The most complete guide to building an operations data analysis system is here!

The operations department has always been the largest demander of data analysis, and students who have done operations analysis often feel that there are too many details and trivialities, and many problems are entangled with each other and difficult to sort out. So how can operations analysis be done in a systematic and efficient way? This article shares detailed content on this issue.

Starting from the characteristics of operations

Operations work is clearly different from sales and supply. For example, sales work focuses on revenue, profit, and conversion rate, and making money is enough. For example, supply work focuses on commodity supply, inventory turnover, and cost control, and not being out of stock and having less backlogs is enough. These jobs have focused goals and clear processes.

Operational work has many goals and flexible forms. For example, activity operation may directly promote sales conversion, or it may encourage users to be active and maintain platform popularity. For example, community operation may be purely for the purpose of getting back public domain users, or it may provide services or sales conversion based on the community.

It is more likely that, for example, when doing a large-scale event, you need to first do traffic promotion, then conversion, and then after-sales support. After an event, all types of work are related. This work feature requires you to pay special attention to the following four points when doing operational data analysis:

First: Clear goals

There are many operational goals, so it is important to have clear goals beforehand. Only with clear goals can we choose the right plan, allocate appropriate resources, and have a direction for post-analysis.

There are three common ways to set goals in operations:

1. Achieve an absolute target, for example: reach a total of 1 million users within 5 months.

2. Achieve ratio/proportion goals, for example: in May, the transaction conversion rate increased from 10% to 15%.

3. Achieve incremental targets, for example: in May, drive an additional 30 million transactions over and above natural growth.

Here, goals 1 and 2 can be directly observed with data, but goal 3 involves the definition of "natural growth", which must be discussed clearly in advance. Otherwise, it may be impossible to analyze afterwards. There are three common definitions of natural growth (as shown in the figure below). Each has its own advantages and disadvantages, and there is no perfect solution. Therefore, you must reach a consensus with your boss and various departments in advance.

There are often people who are lazy and do not set quantitative goals beforehand. They just say something in general terms: "to improve performance" or "to increase user activity"... and then try to use data analysis to distinguish between natural growth and activity results afterwards. As a result, they often fail to distinguish and end up shooting themselves in the foot. Everyone should take this as a warning.

Second: sort out the indicators

Operational work is flexible, so it is often necessary to set data indicators based on the actual workflow. This is to facilitate monitoring of execution progress and to observe which link has problems during review, so as to find the causes and opportunities.

For example, the operations department held a "sign in to get benefits" activity, where users can get rewards if they sign in 7/14/21 times a month. The rewards include coupons, hoping to stimulate user activity and conversion. At this time, the following process indicators should be sorted out to facilitate subsequent data tracking and review of the results (as shown in the figure below):

1. From what channel the information is output?

2. How many users are attracted to participate?

3. How many people completed each stage.

4. How many people consume after receiving the reward.

Note that if there are continuous operational activities, it is necessary to continuously track user participation. For example, after the above-mentioned clock-in sign-in continues for n months, there will be considerable data accumulation, and you can observe:

1. The overall impact on whether the number of users increases.

2. How many users participate repeatedly.

3. How many users never participate.

Different data trends can lead to different conclusions (as shown in the figure below). Combining the data trends can help us better judge whether an operating method should be continued or adjusted.

Third: Label well

There are many factors that affect the operation effect, such as promotion channels, promotional copy, activity form, operation steps, conversion products, discount strength, etc., which will affect the effect. Therefore, before starting to work, it is necessary to label key factors such as promotional copy, recommended products, and operation procedures, so as to analyze them afterwards (as shown in the figure below).

In addition to individual labels, you can also label the overall operation measures to judge the overall situation. For example, to increase user activity, you can use several methods such as receiving red envelopes, signing in, big turntables, scoreboards, etc. Each method can be configured with rewards. At this time, you can use labels to group and compare various methods, so as to understand the scope of the effect of each method and provide experience accumulation for subsequent operations (as shown below).

Fourth: Follow the map

If you do the first three steps well, it will be very easy to analyze operational data.

1. Compare with the goals, see how much has been accomplished, see if the investment has exceeded the budget, and make a judgment first: whether this time it is done well/poorly.

2. Compare the results of previous operational activities under the same goal to see whether this one is at an upper, medium or lower level.

3. Check the process indicators to see which link has problems: traffic diversion → acceptance → conversion.

4. Compare the conversion effects under different tags to see which method is effective/not effective.

This will output the conclusion.

In work, we often see that operational analysis has no conclusions because:

1. The goal is not clear, or there is no goal at all. There are only a bunch of data but no conclusion.

2. Few process indicators are collected, and we only know that the final conversion fails, but we don’t know why.

3. Due to the lack of labels, it is impossible to quantify operating methods and evaluate whether they are good or bad.

Only by being well prepared can you get good analysis results, remember this.

summary

From the above, we can see that in order to do a good job in operational analysis, we need to master comprehensive capabilities such as data indicator system sorting, label making, and analytical thinking in order to adapt to the requirements of various scenarios.

Author: Down-to-earth Teacher Chen; WeChat public account: Down-to-earth Teacher Chen