Catalog structure
1. Ensure consistent and clear product titles
1.1. Include the core product name
Every product title should clearly state the main product (e.g., "sofa", "lamp", "cover") to make it easy for search engines to understand what is being sold. Our code internally identifies this main product noun from every title and prioritizes relevant products when the respective product noun is searched.
Examples:
- Men's Wool Sweater
- 32-Inch LED TV
- Cordless Drill with Battery
- Running Shoes for Men
- Organic Whole Milk
1.2. Stick to consistent phrasing
Use a logical order in titles, such as "Product for [Use Case]" (e.g., "Slipcover for Sofa" or “Sofa Slipcover”). Avoid irregular structures like "for Sofa, Slipcover" which can reduce clarity and search relevance.
Examples:
- Leather Jacket for Men instead of "For Men, Leather Jacket"
- iPhone 13 Screen Protector OR Screen Protector for iPhone 13 instead of "For iPhone 13, Screen Protector"
1.3. Keep titles focused
Avoid adding unnecessary information like ingredients, technical specs, or additional parts directly in the title. Use the title to focus on what the product is (e.g., "Chocolate Cake", "Leather Sofa Cover"), and move supplementary details to their respective attributes (e.g., color, material).
Examples:
- Bluetooth Headphones (avoid: "Bluetooth Headphones with 20-Hour Battery Life, Black")
- Chocolate Cake (avoid: "Chocolate Cake (egg, sugar, chocolate)")
2. Handling product variants
By default, only one variant (e.g., one color or size) per product is shown in search results. If you want multiple variants (e.g., all colors) to be displayed, ensure they are assigned as separate products or given different parent groupings in your catalog.
Examples:
- A t-shirt may be available in 3 colors (White, Maroon, and Blue) and 3 sizes (Small, Medium, and Large). Assuming all colors are available in all sizes, it makes it in total 9 variants.
- If the query term is “t-shirt”, only one of the 9 variants will be displayed. Here, the variant that is most popular on the website is displayed.
- If the query term is “white t-shirt”, one of the 3 white t-shirt variants will be displayed, depending on which size is the most popular.
- Here, let us say when a query is “t-shirt” and you want to show all colors with the respective trending size in each color to be shown to the customer, it is expected that you will create separate product groups for individual colors.
3. Standardize and use attributes effectively
3.1. Ensure consistency across attributes
Attribute fields (such as color, size, and material) should be uniformly filled across all products. Not only inconsistent attributes reduce the ability of customers to refine search results effectively, inconsistency across where they are added (e.g. attribute vs description vs title) also impacts their ranking in the search results.
and Similarly, please try to be consistent with attribute values as well. For example, if you are using S, M, L symbols for size, avoid using Small, Medium, and Large for others. It is also recommended that you use full forms instead of abbreviations.
Some normalisations, for example, S → Small, Men → Man, Male, Gents, etc are taken care of by the Klevu system. However, should you want your shoppers to be able to search with different forms, please use the manual synonyms options.
4. Assign products to the correct categories
4.1. Granular categories improve relevance
Assign products to the most specific categories possible. For example, place phone cases in "Phone Accessories" rather than just "Phones" for better search precision.
4.2. Avoid mixed categories
Broad categories like "Phones & Accessories" can confuse users and dilute search results. Use more specific categories like "Phones" and "Phone Accessories" to help customers find what they need.
4.3. Align category names with product types
Category names should reflect the primary product type (e.g., "Lamps" instead of "Home Decor") to improve search relevance.
5. Leverage descriptions and metadata
It is important to reiterate that please focus on including key information such as product type in the title, and product attributes in attributes. Descriptions should complement, not replace, this information.
When writing descriptions, please provide as much detail as possible about the product itself and its potential use cases and suitability criteria in the description. Write them using natural, clear language. Avoid overly technical or incomplete sentences, which can confuse search engines/vector databases and hurt relevance.
6. Use synonyms
Klevu tries to identify and add commonly accepted synonyms (e.g. “sofa” vs “couch”) as part of its enrichment process. However, for any domain-specific or technical terminology, we encourage you to use the manual synonyms feature.
