Product and operation, why can’t you do conversion analysis well?

This article will share with you the necessity and importance of conversion analysis, how operations and products conduct conversion analysis, how to break down the conversion steps, and the four major reasons for user loss during conversion and corresponding solutions.

Why do we need conversion analysis?

For a product, how to better convert users is the key factor in whether users can stay and create value.

For operations, it is necessary to open up a wide range of channels to find target users, and even select some paid channels (as shown below) for delivery in order to achieve better results. Here we need to evaluate and measure the effectiveness of the channel in bringing in volume. We need to know not only the quantity but also the quality. Otherwise, the question will arise as to why there is a daily increase in users but the benefits are still poor.

The quality of users coming from different channels is different, which will also affect their conversion within the product. At this time, the product manager has to think about it. In addition, he also has to think about how to enable users to better use my products and better improve product conversion. There are many practical problems here, such as:

• Why is it that the number of product users still cannot be increased even though operations are redirecting traffic every day?
• How to make product conversion (service/product purchase) better?
• How is the effect of this new feature?
• Which of the two product solutions is better?

Therefore, the conversion effect is closely related to every operation and product manager. Many of the things we do are ultimately aimed at achieving better product conversions. However, conversion is not a one-step process, and optimization of each link may bring better results.

Take user behavior in transaction product processes as an example: traffic comes from various channels and reaches our landing page. Interested users begin to browse the page and even start the purchase process until the purchase is finally successful. In this process, marketing, product managers and operations must make full use of tools to achieve better conversions.

Next, we will disassemble it according to the conversion process:

First of all, make sure the source of traffic is good . If the source of traffic is problematic, it will be difficult to convert. For example, the users who come are not the target users at all. We see that the users come, but they cannot be converted.

Secondly, ensure that users’ conversion within the product is smooth , and they do not become lost users due to issues such as module settings, bugs, etc.

How to conduct conversion analysis of channel traffic?

When we first measure the channel's ability to divert traffic, we may look more at the amount of traffic. Next we can see more data indicators, including bounce rate, dwell time, and page views to measure traffic. However, these indicators are auxiliary indicators. The bounce rate of a channel may not be low, but you cannot guarantee that the conversion rate of this channel will be good.

Therefore, we need to associate traffic sources with conversion data, which is the basis for refined operations.

1. Channel conversion analysis

The conversion here can be product conversion, such as registration, or successful order payment, or some user behaviors used to judge and analyze the quality of the channel: collecting products, liking, reading comments, etc. The more actions a user takes within the product and the more conversions they make, the more likely they are to stay.

At this point, we need some data to make an evaluation.

2. Distinguishing malicious traffic

The process of monitoring channel traffic also involves malicious traffic that operators hate, and how to distinguish low-quality traffic from high-quality traffic, because malicious traffic always has some characteristics, such as a certain group of people accessing the site in a concentrated manner at a certain time, relatively fixed hardware devices, using specific browsers , etc.

Last year, some APP companies found through data analysis that if a large number of iPhone 5C users appeared in a certain channel, there might be a problem. In order to make the fake look more genuine, some traffic-boosting companies would use real phones to cheat, but considering the cost, they would choose the relatively cheap iPhone 5C, so that this model appeared frequently.

Malicious traffic always leaves traces, so how should we investigate it?

Step 1: Multi-dimensional comparison to find related features

The picture below shows the conversion situation of each channel. The conversion efficiency of the first step of channel 3 is very low, so there may be a problem at this time.

We took out the users from channel 3 and segmented them using other dimensions, such as region, browser, etc. When we looked at the conversion rates of different browsers, we found that the conversion rate of a certain browser was also abnormally low.

Step 2: Cross-compare related features

After finding the browsers that might have problems, we did a distribution of browser versions and discovered that among the browser versions used by users from this channel, some older browser versions had increased dramatically.

Step 3: Confirm through the time dimension

Normal access time should be M-shaped, with two peak periods in the morning and evening, but midnight is when traffic is the lowest. Channel 3 users have access time 24 hours a day, and even in the early morning and midnight it is very high, so we can basically lock in that this channel has problems.

Therefore, we need to integrate traffic conversion data, conduct channel evaluation, and use data analysis tools to locate and discover problems.

How to break down the transformation steps?

When users come through various channels and enter the product, we begin to break down the conversion steps within the product and look for room for optimization.

There are two ways to break down the transformation steps:

• Vertically, we break down the process to study the number of completed registrations and the number of successful purchases;
• When breaking down horizontally by dimensions and population groups, the conversion results are different from expectations, so horizontal breaking down and comparative analysis are carried out.

We have heard many users ask questions like this before: Is my conversion rate high or low in the industry? Is there room for optimization? In fact, we don’t necessarily need benchmark data. On the one hand, a lot of such data is company confidential. On the other hand, the third party’s calculation method may not be the same as yours.

Without industry data, your own conversion data contains a lot of information. You can analyze it from other data dimensions such as time dimension (analyze daily conversions, analyze activities and daily conversions), platform dimension (iOS and Android situation), etc., and you can find a lot of room for optimization.

Four major reasons and countermeasures for user conversion loss

The reasons for loss at different conversion steps vary. Some are due to lack of appeal, and sometimes there are problems encountered during use.

Although the reasons for churn vary, there are four main categories of churn:

• Mismatched requirements
• Product features/services/goods do not meet expectations
• Poor interactive experience
• "Mysterious" reasons

Next, we will analyze these user churn operations one by one and explore solutions based on cases.

Category 1: Mismatched Demand

There are many situations of demand mismatch. One is that the product can stimulate user demand, but the user does not see it; the other is that the product simply does not meet the user's needs.

If our product can stimulate user demand, but the user does not see it. At this time, we need to let users see the right content in the right place. We can analyze the user's click behavior on the website through key location data statistics and heat maps to find the golden location and put the most suitable content.

If the user's needs are not met, then it is necessary to analyze this part of the needs. In fact, more users have experienced no search results than you think. We can find out the specific unmet needs of users by collecting search term information for websites that have no result pages. At the same time, compare it with the number of searches in the search box to measure whether special content, services, and products are being developed to meet this demand.

Category 2: Product functions/services/goods do not meet expectations

When a product’s functions, services or products do not meet user expectations, you can analyze whether the various functions within the product that help convert users are not playing a positive role. At this time, users who have performed a certain operation can be grouped together. For example, an e-commerce platform can group users who have viewed a product review page to verify whether the review promotes the final conversion.

As shown in the figure below, the conversion rate of users who have read the comments and completed registration (on the right side of the figure below) is 53.1%, which is higher than the overall rate; if it is lower than the overall rate, then there may be problems with some settings in the comments.

Category 3: Usability Interaction Experience Issues

A common situation is that the device or browser is not compatible. When differentiating by browser or device dimension, the conversion rate is very low, and the problem should be discovered, located and repaired in time.

Category 4: Other reasons for user loss

After checking, if you find that there are no problems with the above, you need to check the user's original access trajectory. Some users have completed most of the conversion path, but gave up at the last step. If there are many such users, you need to look at it in combination with specific products.

If the average order value is too high and the user hesitates at this step, we need to give the user a push. It is necessary to connect user behavior and user ID, and then carry out further targeted operations; for example, send some vouchers or discount coupons to these users to stimulate conversion.

Conversion analysis is a systematic task that involves all aspects of products and operations. The prerequisite for improving conversion is to fully break down the conversion steps, evaluate the conversion effect from the source of traffic, and locate the problem with the help of multi-dimensional data.