Dynamic Reranking| Hitchhikers Platform

What You’ll Learn

In this section, you will learn:

  • What is Dynamic Reranking
  • Use Cases
  • How to configure Dynamic Reranking

Overview

Dynamic reranking uses machine learning to rerank results and answers based on user engagement data. Our team has trained a complex model that reranks entities in order of their likelihood to be clicked by the user. This model will become smarter over time as it is fed more user interaction data.

If our model detects that the top-clicked on entity is the one returned in the fifth position, this new feature will allow the algorithms to dynamically rerank the results and bump that result to the top since the goal is to present the most relevant result first, as indicated by user engagement.

This will be a powerful feature for increasing clicks and conversions, making your search experience all-around more effective. When users see the result they intended to find in that first spot, they trust the accuracy of the search algorithm, and are more likely to use it in the future!

Example Use Cases

Let’s say you’re a restaurant, like Turtlehead Tacos. If I search for the word “tacos” how do we know which of the many tacos you sell should show first? Taking a look at the query “tacos” – all the menu items are equally semantically relevant from a natural language perspective.

This is where dynamic reranking excels. We can use user engagement data to figure out which tacos to show first. What we want to do is show the taco that’s most engaged with first and the one that’s generating the most revenue.

tacos-reranking

Let’s take another example using FAQs – in this example, you’re a healthcare company surfacing answers to questions on strokes among other diseases, and conditions. We can see how dynamic reranking has improved the results below with the following example: a patient searches for “stroke risk factors”. Before, it returned the FAQ about older people’s stroke risk factors first. However now, the second FAQ, assessing risk for strokes, which drives more clicks, returns first:

stroke-reranking

  • Note: Our reranking model is now able to rerank more broadly across all entity types–not just FAQs!

Search Configuration

You can activate this algorithm via a new top-level object within each vertical called dynamicRerank.

Here’s an example of how a search config using this feature might look for FAQs:

  "verticals": {
    "faqs": {
      "entityType": "faq",
      "dynamicRerank": {
        "target": "CLICKS"
      },
      "searchableFields": {
        "name": {
          "semanticTextSearch": true,

        },
        "keywords": {
          "phraseMatch": true
        }
      }
    }
  }

In the target attribute, the user can select which user action is being optimized for. For now, there is only one option - clicks.

Our tests show that Dynamic Reranking can boost clicks by as much as 10%! When a user has a successful search experience, they will return the next time they have a question, and to our customers’ delight — will stay in the conversion funnel at all times.

Test the Difference

If you want to see the differences before and after implementing Dynamic reranking, run a Version Comparison using Search Term labels. For more information on how to do so, visit the Version Comparison unit in the Post-Optimization module.

unit Quiz
+20 points
Daily Quiz Streak Daily Quiz Streak: 0
Quiz Accuracy Streak Quiz Accuracy Streak: 0
    Error Success Question 1 of 3

    Dynamic reranking uses which of the following to rerank results:

    Error Success Question 2 of 3

    dynamicRerank is configured...

    Error Success Question 3 of 3

    True/False: The target attribute allows the Admin to choose which user action the dynamic reranking is optimized for.

    Climbing that leaderboard! 📈

    You've already completed this quiz, so you can't earn more points.You completed this quiz in 1 attempt and earned 0 points! Feel free to review your answers and move on when you're ready.
1st attempt
0 incorrect