How does operations perform data analysis? 3 ideas and 8 methods!

I've been reading "Chief Growth Officer" recently, and some of the content is very instructive, so I created a new collection and put some reading notes here. I will not include the more common ones like the AARRR model . I will mainly record some methodologies or thinking frameworks that I find refreshing. 1. Three ideas for data analysis

According to the examples given in the book, these three methods may be more suitable for macro-level decisions, such as decisions when placing products on multiple channels . When it comes to more precise analysis of landing points, the methods may need to be adjusted, such as determining whether the button should be placed in the upper right corner or at the bottom.

1. Basic data analysis methods

Business-centric data analysis should start with business scenarios and end with business decisions . This is also a relatively common method. The following five steps can be used for inference in many scenarios.

The book mentions five basic steps: exploring business implications - developing an analysis plan - splitting query data - extracting business insights - making business decisions.

I haven’t thought of a good example for this method, but there is a good example about multi-channel delivery decision-making in the book. I recommend you to take a look at it.

2. Internal and external factor decomposition method

The scenario of this method is to find the factors that affect the North Star indicator, that is, to distinguish them from two dimensions (internal factors + external factors, controllable factors + uncontrollable factors), and divide the influencing factors into four categories.

This is essentially an exhaustive enumeration. The biggest drawback of this method is that it is prone to omissions. Regardless of whether the analyst is a single person or a team, omissions are inevitable. Of course, the advantage is that by distinguishing the influencing factors from four intervals, you can conduct a more intuitive analysis, so that you can prescribe the right medicine in a way that you can control.

3. DOSS idea

Detailed Question

Overall Influence

Single Answer

Scaled Solution

8 methods of data analysis

Suppose you have all the data of the product in front of you, including the general data such as DAU, new additions, retention , channels, etc. in the past year, as well as the UV/PV of all pages of the app from the opening screen to the core functions, and the UV/PV of all CTAs. In other words, all the data about the app can basically be obtained. Now you want to build a data system. What are the ways?

1. Numbers and trends

The most basic form of data presentation can be accomplished using Excel.

Key value. Trend charts, histograms, line charts, stacked charts, pie charts.

2. Dimensional decomposition

When a single value or trend is relatively macro, it needs to be broken down.

For example, daily additions can be broken down from the dimension of channels, such as App Store , 360 Mobile Assistant , etc. Based on this, we can view the additions in different dimensions and make some decisions on channel promotion ;

The daily additions can also be broken down from the dimension of time. For example, if the peak time period for new additions is found to be 1 p.m., then operational activities can be considered to be concentrated at this time.

3. User Segmentation

Group and categorize users who meet certain specific behaviors or labels, such as "new users of App Store from Beijing in January."

We can conduct in-depth analysis on users in this group, such as their hobbies, consumption levels, and high-frequency behaviors, and then carry out targeted user operations or marketing promotions , such as issuing coupons to "users who put items in the shopping cart but did not pay."

4. Conversion Funnel

This should be the most well-known method. Basically all user behaviors can be expressed by funnels, whether it is a registration conversion funnel or an order payment funnel.

Focus on three issues:

- What is the overall conversion rate from start to finish?

- What is the conversion rate for each step?

- Which step has the most churn, why, and what characteristics do the churned users have (user segmentation can be used for detailed analysis here)

5. Behavior Trajectory

From the UV/PV values ​​of several pages, we can only see the overall conversion rate, but there may actually be deviations. By looking at the user's behavior trajectory, we can understand the product from a more practical perspective.

6. Retention Analysis

Focus on two types of retention:

-New user retention : next-day retention, 7-day retention, 30-day retention, and retention rate change trends.

- Function retention: users who use function xx will perform the operation again the next day

7. A/B Testing

There is no need to explain the significance, but the number of people who have actually implemented A/B testing should actually be limited.

There are two essential factors for A/B testing:

- Allow enough time for testing.

-High data volume and data density. When product traffic is not large, the statistical results of the test are actually highly random.

8. Mathematical modeling

When a business goal is related to information such as user behavior and user portrait , mathematical modeling, data mining and other methods can be used to build a model and then conduct predictive analysis.

This is a relatively advanced operation, which requires the company to be of a certain size and have sufficient budget investment.

Source: Rockelbel