design should not be based solely on experience and intuition, because of the different goals, circumstances, and practices involved. In order to obtain more accurate and effective information to assist and detect design, designers will choose qualitative (user interviews, focus groups) and quantitative (questionnaire survey, website data analysis) way to conduct user research. Among them, "website data analysis" this method does not need to spend a long time and labor costs, while avoiding the user and the environment and other unstable factors on the analysis of the results of interference. As long as we have accurate and applicable data, we should give priority to this method to assist the design.
what data can we usually get,
1, site data
The common data for
search is as follows:
Query – search key word
PV (Page, View) – page views, and each refresh of the page is calculated once
UV (Unique, Visitor) – the number of user accesses
Click – the total number of hits per page, each function will have a corresponding number of clicks
L-> D – search list page to detail page click data, that is, conversion rate, different pages have different data.
CTR – Click/LPL, LPV, that is, the search on the list page of the browser, CTR, that is, the number of hits per visit.
2, user interviews, qualitative research, focus group,
3, reports that have been concluded,
4, on-line testing (such as A/B, test, search, commonly used in internal development of multi program on-line testing buckettest)
What information can be found in the
1 keyword loss rate analysis
Figure 1 is the
user input "shoes" related keywords and corresponding key UV loss rate (the ratio of the number of users accounted for all of the search user is not in the search page to carry out any operation behavior), data from the point of view of key words added leather, Guangzhou, etc. are attribute words much lower turnover rate.
The more detailed the
keyword description, the more accurate the search matches the product, and the faster the user can find the target product. But it’s more expensive for users to accurately enter key words (such as the user doesn’t know which descriptors to use better, etc.). How to reduce this cost? We can use suggestion (keyword recommendation) (see Figure 2) and SN area (class >)