Why submitting analytics events is crucial
1. Unlock actionable insights from your data
By submitting key analytics data such as search terms, clicks, ratings, and checkouts, you gain access to powerful insights that can transform how you run your store. We use this data to show you what’s really happening: from which products are trending to which searches are leading to zero results. This would help you make decisions faster and ensure your store stays ahead.
The analytics data also allows us to identify trends – something that we use to fine-tune the algorithms automatically.
2. Pretrain the algorithm with historical data
If you’ve got any historical data, we request you share it. Search terms, product clicks, product purchases, etc. Sharing this allows us to pretrain our algorithms, ensuring that your store benefits from more accurate recommendations and improved search results right from the start. This data is useful and used thoroughly across all the solutions – search, category merchandising, recommendations, and personalisation to help our digital assistants.
Personalization for your store
1. Start personalizing from day one
Another reason why you should share your historical data is that it helps in driving personalisation. When a shopper is new, we would use collaborative filtering to recommend products based on what similar customers have shown interest in. This ensures that even first-time visitors are presented with relevant products, making their shopping experience more engaging right from the start.
As customers click on more products, our algorithm learns in real time. The more they engage, the more personalized their experience becomes. Our system gets smarter with every interaction, helping you deliver a customized shopping journey.
2. Boost or Filter attributes
If a customer is searching for a "red shirt," we’ll automatically boost similar shades like maroon. But maybe gender is a key attribute for your store - tell us, and we’ll apply a filter to ensure shoppers see only gender-specific products when appropriate. Our systems will eventually learn but you know your products and your customers the best! Share your understanding, use cases, and different patterns you have observed in shoppers behavior and we will use it to fine-tune your system.
Here are some examples of what other merchants have shared in different domains:
- Fashion retail: prioritizing style over brand
A fashion retailer noticed that their customers care more about style and fit than brand loyalty. They asked us to boost products based on attributes like "slim fit" or "casual style" rather than brand names. Now, when a shopper looks at casual wear, the system emphasizes style-related features instead of focusing solely on the brand. - Electronics: filtering by device compatibility
In the electronics domain, a merchant specializing in phone accessories requested that we apply a strict filter for device compatibility. When a customer searches for an "iPhone charger," they want to ensure the results show only products compatible with that specific device, filtering out all irrelevant products for other phone models like Android. - Luxury goods: focusing on price range
A luxury goods retailer knew their customers frequently shopped for high-end, premium products. They asked us to focus on price as a key attribute. If a shopper often purchases high-ticket items, we boost products that match their preferred price range, such as jewelry or designer accessories above a certain value, ensuring the shopper is shown the most relevant products. - Home decor: material preference over design
A home decor store noticed that customers cared more about material (e.g., wood vs. plastic) than design aesthetics. They asked us to boost products based on material first, ensuring that when a shopper searches for "dining table," they see options that match their material preference, like "oak" or "solid wood," instead of focusing solely on the design or brand. - Beauty products: applying gender and skin type filters
A beauty products retailer asked us to filter products based on both gender and skin type. When a customer looks for skincare products, the system now filters out irrelevant gender-specific products and focuses on items suited to specific skin types, like sensitive or oily skin, boosting the most appropriate matches for each shopper.
By sharing these kinds of insights, you ensure that our system is perfectly tailored to your store, providing personalized shopping experiences that drive engagement and increase conversions.
3. Submit shopper - Specific preferences
If you already have insights into your customers' preferences, for example, you are collecting them or using a 3rd party software to help you out with that, our API is flexible and allows you to submit shopper-specific preferences directly when a query is fired. You can submit key-value pairs with weights for different attributes. For instance, if you know a shopper prefers a specific brand, you can supply that information, or if a customer often buys luxury gifts, you can submit their preferred price range. This ensures we can fine-tune recommendations to cater to each shopper.
The more you share with us, whether it’s specific patterns, filters, or features to prioritize, the better we can customize the experience for your customers. Together, we can create personalized journeys that keep your customers engaged and boost sales.