Overview

Providing transaction history is optional but recommended since it allows training recommendation models on historical user data. Without providing this data, it may take considerable time before models train and adjust properly on the user data collected from the web, thus providing sub-optimal performance until adjustment (i.e., enough data is collected with the data collector deployed on the web) is complete.

Please provide a one-off export of transactional data in the following CSV format.

Columns

Name	Description
`date`	Date of transaction
`user_id`	Unique user ID (e.g., cookie, email, etc.). Ideally, the same identifier that the Perselio data collector uses (typically information is grabbed from the website from Google Analytics). Alternatively, an arbitrary anonymous ID can be used.
`item_ids`	List of item_ids purchased in the transaction. Item_id must match with item_ids provided in item feed integration.

Example

date,user_id,item_ids
"2023-11-04","fd834fa733682951518d85f52daee","498,508,1048,114463,133587,109515,109516"
"2023-11-04","5ccbe77251f08568739270a4a409b","125881,119012,100294,96137,132817,85226"
"2023-11-04","819fe5ea927b479f6b938e20ecc78","102629,42948,88499"
"2023-11-05","4c876c865200c38ec58de993710c0","119476,135637"

FAQ

Can I provide a shorter period of transaction history?

Yes, you can, but having at least one year of history covers at least one whole year season.

What exactly is the user_id that Perselio uses?

Please ask the Perselio team, it depends on the way how the onboarding was done (depending on current website architecture).

Can I provide the custom format of transactional history?

Yes, but it must first be agreed whether the Perselio team will create a custom adapter.