Analyzing the merging and comparing measures of e-commerce conversion rate

For e-commerce, often we need to do analysis in conjunction with various basic indicators to carry out two calculations combined to get a can be used for comprehensive evaluation or comparison of the measures, this process needs to involve a number of indicators of the combination of skills, and baseline settings. In fact, the previous series of "data context" has repeatedly emphasized that we need to set a reasonable reference for the indicators to evaluate the trend of indicators or the performance of good or bad, previously provided a series of methods, the following approach should be the simplest and most convenient, without losing practicality, thanks to the "User experience metrics" in the Book of the introduction, So this article is more like a reading notes, the content of the basic collation summarized from the "User Experience Measurement" 8th chapter-Merging and comparison measures, of course, no longer limited to the user experience level, combined with the site analysis level of thinking.

Here I divide the content according to the differences between the merging and comparison types: Simple merge metric target comparison, percentage score metric mean comparison, comparison between standardized measurement groups and comparison with expert performance.

Simple Merge metric target comparison

This is the most concise and effective KPI evaluation model. First of all, a simple merger, for example, e-commerce site produces a lot of orders every day, these orders by a large number of Web site access to bring, once a visit produced an order, we said that the visit produced a transformation, so the most basic statistical indicators have no conversion rate, only the site's access volume and order number, conversion rate is by "order number /access counts. So why to calculate the conversion rate, the use of orders to evaluate the performance of the site is not also OK? Very simple, because the amount of the order will be affected by the amount of site traffic, and many times the amount of access to the site can not be controlled, so we can not say 100 visits generated 10 orders must be more than 150 visits generated 15 orders to be worse, If we use the combined metric of two indicators--the conversion rate--to evaluate it, it is obviously more scientific, because it is 10%, and the performance is equal.

Because the combination of measures, such as conversion rate, per capita consumption and other indicators are generally more stable, fluctuations will not be too large, we will generally use these indicators as a site KPI, and we will set a target for each KPI, such as our set site conversion rate target is 10%, So let's take a look at how much of a goal is achieved within one months of the site:

We can see that in the 31 days of May, the conversion rate of 26 days is equal to or exceed the target value, so that the target rate is 83.87% (26/31), but also a good result. Perhaps your team is complaining that the KPI is too harsh, after all, the data will be affected by a number of factors, to ensure that more than 10% of the day is indeed difficult, if you can fully achieve, that is the goal set too low, the goal should always be critical to reach and can not be reached between is reasonable So it might be more reasonable to use the goal-reaching rate here.

Percentage score metric mean comparison

But sometimes we can not set goals for each indicator, after all, the goal is more to control the overall performance of the site or KPI, for some based on the subdivision of the measurement, we need to use a different combination and comparison methods, so there is a percentage based on the score.

Equally simple, a percentage score is a form that converts the value of an indicator to a percentage, equivalent to how much it scores under the 100-point system. How to convert an indicator to a percentage value is a very simple method where all the indicators are at the maximum of the population, and this method is valid for all indexes that are greater than 0 and are not particularly discrete. For example, we evaluate the quality of the website merchandise, weigh the volume of merchandise and conversion rate of these 2 indicators, we know that the conversion rate itself is a percentage, but obviously also need to be transformed, So we divide the browsing volume and conversion rate of each product by the maximum value of the total view and the total conversion rate, respectively, to get the corresponding percentage score:

The above mean is based on a simple average calculation, we can also introduce weights to each indicator weighted average, for example, we give the browse volume and conversion rate respectively given 40% and 60% of the weight:

This allows us to search for high-quality products based on the final scoring mean.

Standardize the comparison between measurement groups

It's a bit too much to be standardized, but it does play an extremely important role in the field of index consolidation, and it also recommends reading-the standardization of data.

method is not described here in detail, or for instance, another noteworthy point is the "inverse index" treatment, the inverse index refers to the performance of the value is just the opposite of performance indicators, that is, the greater the value of performance, the lower the value of the better performance, the site analysis of the typical inverse index is bounce Rate. Because the standardization of the index after the average is 0 standardization is 1 normal distribution, so the processing of the inverse index only need to standardize the data multiply-1 on it, but also very simple. Here take the website landing page optimization For example, to see how to effectively evaluate the 3 Landing page optimization solution which is the best:

It also uses the method of finding the mean value, after the standardization of each indicator to compare the mean value (note that the results of the standardization of BR in this case multiplied by-1), we can easily see that the effect of a scheme is optimal, which is also a "goal decision" of the simplest application. It should be noted here that the distribution of the values after the standardization of the indicators is uncertain, unlike the percentage above must be in [0,100], so the normalized value itself does not have practical significance, only to put it into the comparative environment to have the value of analysis, so the standard method of standardization applies only to the comparison between groups.

Comparison with expert performance

If you can get a team of experts who are very familiar with the expertise and skills of the field, it will be much easier to evaluate the site, as the expert's use and scoring of the site can be the best standard for the site.

In fact, we generally think that the performance of experts is an idealized state, for example, experts can in the shortest possible time through the least operation to complete the expected task, so we can be the expert group of data as the site can achieve the optimization goal, and the current data and expert data is the difference is the current site optimization space, This analysis allows us to be clear enough in which modules are farthest from the optimal level, while optimizing the largest space, we can start from these aspects to achieve the fastest maximum ascension.

In addition to the above methods of measuring consolidation comparisons, the book also introduces the use of columnar line combinations and radar charts to show multiple indicators of the comparison environment, as these are mentioned in previous articles, this is no longer introduced here.

In fact, the above mentioned are the most simple and practical indicators of comparison processing methods, these methods no matter what kind of companies or what indicators can be applied, even based on these most basic percentage, average, target comparison and other methods can analogy expand more flexible and effective analysis methods. This is the time to use your intelligence to choose the most appropriate method for your data.