📘

Instructions match to QuickStart data

As with other sections, the data and examples provided are designed to be used with the data file provided in the QuickStart section. If you are using your own dataset, then adjust the API calls accordingly.

What is the AI Multiplier?

The AI Multiplier is a powerful feature that boosts search results based on behavior on your website. This is like any other boost rule in HawkSearch - it's just that in this case, the boost is added based on site activity. There are four AI Multipliers:

The Learning Search Multiplier looks at what results are being clicked on your search results. If people are searching for and clicking on a product frequently, that product will get a boost and will appear higher in the next search results. The Boost for Recommended Items refers to individual boost based on a specific person's activity - namely a purchase. If HawkSearch tracks a purchase by an individual of a product on your site, the next time they visit your site, that product will be boosted. The Orders and Add2Carts multipliers are similar boost results based on those events.

In order to see the impact of these actions, you can go to the Preview screen and view the boost details for each product. This will break down the impact of each AI Multiplier rule:

Before you get started

There are a number of moving parts to the AI Multiplier capability and this feature only works in Production and Test environments. All other documentation in the QuickStart section of our Developer Portal are on our Dev Environment, but this must run in test. If you are interested in testing this, please reach out to your HawkSearch project manager to copy your engine to our Test environment.

These are the components of the AI Multiplier

• Settings in HawkSearch - see the screenshot above - make sure you select some options
• The API tracking endpoints - these will typically be built into your website, but in this documentation we'll mimic the activity using Postman. These are written to MongoDB
• The Summary Jobs - these run each hour in MongoDB (keep this in mind during testing) to aggregate activity
• The Harness - used in personalization strategy - it performs further machine learning processes focused on individual behavior. This is performed each night.
• Rebuild All API - if you are not creating a new index and reindexing each night for example, then you should run the Rebuild All API to get the machine learning results from Mongo and the Harness to update your index with the appropriate boost values

Testing this functionality can be challenge as you try to mimic site behavior and considering the automated processing each our and each night - please reach out to your HawkSearch contact to help you with this process.

📘

Just the basics are highlighted here

There are additional variables and options you can include in your tracking. This documentation is to provide the basic tracking to give you an idea of how the feature works.

Learning Search Multiplier

The Learning Search Multiplier is determined based on search and click activity. It is critical though to start with a trackingId to connect your API update to tracking information already captured in HawkSearch.

Step 1: Get the TrackingId

You can run this in Postman just like the other QuickStart processes.

POST 

Header 

X-HawkSearch-ApiKey 

https://searchapi-test.hawksearch.net/api/v2/search

{ 
  
	"ClientGuid": "xxxxx",
    
	"IndexName": "xxx.xxx.xxx",
    
	"Keyword": "xxxxxxxxxx"
  
} 

Look at the search results in Postman and search for the TrackingId - it is a GUID - you need this to now track this request (the search above by default does not track search activity).

Step 2: Base64 encode the Search Call

The JSON below should be encoded so you can send it in the actual tracking API. You can use any tool for Base64 encoding, for example https://www.base64encode.org/

{  

	"TypeId": 1,  

	"TrackingId": "<<returned from search result>>",  

	"QueryId":  "<<generate yourself>>", 

	"Qs":  "Clem Hat" 

} 

You should get a result that looks something like:

eyAgCgoiVHlwZUlkIjogMSwgIAoKIlRyYWNraW5nSWQiOiAiN2Q0MWMyZmQtZDNlZC00MmY3LWIyN2YtOWQ3NTk1MTJhZjI4IiwgIAoKIlF1ZXJ5SWQiOiAgIjI1ZTc4MTdlLTlkZjMtNDNjYS04MGQwLTI4NGFhNzM2OWJiNSIsIAoKIlFzIjogICJDbGVtIEhhdCIgCgp9

Step 3: Send the Event Data through the Tracking API

POST 

Header 

X-HawkSearch-ApiKey

https://tracking-test.hawksearch.net/api/trackevent 

{  

	"ClientGuid":"<<from your test account>>",   
	
	"VisitId":"<<generate yourself>>",  
	
	"VisitorId":"<<generate yourself>>",  
	
	"EventType":"Search",  
	
	"EventData": "<<from the base64 encoding above>>" 

} 

Click Tracking

The click tracking data will help indicate how much to increase the boost for the Learning Search functionality.

Step 1: Base64 Encode the Click Tracking Data

{  

	"ElementNo": 15,  

	"TrackingId": "<<from the original search result>>",  

	"UniqueId":  "Item_77476",  

	"Url":  "https://preview-dev.hawksearch.net/elasticdemo/details?itemid=Item_77476"  

} 

After running that JSON through Base64 encoding you should a similar result to above.

