With the release of pipelines last year, we made it possible for any team to optimize search results without the need for hiring a cadre of engineers.
Pipelines offer tremendous versatility in how you configure search. What may have taken weeks or months of engineering can be accomplished in minutes. For example, with pipelines you can:
- Improve visibility of best selling, higher rated, or highest margin items
- Promote low or high-inventory items
- Merchandise seasonal or on-sale items
- Personalize search results
- Or all of the above!
In this article, we’re going to show how textual and ranking adjustments, which are built on top of pipelines, give you control over the search algorithm to display the results you want.
But first, let’s briefly explain pipelines.
What are pipelines?
Until we released pipelines, every search engine required a significant investment of time and engineering resources to manage search result ranking. A team of people would be needed to add complex rules, modify the underlying algorithm (a bad idea!), and/or write libraries of search APIs to modify and improve results from data scattered across systems.
In contrast, we moved search configuration into what we call pipelines, an easily accessible and easily modifiable set of steps that can be run during indexing or at query time. Pipelines allow anyone to do advanced search configuration with either non-technical relevance settings or a few lines of YAML script.
Pipelines can be saved, versioned, and even A/B tested to determine which configuration performs best. They take minutes to create and new results can be previewed in real-time.
Search ranking algorithms
For most search engines, the algorithm that determines search results ordering can be a mystery. Why did Product X get listed before Product Y? The search algorithm is looking at a variety of factors — from keyword density to product popularity. But it may put too much emphasis on one factor over another, which can cause unexpected results.
And what if you want to control how search results are ordered to ensure the right brands or product types are listed?
You would need to write rules. Lots of rules. Running a new sale? More rules. Added new products or an entirely new category to your catalog? More rules. It’s not long before you can be overwhelmed with managing rules.
We took a different approach. Pipelines give you a way to determine relevance and manage ranking. Textual and ranking adjustments are just one of the many ways you can use pipelines to deliver better results. Let's look at them in more detail now.
Under Relevance Settings you will find textual adjustments, available as a no-code feature or pipelines to configure which attributes are most important for ranking results. It gives you greater control over determining how you want search results to rank.
For example, you may want to put extra weight on ranking for product title, page or product tags, category, URL, user ratings, or some combination of those attributes. With Sajari, you can easily test different settings and preview how your changes affect results.
Textual relevance settings put the power of pipelines into the hands of everyone at your organization. More technical users can switch to the Advanced tab to make the same adjustments using YAML.
To give you even greater control over results, there are also Ranking Adjustments. It includes the ability to change results based on attributes such as brand name, category, or product type, and range boosts to control results based on just about any criteria you choose.
For example, you can use range boosts to push popular, best selling, review count, or highest rated items — or any other attribute you think is important — to the top of the search results.
Note in the screenshot above that we’re adjusting the search results for all four of these example criteria and that each criteria is weighted, or scored, in importance. Pipelines will execute the query in this exact order (1) popular (2) best selling (3) customer reviews (4) highest rated, or you can change the order you want the pipelines to run.
What if you add a new field, re-scored the fields — increased or lowered the importance of an attribute — or removed an attribute from your boosts? The search preview along the right hand side shows you exactly how it will affect results.
Preview, save, and publish
When you’re happy with how the results are ordered, you can publish the new search algorithm live to your site with one click.
Everything we described above can work for any number of criteria.
- Most popular
- Highest margin
- Customer reviews
- Most rated
- By geo
- Date product was added
- By tag
You could even create a pipeline of everything mentioned above and compare it against a pipeline in a different sequence or with fewer/more steps.
You can easily create and save multiple variations for different times of the year or new sales. For example, you might have a “Spring Fashion” version, a “Black Friday” version, and a default version, any of which can be easily pushed into production whenever you’re ready.
Every pipeline version can be saved and easily rolled back.
Where rules require painstaking development, can conflict with one another, and require maintenance, Sajari relevance settings are easy to manage, offer real-time previews, and give you tremendous control over the search algorithm.
Ultimately the goal is to increase search revenue. With ranking and textual adjustments, there are no limits on what you can achieve. Signup for a free 14-day trial to see for yourself or contact us for a personalized demo.