Evaluate whether the article hits with the 'One Metric method' and make it a single index properly
GIGAZINE delivers various articles every day, but what worries me about creating articles is 'how can many people read them?' However, there are various indicators that show how much an article has been read, and it can be difficult to decide what to focus on when determining a 'good article.' GIGAZINE has created a mechanism to automatically evaluate articles by generating a single index from such various data, and this time I will explain it.
Various data related to access to pages can be obtained by using services such as
I used the following site to help me with this issue.
'One Metric', the Only Metric to Evaluate the Results of Web Content (Part 1)
To summarize the main points of this idea,
- Determine weights for some indicators
- Process the actual measurement value of each index so that it will be allocated points like that
- The final single index is the sum obtained by adding weights to the points assigned to each index.
Here, the determination of the weight is the most important point because it has implications for determining from what perspective the target is evaluated. In addition, if weighting is not taken into account, the evaluation balance will be out of balance unless each indicator has approximately the same weight (in other words, the weighting will not work correctly), so it can be said that processing the actual measured values is also quite important.
Using the previous example to simplify the series of operations, the figure below is shown.
To make it easier to understand, each index is processed to 100 points and weighted by an integer multiple, but in reality more detailed adjustments are required. For example, when adjusting values, there may be cases where the variance of each measured value should be considered, and if the weighting factor is too large, the meaning of using multiple indicators may be diminished. In the above example, the evaluation is determined almost solely by the number of impressions.
GIGAZINE performs daily batch processing using measurement values obtained from Google Analytics, and automatically calculates evaluation values for articles that have been accessed in the past month.
As an implementation content, in Python
- Get actual values from Google Analytics
- Aggregate measured values by article/measurement date
- Calculate the evaluation value by processing and weighting the aggregated actual measured values and then totaling them
- A job controller that controls the execution of each script
Regarding the content of data processing, in the case of GIGAZINE, the deviation value is obtained for each index, and the average value becomes 90 by raising the value (in other words, if the calculated value is 100 or more, it can be said that it is excellent in that index) I am making adjustments.
In addition, the indicators used for evaluation and their weighting settings are as follows.
In addition, we provide web tools for editorial staff so that they can quickly check the evaluation values that are updated daily.
In addition to directly creating articles at GIGAZINE, we are also doing development work to support article creation in parallel, and we are looking for people who are interested in such content from the following. We are waiting for your application.
GIGAZINE recruitment information. – There are things that can be done because it is GIGAZINE.
https://gigazine.co.jp/
Related Posts:
