to make, and how to act," rather than being overwhelmed by the data.

Data analysis fails due to lack of hypotheses

We will introduce the cases of Companies A and B, who are working on data analysis of their own e-commerce sites.

Company A's marketing team was faced with a huge amount of data on the EC site, and they started by looking at all the indicators. They spent three weeks analyzing the data from various angles, but the report they came up with was just a list of obvious facts, such as "high dropout rate" and "low repeat rate," and did not lead to any concrete actions.

Meanwhile, Company B's marketing team hypothesized that "new customers were dropping off before completing a purchase, but the membership registration required to make a purchase was a hurdle, causing them to drop off before the registration stage," and conducted an analysis focusing on the dropoff rate at each step of the purchasing flow.

As a result, they discovered that there was a sharp increase in dropouts when the shipping fee was displayed before the member registration process, which led them to take the next action of reviewing the way shipping fees are displayed. In this way, by narrowing the focus of the analysis based on a hypothesis, they were able to eliminate unnecessary testing and take appropriate action in a short period of time.

5 benefits of creating hypotheses

As you can see, formulating hypotheses before analyzing data in marketing is extremely important for improving the quality and speed of decision-making. Other benefits include:

1. It leads to earlier problem detection
By analyzing data based on a clear hypothesis, you can easily find new problems in addition to the problems you had anticipated. This allows you to detect problems early and deal with them quickly, which helps you maintain and strengthen your competitiveness.

2. The data to be collected and analyzed becomes clear

Modern marketing deals with huge amounts of data. However, having a hypothesis makes it clear what kind of data is needed and for what purpose, enabling efficient data collection and analysis without waste.

3. Give meaning to numbers
A hypothesis helps you to interpret why a certain number was obtained and how to interpret the result in light of your marketing objectives. It goes beyond a simple list of numbers and allows you to draw strategic implications.

4. Your analysis results will be more persuasive
Analysis designed based on a hypothesis is consistent in its questions, analysis, and conclusions, making it highly persuasive when used to explain things to stakeholders and support decision-making.

5. It’s a common language for the team

By sharing the same hypothesis among all members involved in marketing, they will be able to see the data in the same way and recognize the issues, which will enable more efficient communication. This will make decision-making and policy coordination smoother, accelerating the movement of the entire team.

Two types of marketing hypotheses

Separate "why it happened" from "how to act"

Hypotheses in marketing can be broadly categorized into two types: "cause hypotheses" and "action hypotheses." Each is used depending on the analysis step and purpose.

A causal hypothesis is a hypothesis that posits the fundamental cause behind a "symptom" (e.g., decreased sales, a drop in the number of branded searches, a worsening conversion rate, etc.) that appears as a result of marketing activities.

For example, the role of a hypothesis is to ask "why" about a phenomenon, such as "Could the price of search ads have risen sharply because competitors are strengthening their bidding?" or "Could the number of store visits have decreased because the amount of TV commercials has decreased?" It is used in the phases of analyzing the current situation and sorting out issues.

On the other hand, action hypotheses are important during the planning and execution phase of a policy, and hypotheses are formulated about what actions will be effective in achieving the set goals.

For example, the action hypothesis is the starting point for thinking about how to reach your goal, such as, "To improve brand awareness, do we need to ensure a certain amount of GRP for TV commercials each week?" or, "To increase customer numbers, shouldn't we change the appeal of our campaign?"

Two perspectives to hone marketers' ability to hypothesize causes and actions

To ensure each hypothesis is of high quality, a systematic approach is required, not just random guesswork.

Causal hypothesis: Create a hypothesis tree

To solve a complex problem, it is necessary to organize and think about the causes one by one. A "hypothesis tree" is an effective way to do this. This is a method for decomposing a problem hierarchically and structurally organizing possible causal hypotheses at each level.

For example, the question "Why aren't sales of new products increasing?" can be broken down as follows:

Structuring the problem in this way makes the true hypothesis to be tested clear, enabling more efficient testing.

Action hypothesis: Always consider causal relationships and mechanisms

An important approach to generating good action hypotheses is to always be aware of the causal relationship (mechanism) of "what happens when you do something?" Rather than simply making a superficial hypothesis such as "broadcasting a TV commercial will increase awareness," it is important to draw a causal flow from actions to results in advance, such as the following:

Air a TV commercial
↓
Builds brand awareness among target audience
↓
When you see the product in the store, you feel a sense of familiarity.
↓
Increased chances of being picked up compared to other companies' products
↓
Increased chances of trial purchase
↓
High product satisfaction leads to repeat purchases

In this way, by mapping out the cause and effect flow in advance, you can understand in advance where problems may occur (bottlenecks), which indicators you should track (KPIs), and when the effects will appear (timeline), making it smoother to design and evaluate actions.

For example, let's consider the case of a new product launch event. Let's say that the result was that the product gained media exposure but the number of stores selling the product did not increase.

In such cases, it is necessary to dig deeper into the question, "Why is adoption not progressing despite the exposure?" One hypothesis is that "buyers do not feel that the product is trustworthy or specialized enough." In this case, a possible solution would be to appeal to the reliability and expertise of the product through information sources that are important to industry insiders (e.g., specialized magazines and industry events) rather than simply advertising or publishing in general media.

It is hypothesized that third-party evaluations and exposure in such forums will encourage distribution buyers to make the decision to adopt the product, which will ultimately lead to an increase in the number of stores that carry the product.

Hold a presentation for industry professionals
↓
Featured in trade papers and specialized media
↓
Establish recognition within the industry, expertise and reliability of your products
↓
Growing interest from distribution buyers
Or, the concerns (= the risks of introducing the product) are resolved.
↓
The number of stores handling the product will increase
↓
Expanding consumer purchasing opportunities

Furthermore, verifying these hypotheses one by one deepens your own knowledge and that of the entire organization, leading to sustainable growth that allows you to flexibly respond to changes in the environment. It goes beyond simply implementing measures, and evolves into marketing activities that continually learn and grow.

Steps to mastering hypothesis thinking in practice

In order to effectively utilize hypothesis thinking, it is important to acquire the skills step by step. Here, we will introduce four specific steps to master hypothesis thinking in practice.

STEP 1: Ask a question

A good hypothesis comes from a good question. One effective way to formulate questions is a technique known as the Toyota Production System, which involves asking "why?" five times. By asking "why?" about a symptom that appears, and then asking "why?" about that answer five times, you can get to the root cause of the problem.

In the example below, we can see that the root cause of low customer satisfaction is an inefficient system. Instead of superficial symptomatic treatment (such as increasing the number of staff), we can focus on the fundamental solution (improving the system).

<Questioning why customer satisfaction is declining>

• Why is customer satisfaction declining? Support is slow
• Why is support so slow? → There is a lack of support staff
• Why is there a staff shortage? → Because of the high turnover rate
• Why is the turnover rate so high? → Because the workload is too high
• Why is the workload so high? → The system is inefficient and manual

STEP 2: Build a hypothesis

Once the question has been clarified, the next step is to consider a specific hypothesis. The following are "conditions for a good hypothesis" that are useful in practice:

<Conditions for a good hypothesis>

1. To be specific:The content of the hypothesis is unambiguous and the actions and effects are clear.
2. It is verifiable:It must be objectively verifiable through data and experiments.
3. It is falsifiable:If the results differ from the hypothesis, it must be possible to conclude that it was a mistake.
4. It is feasible:The content must be something that can be tackled within a realistic scope, taking into account resources and constraints.
5. Leading to decision:The results of the verification will serve as a basis for making decisions that will lead to concrete actions

It is also important to consider multiple hypotheses for a single problem. For example, for the problem of "high bounce rate on website," consider multiple hypotheses simultaneously, such as "slow loading speed of landing page," "headline does not match user search intent," and "mobile display is not optimized." This reduces the risk of falling into a one-sided view.

STEP 3: Design a verification method

Once you have formulated a hypothesis, it is time to consider "how to verify it." If you begin your analysis without clarifying this point, the analysis that was meant to verify the hypothesis will end up causing you to go astray. Designing a verification method means planning in advance "what data, from what perspective, and how to compare them to verify the hypothesis."

