Top 10 methods of data analysis, the last one is amazing!

In the Internet age, there is an explosion of information and a dazzling array of things. However, the ones that are really useful are not necessarily those that look cool. Many methods are simple, but they solve a lot of problems.

The following ten methods are the most classic methods that I have used when doing Internet operations analysis over the years. If you are familiar with these methods, you will have mastered the core parts of Internet operation analysis. It’s really not that complicated.

We will talk about them from the tenth method backwards. There is no order of importance, but the last one is often the most important.

Method 10: Traffic marking of Link Tag

Link tag marks the source of traffic and is definitely the most basic and important method of all. This method is not only applicable to the traffic source of the website, but also to the monitoring of the app download source (but the latter needs to meet certain conditions).

Link tag means adding a trailing parameter to the outbound link from the traffic source (i.e. the access URL). Not only do these parameters not affect the link jump, but they can also indicate the traffic source to which the link belongs (theoretically, the number of attributes that can indicate traffic sources is infinite).

Link tag cannot work alone and must work in conjunction with website analysis tools or app analysis tools.

Link tag is the basis of traffic analysis. To seriously analyze traffic, not only conventional analysis but also attribution analysis, you need to use the link tag method.

Method 9: Conversion Funnel

The basic model for analyzing conversions is the conversion funnel, which everyone should be familiar with.

The most common conversion funnel is to set the final conversion as the realization of a certain purpose, the most typical of which is sales, so people often confuse conversion and sales. But the final conversion of the conversion funnel can also be the realization of any other purpose, such as using the app for more than 10 minutes at a time (session duration >10minutes). Building funnels is the most common task for growth hackers .

The funnel helps us solve two problems. First, whether leakage (i.e. loss) occurs in a process. If there is a leakage, we can see it in the funnel and plug the leakage point through further analysis. Second, whether other processes that should not appear in a process appear, causing damage to the main conversion process.

Building a funnel is simple and is one of the most useful methods, whether it is web or app. But there are many secrets to using funnels. Moreover, the funnel method can be mixed with other methods, which makes it fun.

Method 8: Micro-conversion

Everyone understands the conversion funnel, but not everyone pays attention to micro-conversions. However, it is too difficult to expect a conversion funnel to continuously improve the conversion rate , but micro-conversions can do it. The conversion funnel solves the big problems in the conversion process, but big problems are always limited. After these problems are solved, you still need to continue to optimize your conversion. At this time, micro-conversion must be used.

Micro-conversions refer to various elements that are outside the necessary conversion process but can also have an impact on conversions. The interaction between these elements and users influences their feelings and directly or indirectly affects their decisions.

For example, some pictures of products are not necessary to see in the conversion process, but will their existence affect the user's purchasing decision? These images are micro conversion elements.

Personally, I think studying micro-conversions is more fun than studying conversions.

Method 7: Combine similar items

Combining like items is a common method that is often overlooked. We tend to pay great attention to segmentation, but sometimes we need to understand the more macro performance.

Combining like terms is one such method. For example, let me ask you, for an e-commerce website, what is the overall performance of all product pages? What is their overall bounce rate, dwell time, user satisfaction, etc.? Can you answer this?

If we check the performance of each product page and then add up the data of all the pages one by one for analysis, it would be too troublesome (it would be impossible to perform the analysis at all). At this time, we must combine like terms.

How to merge? Use the analysis tool's filtering tools or search and replace functions.

(PS: You don’t need to consider tools that don’t support such functions, because they should not be included in the professional equipment box of growth hackers at all).

There are many uses for merging similar items. For example, if you want to understand the overall performance of a section (channel) of a web or app, or if you want to understand the usage of the entire navigation system, this is a must-use method.

Method 6: A/B Testing

The logic of optimizing operations and products through data is simple - see the problem, come up with an idea, make a prototype, and test and finalize it.

For example, you find a hole in the conversion funnel, so you think that the price of the product must be wrong, which makes people not want to buy it. You see the problem — the funnel — and you come up with an idea — change your pricing.

But whether this idea is reliable or not is not something you can come up with. It must be used by real users. So you use AB testing, some users still see the old price, and other users see the new price. If your idea really works, the new price should convert better. If so, the new price is set (finalized) and begins operating at the new conversion level until you discover a new problem that needs improvement.

One of the main ideas of growth hacking is not to make something big and comprehensive, but to constantly make small and precise things that can be quickly verified. Quick verification, how to verify? The main method is A/B testing.

In today's Internet world, due to the end of the traffic dividend era, the requirements for rapid iteration have greatly increased, which also makes us pay more attention to the power of testing.

It is easy to perform A/B testing on the web, but it is much more difficult on the app, but there are still many solutions. Classic foreign apps and profitable games are A/B tested almost every day.

Method 5: Heatmap and comparative heatmap

Heatmap is a feature that everyone likes. It is the most intuitive tool to record the interaction between users and product interfaces. However, when it comes to real use, perhaps few people really delve into it!

Heatmaps are very important for web and app analysis! Today's heat maps have greatly improved their functionality compared to past heat maps.

On the web, some problems that were difficult to solve in the past, such as only being able to see whether a link has been clicked, misplaced click positions, marking clicks on floating parts, marking outbound links, etc., now have good tools that can provide many new ways to solve them.

On the app side, there are two situations: content apps have a weaker demand for heat maps, but tool apps have a significant demand for heat maps. The former’s screen is mainly composed of parallel content, and the content changes dynamically, so the application value of heat maps is not high; the latter especially needs to reflect user usage habits through heat maps, and combine with other engagement analysis within the app (in-app engagement) to optimize functions and layout design, so heat maps are very important to them.

