[Data Analysis from Scratch #5] 4 Tips for Field Staff to Involve Management and Organization Even if You are a Data Analysis Beginner

This series explains the basics of data analysis for those who have never studied data analysis before. In the fifth installment, we will focus on "involving management and the organization" and introduce four tips for involvement that will be useful in the field.

Data analysis does not end with the analysis. Data can only be considered useful when action is taken based on the analysis results and results are achieved.

In order to take action, it is important to involve the entire organization, including management and other departments.However, there is little information available on how business people in charge of data analysis can get management and other departments involved in the analysis process. I hope that this article will help you put into practice data analysis that gets your whole organization involved.

↓ Click here for a list of articles on "Data Analysis from Scratch" ↓

#1 "8 steps of analysis" that beginners should know first
#2 Three analytical techniques that data analysis beginners should remember
#3 Communication tips to get management involved that data analysis beginners should know
#4 What data analysis beginners need to know: What management expects from data analysis and analysts
#5 4 tips for field staff who are new to data analysis to get management and the organization involved

In this final installment, we will introduce four useful tips for Step 8, “Involvement,” of the eight steps of data analysis.

Tip 1: Talk about your vision and results

What are the concerns of management and other department members?

It's not a question of whether or not advanced analysis can be performed,"Will it lead to success?".

Data workers are responsible for connecting data to outcomes and the vision beyond.

Work backwards from your vision of "what kind of world do we want to create?" and discuss "what results will this data bring to achieving that vision?"This is the first step in getting management and other departments involved. 

Data analysis can easily stray from its purpose, even if you are conscious of it. I often hear stories of people losing sight of their vision and purpose and getting set back because they were unable to collect the data needed for analysis, or the analysis results were different from what they expected. 

In this way, in order to prevent the situation where "analysis" itself becomes the goal and leads to failure,It is important to communicate data in terms of vision and results and to spread it across the organization.

Tip 2: First, create a successful experience

The quickest way to get the organization involved is toTo give people the experience of success, where they realize that taking action based on data produced results.

So firstCreate results through projects that generate small successes and bring success experiences to the organizationThis is the second point.

Data analytics doesn't have to cost huge amounts of money, and data that is cheap to collect and analyze is often enough to achieve small wins.

As small successes pile up and the entire organization begins to realize the benefits of data analysis, the importance of data will gradually increase within the organization. Cultivating a data-driven organizational culture where decisions are made based on data starts with these small steps.

Once a data-driven organizational culture is created, translating data into business outcomes will become the norm.

Tip 3: Appeal to the visuals

Reports summarizing the results of data analysis often contain unfamiliar formulas and complex graphs, and in some cases, even if shown as is, participants may be dismissed as "kind of difficult."

In such a casevisualLet's make use of it.Designed to be easily understandable to everyoneI can. 

Let's look at a simple example. The two graphs on the left side of the figure below show the "amount of TV commercials" and the "number of impressions of banner ads," respectively. If you were asked to look at these and find a correlation, it would be quite difficult, wouldn't it?

Now, let's look at the graph on the right. This is a combination of three graphs showing "TV commercial advertising volume," "number of banner ad impressions," and "number of conversions." Comparing the waveforms, we can see at a glance that TV commercials are more closely linked to conversions than banner ads. 

In this way, by overlaying waveforms on waveforms rather than showing several graphs separately, it is possible to communicate in an easy-to-understand manner even to those who are not familiar with analysis. This is the third point. 

Tip 4: Use data to your advantage

The premise of analysis is "data collection." However, it is rare that the necessary data is perfectly prepared when you are about to start data analysis. In addition, while cooperation from the organization is essential at the data collection stage, it may be difficult to ask for cooperation in data collection at a stage when results have not yet been seen.

In such cases, aim for the small successes listed in Tip 2.Use existing data creatively and conduct analysisIt also requires effort.

When you can't collect the data you want, or when you don't have the required number of samples...Consider whether there is alternative data or whether you can analyze a smaller amount of data. 

You may think, "I have to have this data," or "I need more than x number of samples," but in fact, you may be able to analyze using different data, even if the number is small, without significant error.Data analysis is possible even if you don’t have perfect data  

(eg) Example of using other data when desired data cannot be collected

If you want data showing "changes in the daily movement of people in a certain area," you could collect GPS data or send people to the area to take measurements. However, these methods are very costly.

In this case, you can use data from Google Trends, a free tool that checks search volume and its trends. This is because you can estimate approximate daily population trends from search volume (*1). This will allow you to easily collect the data you need and significantly reduce costs.

(*1) Based on the assumption that many people will search for transportation methods, routes, and time from their destination, a search was conducted by entering "From XX (area name)" on the Google Trends homepage.

For example, the graph below shows the results of a search for "from Shibuya Station." After the first case of COVID-2020 was confirmed in Japan on January 1, 15, there was a major outbreak in Tokyo from late March to late May 19, and especially from April 2020 to May 3, 5, when the first state of emergency was declared. However, there has been an increase in the number of searches since around September 1, when the state of emergency and key measures were fully lifted. From this, it can be confirmed that there is a correlation between search volume and population inflow, and population inflow can be estimated.

Point) Points to note when using alternative data/analyzing small sample sizes

Not all data can be substituted unconditionally. When using alternative data or performing analysis with a small sample size, be sure to check the following in advance:

◆ When using alternative data

-Verify that there is a correlation between the desired data and alternative data (see example above)
- Check whether there are any abnormal values ​​in the alternative data

◆ When analyzing a small number of samples

・Check the "standard error" figure, which indicates the difference in accuracy between the number of samples required and the number of samples actually collected (*2)

(*2) The smaller the standard error, the higher the accuracy, and the larger the standard error, the lower the accuracy. Standard error can be calculated using Excel formulas, but this is a bit technical, so I will explain the details on another occasion.

Two recommended books for those who want to learn data analysis from scratch 

Kida, Hiromichi, Ito, Takeshi, Takashina, Hayato, and Yamada, Hiroshi (2020) "Becoming a data analysis talent. Aiming to become a 'business translator'" Nikkei BP 

This book was written by data scientists who are promoting digital transformation at Mitsui Sumitomo Insurance. It explains the important ideas that people in charge of data analysis at companies should keep in mind using a methodology called the "5D Framework." Since it was written with a focus on the business world in mind, this is a book that we would recommend not only to data analysts, but also to management and executives. 

Kaoru Kawamoto (2017) "The Strongest Data Analysis Organization: Why Osaka Gas Succeeded" Nikkei BP 

Kaoru Kawamoto of Osaka Gas, the first recipient of the Data Scientist of the Year award, talks about how he overcame obstacles from the launch of his data analysis organization to the present day. You'll get some hints on how to create an organization and get people involved.

*Link to Amazon product page is provided.

In conclusion　 

Over the course of five articles, we have introduced the basics of data analysis.

What I have been telling you through five times is,Data analysis is only meaningful if it is linked to business resultsAnd the data analysis is,Anyone can do it if they understand the basicsThat is to say.

Some people may think that "only experts can do data analysis" or "you need perfect data," but that's not actually the case. I would like you to try it out from what you can do. 

↓ Click here for a list of articles on "Data Analysis from Scratch"

#1 "8 steps of analysis" that beginners should know first
#2 Three analytical techniques that data analysis beginners should remember
#3 Communication tips to get management involved that data analysis beginners should know
#4 What data analysis beginners need to know: What management expects from data analysis and analysts
#5 4 tips for field staff who are new to data analysis to get management and the organization involved

↓ The business media "PIVOT" explains how to utilize data science in business.

.has-white-color{ color:#fff; }