Update: Ontology

We added two features so you can define metrics more precisely. With percentile calculations, you can freely set any percentile you want, and by applying condition filters when creating a metric, you can build a metric that applies only to specific properties.

Why we need this

Averages alone make it hard to grasp the actual distribution of your data. When a few extreme values distort the average, the basis for your decisions becomes shaky.

Also, even the same metric sometimes needs to be measured differently depending on context. There are situations where you need to track only the revenue of a specific channel rather than total revenue, or only a specific type of inquiry rather than all inquiries. Previously, it was difficult to configure such granular metrics.

With this update, you can define metrics that are more accurate and more context-aware on your own.

Percentile calculations

Percentile has been added to metric calculations. You can set metrics that more accurately reflect the data distribution, such as the median, the interquartile range (IQR), and the Nth percentile.

• 50th percentile (median): Use it when you need a representative value that is not affected by extreme values. Measuring the median of support handling time prevents a few abnormally long inquiries from distorting the overall figure.
• 5th / 95th percentile: Use it to identify the boundaries of the bottom 5% and top 5%. Setting the 5th and 95th of order amounts lets you define a typical order range excluding outliers.
• 10th / 90th percentile: Use it to grasp the level most customers experience. If the 90th percentile of delivery time is 3 days, it means 9 out of 10 customers receive their order within 3 days.
• 25th / 75th percentile (quartiles): Use it to see where the middle 50% of the data is distributed. A large gap between the 25th and 75th of response time is a signal that the variation across agents is large.

Metric condition filters

When creating a metric, you can set the same condition filters used in search. You can define a metric that applies only to specific properties and track it directly on the dashboard.

• Track only the revenue of a specific channel: Set up own-mall revenue, marketplace revenue, and offline revenue each as separate metrics to compare performance by channel.
• Aggregate only a specific type of inquiry: Separate exchange/return inquiries, delivery inquiries, and product inquiries to monitor the trend of each type individually.
• Measure only the marketing performance of a specific campaign: Manage the conversion rate and ROAS of brand campaigns and performance campaigns as separate metrics to analyze efficiency by campaign type.

Filter conditions support both AND/OR combinations and filter groups. You can apply the same filter logic you use in search directly to metric definitions.