Using the Website Data Analysis assistant work to carry on the good design

Article Description: Data-aided design-practice in search.

Design can not be based on experience and intuition, because the target groups involved, the scene, the operation of the different habits. In order to obtain more accurate and effective information to assist and detect the design, the designer chooses the qualitative (user interview, focus Group) and quantitative (survey questionnaire, website data analysis) in the way of user research. The "Website Data analysis" This way does not need to spend a long time and human cost, while avoiding the user and the environment and other unstable factors of the analysis results caused by the interference. As long as we have accurate and applicable data, we should choose this method to assist design.

What data do you usually get?

1. Website data

The search for common data is as follows:

query– Search Key Words

PV (Page view)-pages view, every time the page is refreshed, it is computed once

UV (Unique Visitor)-Number of user visits

click-page Total number of clicks, each function will have a corresponding number of clicks

l->d-Search the list page to the detail page of the click Data, that is, conversion rate, different pages have different data.

CTR-CLICK/LPL,LPV is the search volume on the list page, CTR the number of clicks per browsing.

2, user interviews, qualitative research, focus groups

3, report on the findings

4, on-line testing (for example, A/b test, a buckettest that can be used in-house to develop multiple-solution online tests)

What information can be learned from the website data?

1 , key word loss rate analysis

Figure 1 is the user input "women's Shoes" related keywords and the corresponding UV loss rate (that is, no action on the search page of the number of users as a percentage of all search users), from the data to see the added leather, Guangzhou, fashion and other attributes of the key word loss rate is relatively low a lot.

The more detailed the keyword description, the more accurate the search match to the product, and the faster the user can find the target product. But it costs more to get users to enter keywords precisely (such as when users don't know which descriptors are more appropriate). How to reduce this cost? We can use suggestion (keyword recommendation) (see Figure 2) and the SN region (Class View property filter area) (see Figure 3) to give users the appropriate recommendation and guidance.

2 , rapid screening after the revision of data analysis

Figure 4 is the filtered item on the search. The goal of the search should be faster and more accurate to help users find products, the filter area is one of the important components, so that users find a faster filter and simple to complete the screening operation, is the central purpose of each revision.

where each filter item should be placed is more appropriate, depending largely on which dimension of information the user is looking for in the product. For the functions that have been on line, we can analyze through the data, such as the CTR data of the filter area, we can find that the user uses the area, the ordering, the unit price, the operation mode to operate more, explained the user to this aspect screening demand is bigger, also concerns these several dimensions information, This can be adjusted to facilitate the user to find the location, but also reduce the user's memory burden, because users are generally from left to right browsing, so you can adjust the important screening to the first or with visual highlights. And some low data filtering, can be hidden or offline according to the situation, but also increase the expansion of the filter area.

Figure 5, based on the data in Figure 4, we have adjusted the location of the filtered items and the way the sort buttons interact.

The new version of the line two weeks later we found that the user's attention to the screen adjusted to the left position after the CTR data significantly increased (green for the obvious increase in data, red data drop, other data a small increase).

To sum up, the design, iterative process is as follows:

Data validation, a relatively successful design this time.

3 , contrast function after online data analysis

At the same time, we look at an unreasonable product design (see Figure 7). The figure is Alibaba's contrast function over the last year, the user tick the product to add the contrast (Figure 7 the 1th step) only occupies the entire search ctr0.6% about, but to the last clicks the contrast button (Figure 7 2nd) The conversion rate only then 10% not to have, the contrast function utilization rate is very low.

According to this discovery, we have gathered 5 testers (1 product managers, 1 operators, 3 users) for focus testing, get the following feedback 1, users just look at a few of the information, do not need so multidimensional information comparison. 2, the user is more accustomed to by the point open detail to compare.

Many vertical industry search has a comparative function, such as the Pacific, Zhongguancun, Taobao mobile phones, and so on, the function through the comparison of information to help users select more consistent with the target products. But not suitable for our website, but also need to consider more. Combined with the above data, the comparison function is more suitable for the vertical industry which needs to pay attention to the multi-dimensional information contrast, and the attention dimension is less, the user can reach through the short-term memory.

[1] [2] Next page