How to guide product optimization with data

A designer is born to solve a problem. How can you solve the problem better and more efficiently? First of all, we can find the problem, data analysis is a common method of discovering problems. Through the data location problem, and then use the design scheme to try to solve the problem, then use the quantitative data metrics to assess whether the problem solved, how much to solve. By iterative optimization, the problem can be solved better.

This article unifies oneself in the login product experience optimization accumulation Some actual combat experience, the reappearance process design drip, has the effect obvious plan, also has the effect not obvious optimization attempt, finally will summarize some general design idea.

• What is the reason for login failure?

Login box is generally a simple interface, it includes: Account name and login password, check code, other account login entry, registration and other related operations and other elements.

Seemingly a simple interface, every day a lot of users try to log on but failed to log in. Some sites with large traffic may have millions of login failures per day. Logging in as a user enters the first door of a product, the experience is critical. Therefore, we use the login success rate as the most important index to measure the quality of the login experience.

After data collection and analysis, we found that the user login errors for example (the figure is blurred): Login password error, account name error, check code error is the TOP3 factor that affects the success of the login.

• Detailed positioning problem, conquer

In view of the above top3 problem, we have made an iterative optimization. In this case, the optimization method is illustrated by the example of error checking. We identified the big problem in front of the data, and then we need to locate the problem in detail. Why is the check code wrong? What are the solutions?

Click on the data and background log through the page we found that the user refresh picture Check code concept is high, the average user will refresh 2 times check code to enter the correct. So the problem can be summed up as: Check code recognition is low, easy to lose.

What kind of solution should be? Try this progressive approach: beforehand , in the event, in the aftermath .

1) Beforehand: We put the input picture check code as an event, that "beforehand" means that does not appear the picture check code, or reduces the check code to appear the probability. The role of the picture check code is to prevent the machine bulk login behavior. Reverse thinking: If we can ensure the security of the system, by technical means as far as possible to identify the person in the request login, or the machine malicious login. Be sure to identify as a person in the case of login, no picture check code (such as), you can greatly reduce the normal user login to enter the picture check code concept.

After the online, by tracking the login success rate data, found that this optimization effect is very obvious.

2) In the matter: Nearly one step found that the check code is the number and the letter at any time mix and match appears, such as the number "0" and the letter "O", the number "1" and the letter "I" is easy to confuse.

Solution: Reduce the probability of confusing numbers and letters, and if the number "0" or the letter "O" appears, the default user input is correct. If the conditions permit, you can also filter out these confusing combinations to increase the success rate of the first input of the user.

There is also a workaround for the user to enter the check code after the instant feedback results: input correct or error.

Enter the correct prompt as follows:

The following are the tips for entering errors:

Through this optimization, the user can know in advance that they have made a mistake, and do not need to log in the request to know the error, resulting in a single login failure. So the user login is more efficient (time is shorter), and the login success rate also has a small increase.

3) After the event: the user lost a picture check code, re-input how to be more successful? As in, the user by clicking on the picture Check code area refresh appears another check code, but relatively obscure, some users error, do not know click to change a check code picture.

Solution: Enhance the Operation button to refresh the picture check code, and add the Voice check code. Although the Voice check code is intended for the blind user design, but normal users in the picture check code is always wrong, you can also use the Voice check code.

Finally, we summarize the general process of product optimization through data.

The first step: determine the product experience of the quantitative indicators

As in the above case we use the success rate to measure the sign-in product experience good or bad index. What metrics can be used to quantify the experience of a product? This is the topic of interactive design in the field of more discussion, not to start the elaboration, only to provide reference suggestions: first, the user experience changes to the index has a critical impact; second, the indicator is very good quantified; Thirdly, the project team reached a consensus on how to calculate the indicator, so that a more targeted design could be carried out.

Step two: Accurate positioning of data analysis

Find out where the product issues affecting the above-mentioned measurements are, and what the ratio of login failures in the above case is. Depending on the product, the data source can be the background log data of the product, the user clicks the behavior data or the user calls for help rate, and so on. Another step closer Analysis: Why the password error, account name why error, picture check code why error. For example, through a large number of users call Customer service recording can find the real reason for the user password error: Confuse the payment password and login password, and the secret of other products, the password is too complex to remember, password error, the password is unsuccessful, and so on.

Step three: Iterative optimization, using data to test the effect

To find the problem, we have to design the project. The solution to the idea can be referred to the above "beforehand", "in the matter" and "afterwards". Beforehand: Whether the problem can be avoided in advance, reduce the concept of occurrence; In the matter: The problem has occurred, how to help the user quickly solve, afterwards: how to avoid recurrence. The optimization program goes online quickly to see if the corresponding experience metrics have changed and how much has changed. It is often impossible to solve the problem one step, need iterative optimization, and constantly revise the design strategy through data tracking. Achieve our final design goals by accumulating results.

How to guide product optimization with data