E-Commerce Index Analysis

Architecture

Log

👉 The log is generated by a simulator

{
  "common": {
    "ar": "110000",
    "ba": "iPhone",
    "ch": "Appstore",
    "md": "iPhone X ",
    "mid": "mid_41",
    "os": "iOS 13.2.3",
    "uid": "369",
    "vc": "v2.1.111"
  },
  "displays": [
    {
      "display_type": "query",
      "item": "6",
      "item_type": "sku_id",
      "order": 10
    }
  ],
  "page": {
    "during_time": 14807,
    "item": "2",
    "item_type": "sku_id",
    "last_page_id": "orders_unpaid",
    "page_id": "good_detail",
    "source_type": "query"
  },
  "ts": 1677828069291
}

👉 Only fields that are useful for our indices are considered. Other fields may be ignored.

Timestamp (ts) Field

This field records the timestamp when certain a certain take place

Common Field

This is used to analyse daily active users.

Only start event (Launching the app) contains this field

The ts field is used and the date is extracted to filter multiple launching event within a day

	Attribute	Comment
ar	area code	the postal code of mainland China, specified in enums/MyRegion
mid	device id	device identifier
uid	user id
ch	install channel	source of download
vc	version control	app version

Display Field

This field exists when items are displayed to the user through promotions, ads and recommendation

(The field display_type does not contains query)

	Attribute	Comment
display_type	type of this display	recommend: system recommendation promotion: promotion content activity: sale query: user search result
item_type	type of the item field	sku_id for a single item, sku_ids for multiple items separated by ,
item	item id(s)	single item id like 1, or multiple item ids like 2,3,4
order	id of the order

Page Field

This field has some similar fields as the display field

(The field source_type contains query)

	Attribute	Comment
during_time	duration	The time during which the user stays and browses the page
page_id	type of page	specified in enums/PageType. Examples are home page, search page, payment page and so on
last_page_id	previous page id	previous page of a browsing route

Indices

Browsing Route Analysis

demo video

We analyse the browsing route to discover user’s behaviour pattern and see the overall conversion rate of the platform

fields needed:

Field	Usage
uid	user id to identify route cache
page_id	current page
last_page_id	previous page
ts	timestamp

We first query the last route history of a user from the Radius cache and get the most recent page of that route.

Then we compare the most recent page from that route with the last_page_id in the current log to see if the route is continuous.

If the route is continuous, we append the page_id to the route and update the cache.

If the route is not continuous, we submit this route history and start a new route with page_id.

Click Through Rate & Recommendation Algorithm Precision

demo video

CRT and algorithm precision tells the successfulness the promotion and recommendation algorithm are.

⚠️ Note that the unit of clicked and shwon data is different, specifically in this case, 1:5.

fields needed:

Field	Meaning	Comment
common.ar	area code to analyse a specific region
display.display_type	the type of promotion/impressions	recommend: system recommendation promotion: promotion content activity: sale query: user search result
page.source_type	the type of promotion/impression from which the user enters into the product page	recommend: system recommendation promotion: promotion content activity: sale query: user search result
ts	timestamp

Daily Active User & Download Channel Distribution

demo video

DAU shows which regions has the highest online users (app launch). The download channel distribution shows the effectiveness of app promotion from different app stores.

fields needed:

Field	Usage
common.mid	device id
common.uid	user id
common.ar	area code
common.ch	download channel
common.vc	app version
ts	timestamp