Step 2: Send the Event Data through the Tracking API

POST 

Header 

X-HawkSearch-ApiKey (from your TEST account) 

https://tracking-test.hawksearch.net/api/trackevent

{  

	"ClientGuid":"<<from account>>",   

	"VisitId":"<<should be same as search>>",   

	"VisitorId":" <<should be same as search>> ",  

	"EventType":"Click",  

	"EventData": "<<from base64 encoding>>" 

}

Add2Cart Tracking

When a site visitor adds an item to their shopping cart, this call can be made.

Step 1: Base64 encode the Add2Cart Data

{   

 	"UniqueId":"Item_77476", 

 	"Price":24.99, 

 	"Quantity":1,

 	"Currency":"USD" 

} 

The results of the Base64 encoding will be similar to above

Step2: Send the Event Data through the Tracking API

POST 

Header 

X-HawkSearch-ApiKey (from your TEST account) 

https://tracking-test.hawksearch.net/api/trackevent

{  

	"ClientGuid":"<<from account>>",   

	"VisitId":"<<should match above call>>",   

	"VisitorId":" <<should match above call>>", 

	"EventType":"Add2Cart",   

	"EventData": "<<from base64 encoding above>>" 

}

PageLoad Tracking

This is a generic event capture which relates to the personalization strategy. It’s different from the click event because there could be scenarios when HawkSearch doesn’t know a click occurs. This call is added here because the personalization requires both the PageLoad and Sale api calls to work.

Step1: Base64 Encode Tracking Data

{ 

	"PageTypeId":"1",  

	"RequestPath":"/elasticdemo/details?itemid=Item_77476", 

	"UniqueId":"Item_76463"

}

Step2: Send Event Data through the Tracking API

POST 

Header 

X-HawkSearch-ApiKey (from your TEST account) 

https://tracking-test.hawksearch.net/api/trackevent

{ 

	"ClientGuid":"<<from account>>",   

	"VisitId":"<<should match above call>>",   

	"VisitorId":" <<should match above call>>", 

	"EventType":"PageLoad",   

	"EventData": " <<from base64 encoding above>>"  

}

Sale Event Tracking

Step 1: Base64 Encode the Event Data

This is tracked when the item is purchased

{  

	"Total":24.99,  

	"OrderNo":"11223344", 

	"Currency":"USD",  

	"ItemList":\[  

	{  

		"uniqueId":"Item_77476",  

		"ItemPrice":24.99,  

		"quantity":1 

		} ]  

}

Step 2: Send the Event Data through the Tracking API

POST 

Header 

X-HawkSearch-ApiKey (from your TEST account) 

https://tracking-test.hawksearch.net/api/trackevent

{  

	"ClientGuid":"<<from account>>",   

	"VisitId":"<<should match above call>>",   

	"VisitorId":" <<should match above call>>", 

	"EventType":" Sale ",   

	"EventData": " <<from base64 encoding above>>" 

}

Ensure Data was captured

All API calls are captured in the Admin section of HawkSearch: https://test.hawksearch.net/admin/trackingPreview.aspx

You can inspect the details of each call by clicking the eye icon on the left of each call.

Wait for Processing

Two processes need to happen for the results to be processed:

1. Raw tracking info for Learning Search, Add2Cart, and Sales are aggregated (this happens every hour)
2. Personalized data tracking data is sent to another system called the harness (this happens every night)

Make sure you speak with your HawkSearch project manager to make sure these processes are configured correctly.

Connect Machine Learning to your Index

Then run this API calls to call the machine learning that has been aggregated and processed into your index. If you are doing a full reindex with new data and a new index, then you can run the standard indexing API call (this is included in that call). If you have not changed your index, then run this call:

POST 

Header 

X-HawkSearch-ApiKey (from your TEST account) 

https://indexing-test.hawksearch.net/api/v2/indexing/rebuild-all

{	

	"IndexName": "xxx.xxx.xxx" 
  
}

Testing the results

The preview screen in the HawkSearch tool will show you the boost results through Machine Learning along with any other boost rules you may have created.

Download Test API Tracking Data

We have created a CSV which you can use in a Postman script to speed up your testing process. Note however, that the Search and the Click events can not be pre-created since they require a TrackingId that is generated by your index. However, the rest of the information will save you considerable time.

You can download the file here: https://hawksearch-sales.demos.hawksearch.com/sales/quickstart-sandbox/AI_Multiplier_Tracking_Data.csv

Note: you will need to generate some of the EventData as explained above.

You can run the Postman script like the rest of the examples in the QuickStart section:

POST 

Header 

X-HawkSearch-ApiKey

https://tracking-test.hawksearch.net/api/trackevent
{  

	"ClientGuid":"{{ClientGuid}}",   

	"VisitId":"{{VisitId}}",   

	"VisitorId":"{{VisitorId}}", 

  "EventType":"{{EventType}}",  

	"EventData": "{{EventData}}" 

}