For example, let's proceed with verification design with the following considerations in mind:

1. Organize the data and indicators required for verification
Let's clarify which indicators should be viewed, for which period, and at what granularity. To do this, it is essential to identify the necessary data sources (e.g. web analytics, advertising volume, sales data, etc.). In addition, the indicators to be viewed and the aggregation units will change depending on the type of hypothesis (cause hypothesis or action hypothesis), so collecting data and designing indicators appropriate to the hypothesis are the keys to improving the accuracy of your analysis.

2. Decide the axis of comparison and analysis
Consider how to divide the data, such as by comparison axis like "Before/After" or "with/without measures," or by perspectives like "by customer segment" or "by channel." If the comparison axis is unclear, there is a risk that you will not be able to understand the meaning of the results of the analysis.

3. Start small
If you aim for perfect verification, it may take weeks just to design the analysis. The important thing is to start with a "minimal analysis that can be roughly checked" and then update the hypothesis again while looking at the results. The first step is not to produce a perfect analysis or answer, but a prototype to "move the hypothesis forward" is sufficient.

STEP 4: Interpret the results and decide on the next action

Once the verification results are available, it is important to interpret them correctly and use them to guide your next steps. Keep the following points in mind:

1. Consider the context of your data
Interpretation should take into account seasonality and external factors (e.g., competitor actions, changes in market environment, etc.).

2. Check for statistical significance
Make sure your results are reliable and not just a coincidence.

3. Define your next action
If the hypothesis is supported, we will consider developing and scaling up measures; if the hypothesis is rejected, we will review the hypothesis or formulate a new hypothesis and proceed with verification.

4. Support your findings with qualitative data
Hypotheses can also be strengthened using qualitative information such as customer interviews and feedback from the sales field.

Worksheet: Applying hypothesis thinking to your company's problems

Even if you've read this far and think, "I see," you won't develop hypothesis thinking unless you actually take action. Use the template below to apply hypothesis thinking to your company's issues.

• Problem you want to solve: ________________________________________________________________________________
• Possible cause(s): ________________________________________________________________
• Hypothesis I want to test first: ________________________________________________________________
• Verification method:______________________________________
• Required data:______________________________________________________
• Action to take if hypothesis is supported: _________________________
• Action to take if hypothesis is rejected: _________________________

Common mistakes in hypothesis thinking and how to deal with them

Hypothesis thinking is a powerful tool, but if not used correctly, it can actually limit your perspective. A common example is being too attached to a hypothesis.

Having a hypothesis is important, but if you become too attached to a hypothesis once you have established it, you run the risk of ignoring objective data and voices from the field.

For example, the following attitudes can lead to analytical bias and ultimately to poor decision-making:

• They assume that this is the exact cause and ignore other factors.
• Dismissing data that doesn’t fit your hypothesis as an anomaly

The following countermeasures can be considered. A hypothesis is merely a tool. It is important to remember that you are using the tool to its full potential, not being used by it.

Be aware that hypotheses are tentative
Keep in mind that these are not confirmed facts, and try to verify them through verification.

Intentionally looking for counter-evidence
Consider what data or results you would get if this hypothesis was wrong.

Conclusion: "Continuing to ask questions" is the most powerful weapon

The key to improving your ability to deal with data is to make use of "hypothesis thinking." Hypothesis thinking is not a special talent. Anyone can develop the habit of asking questions rather than relying on random ideas, and this can further strengthen their decision-making abilities in marketing.

That said, there will likely be times when a specialist perspective and practical support is needed to actually formulate a hypothesis, select the appropriate data to verify it, implement analytical methods, and translate that into concrete action.

If you need help implementing data-driven marketing using hypothesis thinking,Please contact XICAWe have over 10 years of experience supporting major companies, and we provide analysis and suggestions to support marketers' decision-making.

Click here for the material version of this article

Introduction to hypothesis thinking for marketers
-- A practical guide to turning data analysis into a weapon

Download Now

XICA Co., Ltd. Marketing Team