Google can be frustrating sometimes, can’t it? Sometimes your perfect result is the top result, but sometimes it’s a few pages down and it takes you a while to find it.
Our customers told us that searching for a product in our gigantic database can sometimes be a lot like Google—sometimes awesome, but sometimes it takes a little digging. We took that feedback and put our best minds in data science and machine learning on the job. Those big brains came up with the answer: the Advanced Search Endpoint, now available to enterprise customers.
The Advanced Search Endpoint (ASE) is one of the most powerful capabilities of the Indix Product API. It takes in incomplete product information, runs it through the Indix data algorithms, and produces a set of matched products that will be more accurate than simple search results.
ASE is the optimal endpoint when a simple query on one of our synchronous endpoints yields results that you’d need a human to carefully look through to find out the best one. A product title is all that’s required (although more information is better), and our advanced search will get to work to find the best match or matches.
To use ASE, developers post a file with all the information they have about their products—up to 10,000—to the endpoint. The algorithms process the information and up to 20 results per product can then be downloaded with information as to whether the result is due to search or matching and the corresponding confidence score for the match.
Here is a screenshot with some truncated example results; this is what a match with a confidence score of 0.5 looks like:
The confidence score is what Indix customers use to determine how much human intervention they might need to choose the best product result—the higher the score, the better the match:
|1.0||UPC, MPN and Title matches 100%|
|0.9||UPC and MPN matches|
|0.8||UPC and Title matches|
|0.6||MPN and Title matches|
|0.4||Title matches exactly|
|0.3||Title matching score is between 0.95 to 1.0|
|0.2||Title matching score is between 0.90 to 0.95|
|0.1||Title matching score is between 0.80 to 0.90|
|0.0||Title matching score is lower than 0.80|
Also published on Medium.