To use heatmaps well, a very important point is that you can hardly solve the problem by using a heatmap alone. I often use the method of concentrated comparison heat map.

First, comparative analysis of various heat maps, especially comparative analysis of click heat maps (touch heat maps), reading line heat maps, and screen pause heat maps;

Second, comparative analysis of heat maps of segmented populations, such as heat map comparisons of different channels , new and old users, different time periods, AB testing, etc.

Third, interactions of different depths will reflect different heat maps. This is also a case where it is worth taking advantage of the heatmap comparison feature. For example, comparative analysis of click heat maps and conversion heat maps.

In short, heat maps are a great tool for analyzing many user interactions, but there’s more to heat maps than meets the eye!

Method 4: Event Tracking

A very important basis for Internet operation data analysis is website analysis. Today's app analysis, traffic analysis, channel analysis, and attribution analysis to be discussed later, etc., are all developed on the basis of website analysis.

However, one feature of early website analysis is that it can only record one type of user interaction behavior on the page, which is clicking on the http link (clicking on the URL). With the development of technology, there are not only http links on the page, but also many flash links (flash is being eliminated now), JavaScript interactive links, video playback, links to other web or apps, etc. Users' clicks on these things cannot be recorded by the old method.

However, if there is a problem, there must be a solution. People invented event tracking to solve the above problems. Event tracking is essentially customized monitoring of these special interactions, and because it is customized, it has the added benefit of being able to add additional descriptions of the activity (in the form of additional properties of the event tracking method). As a result, this method has even become a bit counterproductive. Even for some http links, many analysis veterans like to add event tracking to them (which is technically feasible) to obtain more additional monitoring attribute descriptions.

With the emergence of apps, due to the special features of apps (small screens, more emphasis on completing interactions on one screen), the importance of analyzing jumps between app pages (actually, the app’s screens) is not as great as the importance of analyzing jumps between pages on the web. However, the importance of analyzing click behaviors on apps is enormous. This means that when we analyze in-app engagement, we must rely heavily on events and relatively less on screens. This means that on the app side, event is the main thing, and page (or screen to be more precise) is the secondary thing!

This is why you must master this method.

Method 3: Cohort Analysis

Cohort analysis does not yet have a single, universally accepted translation. Some say it is cohort analysis, some say it is generational analysis, and some say it is cohort time series analysis. You can refer to Wikipedia :

https://zh.wikipedia.org/wiki/%E9%98%9F%E5%88%97%E7%A0%94%E7%A9%B6, find a translation that you think is suitable.

Regardless of the name, cohort analysis has become very important in the field of data operations . The reason is that with the decline of the traffic economy, intensive Internet operations require careful insight into retention . This is also the greatest value of cohort analysis. Cohort analysis compares the retention of comparable groups with exactly the same properties to discover which factors affect short-, medium-, and long-term retention.

Another reason for the popularity of cohort analysis is that it is very simple and intuitive to use. Cohort uses only a simple chart, without even using arithmetic operations, to directly describe the changes in user retention (or churn) over a period of time. Cohort can even help you make predictions.

I always feel that cohort analysis is the most typical method that embodies the beauty of simplicity.

Method 2: Attribution

Not everyone has heard of attribution, and even fewer use it well. However, considering that people’s decision to buy something may be influenced by multiple factors (digital marketing media), such as seeing an advertisement to learn about the product, using search to further understand the product, and then seeing the official account of the product on social channels, etc. The combination of these factors makes a person determined to buy.

How to understand this sequence or interaction between digital marketing channels? How to set up a reasonable digital marketing channel strategy to promote this relationship? When evaluating a channel, how can we take attribution into account so that we can measure it more objectively? All of these require attribution.

If you are responsible for Internet marketing , attribution analysis is an essential analytical method. For more details, please refer to Aiqi’s previous article: Growth of SEM Director | Attribution Model in Marketing Activities

Method 1: Segmentation

Strictly speaking, segmentation is not a method; it is the origin of all analysis. So it deserves to be ranked first.

My motto is, I would rather die than live without distinction. How can you do analysis without segmentation?

There are two types of segmentation, one of which is segmentation under certain conditions. For example: a visit (session) that stays on the page for more than 30 seconds; or only visitors from the Beijing area, etc. It’s actually filtering. The other type is the intersection between dimensions. For example: new visitors in the Beijing area. That is segmentation.

Segmentation helps us solve almost all problems. For example, the construction of the conversion funnel we talked about earlier is actually to subdivide the conversion process into steps. The analysis and evaluation of traffic channels also requires extensive use of segmentation methods.

The intersection between dimensions is a segmentation method that better reflects a person's analytical level. For example, my friend Sun Wei (data manager of Truck Home) submits user feedback as an event tracking attribute (placed in the event action attribute) to GA, and then in a custom report, he crosses user feedback with other user behaviors to see the behavioral trajectory of users with a certain type of feedback, and thus infer what problem has occurred.

When analyzing bounce rate, we also cross-check the landing page and its traffic source to check whether the high bounce rate is caused by the landing page or the traffic source. This is also a typical application of cross-dimensional segmentation.

I would rather die than not be divided.

The author of this article @Website Analysis in China compiled and published by (Qinggua Media). Please indicate the author information and source when reprinting!

Product promotion services: APP promotion services, information flow advertising, advertising platform