Reading out of the same data in different data analyst conclusion may be different, even completely opposite, but the conclusion itself is not right or wrong, so the objective data from subjective to people, need to have some scientific method of analysis as a bridge help information to better data, more comprehensive, faster delivery. So, for the product, commonly used data analysis methods, what does? Today we explore together with you by Chen in the big data cube it!
Product Data Analysis
A trend analysis
Trend analysis is generally used for long-term tracking of core indicators, such as: hits, GMV, the number of active users. Generally made simple data trends, but the light made into a data trend analysis is not too, have to be seen on which the data changes on the trend, there is no cyclical, there is no turning point, and analyze the reasons behind it, whether it is internal reasons or external reasons.
The best trend analysis is the ratio of the output. There chain, up, than the fixed base. For example, in April 2017 compared with March GDP growth number, which is the chain, the chain reflects the recent trend of change, but there are seasonal influence. In order to eliminate the effects of seasonality, it launched a year, for example: in April 2017 increased by much more than the 2016 April GDP, this is the year. Fixed base is better understood than is a fixed point, such as the January 2017 data as a starting point, fixed base ratio, compared with January 2017 data and data in May 2017 to do comparison.
Second, comparative analysis
Horizontal comparison: horizontal contrast is better than with their own. The most common is the need to index data with the target ratio, to answer we have not completed the goal; with us last month than to answer our growth in the number of North Central.
Vertical contrast: in short, than with others. We'll talk to competitors than to answer our share and position in the market is like.
Common applications are comparing A / B test, the key A / B test is to ensure that the two groups only a single variable, other conditions remain the same. For example, the home page revision test results, we need to ensure that the same sources, the same user quality, on-line time remains the same, so test out the data makes sense.
Third, quadrant analysis
Depending on the data, comparing the respective body is divided into four quadrants.
General p2p products are all registered users of third-party channels drainage, if the total amount in accordance with the quality and quantity of traffic sources can be divided into four quadrants, and then select a fixed point in time, compare the various channels of traffic cost-effective, quality can be retained with this dimension as a standard. For high-quality high number of channels continue to maintain high quality for a low number of channels to increase the number of introduced, the number of low-quality low-pass, high-quality low number of strategies put in to try and requirements, such quadrant analysis allows us to compare the analysis of time to get a very intuitive and quick results.
Fourth, cross-analysis
Comparison of both lateral comparison, another vertical contrast. If you want both horizontal contrast, want vertical contrast, there is a cross analysis. Cross analysis of the data is from a plurality of dimensions of cross-presentation, multi-angle analysis of the binding.
The main role of cross-analysis of data from multiple dimensions is broken down, and found the most relevant dimensions to explore the reasons for changes in the data.
Common dimensions are:
Sharing: Are there different time data changes.
Sub-channels: Are there different traffic source data changes.
User points: whether there are differences of new registered users and old users compared to users of high-grade and low-grade users whether there are differences in comparison.
Region: data from different regions if there is a change.
Cross analysis is from a coarse to fine process, also called segmentation analysis.
Trends contrast, the quadrant, the cross section containing the most basic analysis of data. Whether it is to verify the data, or data analysis, trend identification, do comparison, designated quadrant, do segmentation, data can play its due role. Hope that through sharing the above, the data can help analysts make better data analysis.