Book a demo

Data Ingestion and Usage with PersonaFin AI

How is Data Captured?

Personafin AI offers two main methods of capturing user interaction data:

Personafin behaviour capture library

The first capture method is through the Personafin event-tracker library (also known as Web Behaviour Capture library), a light-weight javascript library that can be installed into your user interface to capture interactions with the platform.

This is suited for customers who have an existing UI that they wish to enhance directly with our API’s.

You can read more about how this works, and see installation examples in the Behaviour Capture Getting Started documentation.

iFrame Solution

If you are using our iFrame solution, we use the same library to capture your interactions. However, this is all handled for you within the easy to install components.
This is suited for customers that have an existing UI or platform, or are developing one, that they wish to develop and enhance with minimal effort.

If these methods are not suitable for your platform or insufficient to capture all the required data, we can offer a number of additional integrations. To find out more about this contact us at help@personafin.ai.

What data is captured?

The behaviour capture library is designed to capture critical information for establishing a users preferences, developing a profile of their behaviour, and generating content and research recommendations.

We strive to minimize the event data captured as much as possible, particularly when it comes to sensitive personal data. We limit identifiers captured to a single user identifier that should be supplied by the customer platform; this identifier is then psuedo-anonymised within our environment and can be further de-identified within your platform using our On-Premises Identify Anonymisation tool. Read more here.

KEY DATA ELEMENTS

An example of data captured (in JSON format) is given on the right. The key elements of this are:

DATA FIELD WHY IS IT IMPORTANT?
client_key

Lets our data ingestion identify where the data came from and route it appropriately.
event_type action

Event type and action performed ensures that the interaction contributes appropriately given its degree of importance
event_meta

Event meta contains all important information about the interaction performed. This can include:

  1. content id: lets us identify a piece of content interacted with. Critical for tracking trending content.
  2. search term: Lets us link a search event to a particular term.
  3. impression id: Lets us link an interaction back to a recommendation provided by our API’s.
related_entities

A list of Tradeable entities associated with the interaction. This is critical for tracking a users entity preferences, and for tracking trending entities.
channel_meta.session_id

Allows us to track interactions across a given session.
channel_meta.client_ref

A user identifier provided by the customers platform. This is critical to allow us to track a specific users interests and preferences, allowing us to provide a personalized experience.

{
“client_key”: <capture key>,
“data_source”: <client platform>,
“send_date_time”: <datetime>,
“event_type”: <interaction type>,
“action”: <interaction action>,
“event_meta”: [
{
// Search type event meta
“key”: “term”,
“type”: “search”,
“value”: <term searched for>
},
{
// Content type event meta
“key”: “content_id”,
“value”: <some content identifier>
},
{
// Meta when interacting with an api output
“key”: “impression_id”,
“value”: <impression id from API output>
}
],
“related_entities”: [
{
“client_ref”: <entity ticker>
}
],
“channel_meta”: {
“session_id”: <session id>,
“client_ref”: <user identifier>
//…. other meta data
}
}

What happens once interaction events enter our system?

During behaviour capture an interaction event undergoes de-identification and anonymisation, as outlined in here.

Once the interaction event enters the Personafin environment and has undergone this anonymisation process, it is then processed and routed appropriately for processing various data sets. This includes all consensus data (trending content, search, etc), and datasets personalized for the user that the event was performed by.

These datasets are stored in presentation ready storage ready for API query. At this point all recommended and content items assigned an impression id(known as `impression_id` or `fxcid` for content), which can be surfaced back onto the customer UI ready to be fed back into the behaviour capture library. This allows us to measure recommendation engagement and update the recommenders appropriately.

The example flow below gives a high level overview of the flow of data around the Personafin system.