User Experience Design: The most commonly used statistical method in work

" statistical methods used in user experience design " see the watercress on the netizen raised the question, see the answer is not many people, can not help to write the following content.

What statistical methods are most commonly used in the work? Give some examples from my own experience.

1. Estimate the overall user situation through a subset of user samples. For example, your site users have 2 million registered users, you need to ask them about the improvement of the site, you have to send them an email questionnaire. But because of the restrictions, you cannot send every user, and you have received more than 1500 valid questionnaires. Make it clear how much more than 1500 questionnaires you have collected can represent a total of 2 million. At this point, you have to estimate the error range (margin of errors). Can you say that half of the people you support are supported if you have an effective 52% support program? Of course not, because your error range may be about 2.5%. Of course the range of error is related to the confidence you have chosen (confidence level), I will not elaborate here due to the time limit.

2. According to the user's behavior and characteristics of the user segmentation. It is necessary to use the concept of clustering to classify individual users with similar behaviors and characteristics by similarity calculation. Comparison of commonly used classification algorithms, such as K-means. For example, you do a business-to-business ecommerce site. You will consider which criteria to use to divide the user is better, such as the size of the manufacturer, the product belongs to the industry, or his business model (manufacturers, foreign trade companies, wholesalers, retailers) and so on.

3. Web design and business objectives of the relevance analysis. Web sites often make improvements, and if you have relevant historical records you can analyze which design improvements improve the user's activity or conversion rate. With the correlation analysis, you can see which design improvements are more active or transformational for the user, and you can also see which design elements improve each other to promote business goals, which are not working, or counterproductive. (when, these analyses are based on what you've done to eliminate the disruption to your operations and promotional activities.)

4. The overall user experience of the site measurement. In the face of hundreds of dozens of of web site data every day indicators, I believe that a large number of people, including bosses are not understand. It is said that Google's various indicators have more than 2000. Perhaps no one can be completely clear about how important these data metrics are to the measurement of user experience. At this time, the site needs a similar stock index indicators to measure the overall user experience of the site, and its changing trends. This requires a long-term tracking of the data, statistics, correlation comparison can be completed.