Signals Data Requirements
Your signals data is historical data describing the behavior of your users.
Required signal/event types
At a minimum, Never Null requires these signal types:
Query response signals |
A response from the search platform that includes search results, if any. |
Add-to-cart signals |
The user adds an item to the cart. |
Recommended signal/event types
The more signal types you can provide, the better your Never Null results. Some recommended signal types are described below. Plan on providing any that your search application supports.
Click signals |
The user clicks a search result. |
Add-to-favorites signals |
The user adds an item to a list of favorites. |
Hover/quick-view signals |
The user hovers over a product, for example to display a short description. |
Purchase signals |
The user purchases a product. |
Remove-from-cart signals |
The user removed an item from the cart. |
Query request signals |
Events that generate a query, such as entering a query string or clicking a facet/filter. If you have response signals with the required fields, then request signals are not needed. |
Required fields
| Field | Description |
|---|---|
Timestamp |
The date and time at which the event occurred, preferably in Solr UTC format, as in |
Signal ID |
A unique identifier for this individual signal record. |
Event type |
The type of interaction recorded by this signal, such as click, add-to-cart, like, purchase, and so on. |
Product ID |
The product ID from your catalog, corresponding to the item with which the user interacted. |
Query string |
The user query terms associated with this event, if any. |
Zero-results flag |
In query response signals only, a flag indicating whether the user’s query returned zero results, if available. This could be a value representing the number of search results, where the value "0" is the zero-results flag. |
Search result ID(s) |
In query response signals only, one or more product IDs that were returned in search results. |
Session ID |
A unique identifier shared by all signals from a specific user’s session, regardless of event type. |
Additional guidelines
-
Signals must be raw, not aggregated.
-
The required number or time range of signals depends on the volume of activity on your site.
One to three months of signals is generally sufficient. Consult with Lucidworks for a recommendation tailored to your site’s activity level.
-
Omit details that could identify a specific user.
Supported formats
-
Tab-separated values (TSV)
See the example TSV data below.
-
Parquet
See the Apache Parquet format specifications.
| timestamp_tdt | signal_id | type | product_id | query_string | num_results | result_docs | session_id |
|---|---|---|---|---|---|---|---|
2020-01-31 15:50:18 |
0c3b742c-77f2-4dc7-8f34-4809be0db103 |
click |
518 |
wool |
q6kowbxwp31hqhiqc5zbcke375kaltq9 |
||
0c401791-d197-4570-a085-b5b4e90b2ced |
session |
0dkjtr6hz75z1eivvzt5eai7xts26w0y |
|||||
2020-01-31 16:09:40 |
13WjqqrKcH |
response |
shorts mens white |
0 |
da645378-cb14-422a-aecd-2b60f6cf83b6 |
||
2020-01-31 16:13:02 |
0c6ac8ea-8bd4-448d-a06a-ac73c7d0e2b7 |
request |
60a3d62a-f5f9-493e-8d10-976f756ddcf0 |
||||
2020-01-31 16:18:29 |
3akIIcgJuT |
response |
red shirt |
4 |
1,932,197,217,281,340 |
20a7fa35-8f4b-4519-b04e-1590f5dd7f7e |
|
2020-01-31 16:18:33 |
0c76c62c-1820-498c-9b85-80d71365d461 |
cart-add |
1992 |
wgvdx117osd9ic63wr0an0ssd3hlzayd |
|||
2020-01-31 16:20:38 |
0c826f75-4836-4812-a163-18ce93639e29 |
purchase |
1165 |
gbrf1z44bewylhwr1271f3nylm25hqpt |
Next steps
-
Once you have planned how to format your product data and your signals data, you can send your initial data stream.
-
After Never Null has performed its initial analysis of your data, you can fetch query results.