AI Search Engine

The feature dedicated to ecommerce business enhances the results of the search engine on your website with product recommendations that match the customers’ preferences. Customers who use the search engine on the website will get a product offer prepared individually to their liking. The mechanism behind the feature is powered by the words included in the attribute names in the product feed. The results are reinforced by the use of customer’ behavioral profiles and the popularity of the products.

Benefits


  • Higher conversion
  • Increased revenue
  • Strengthened relationship with your customers

Requirements


  • Delivered product feed
  • Integration of events via AI tracking or API

The full list of requirements and the AI implementation schedule can be found at these links.

Note: AI Search Engine will work after delivering the product feed, however, supplementing the system with transactional data will facilitate the personalization of results.

Implementation


  1. Paste the tracking code to your website.
  2. Add the product feed to the application.

    Note: You must have the links to our API, which you can find here.

  3. Go to Assets → Search Engine → Integration and define selectors in the Search Input Selector field, so that AI Search can be triggered.

  4. Choose the attributes of the products you want to display as a search result.

  5. Include the method called window.snrAIOnSearchResponse in your page’s JavaScript. This function will be triggered whenever a search query has been entered in one of the input selectors and will return search results.

  6. Call a function named snrAI.search.init() in your code. This function starts watching all the input selectors you defined earlier, fetches search results and triggers window.snrAIOnSearchResponse.

  7. You must render the search results in window.snrAIOnSearchResponse on your own so that the Synerise suggestions/recommendations appear on your page. When conducting the query you can specify multiple parameters (e.g. query, offset, limit, client UUID, min. price, max. price, etc.).

  8. The response JSON gives you your token, returns the number of products found, as well as displays the products, after searching through your specified filters. The response also shows you if a suggestion was used. For example, if a typo occurred in the search query, it returns information about the word or phrase that it searched instead, as well as the score for that word or phrase. Apart from that, the response JSON gives you facets. These are the attributes for which the results are aggregated. For example, if a customer searched for a product, you can implement the results to be returned with information about the number of products in given colors, or number of products of some brands. To personalize the search results, you need to retrieve the customer’s UUID. In order to do that search for the snarAI-uuid in the browser’s Local Storage.

Configuration


In order to start configuration of the search engine powered by AI, you need to follow this path:

Assets → Search engine

The configuration will involve 4 steps.

Tip: You can use query rules to interfere with the search results so you can meet specific business needs (for example, to promote specific products in search results).

Ranking


First, adjust the ranking settings. You need to indicate how you want to customize search results, whether you want to put stress on the products which were viewed or purchased the most frequently.

Image presents two bars on the white background, the first bar presents four headers with marked Ranking header, and the second bar shows the settings of Ranking. There are three sliders, each can be set from 0 to 100.

  1. Past page visits - Use the slider to indicate how much you want to prioritize products which were viewed the most frequently. EXAMPLE: If you set the slider to a high number (e.g. 70), the search engine will prioritize products that were viewed the most frequently by the visitors on your website.

  2. Past transactions - Use the slider to indicate how much you want to prioritize products which were purchased the most frequently. EXAMPLE: Provided you set this slider to the high number (e.g. 87), the search engine will prioritize products which were purchased the most frequently by your customers.

  3. Personalization - Use the slider to indicate how much you want to prioritize products based on the behavioral profile of the customer. EXAMPLE: If you set the slider to the high number (e.g. 70), the search engine will prioritize the display of product recommendations to an individual visitor based on the behavioral profile.

Image presents three lists, each list has several positions with checkboxes.

Then you need to determine the importance of searchable terms (attributes in product feed).

If you want to add new term, click the button + Add attribute.

Image presents the dropdown list with 8 positions.

Standard adding

From the dropdown list, choose the attributes you would like to add.

Each column (high importance, medium importance, and low importance) is a list of attributes with assigned priority search. Remember that attributes enlisted in the same column are of the same importance.

EXAMPLE: You can prioritize a brand or product name in search results:

  • Product 1: Title: Laptop 123 Brand: Lenovo;

  • Product 2: Title: Lenovo Laptop Charger Brand: Unknown;

The Product 1 will be displayed in the search results, because the brand is more important than the title of the product.

Prefix adding

Type in the prefix of the attribute. As a consequence, all attributes that start with this word will be added to the selected Importance column.

Synonyms


The next step is to determine the substitutes for words and phrases. Sometimes customers use slightly different words or abbreviations while looking for products, which may prompt no results or inaccurate results. To avoid such a situation, you can prepare a list of synonyms for the given words.

Image presents two bars set on the white background. the first bar presents four headers with marked Synonyms header. The second bar includes a table, with four columns: checkbox, phrase, synonym and date of creation.

For simple typos, recognizing singular words from plural, etc. the search automatically suggests what it considers to be correct. When the search suggests another word or phrase, it informs the user about it by displaying a message.

  1. Click the Add parameter button to add a substitute.
  2. Type a word or a phrase you want to substitute.
  3. Type a word or a synonym that will replace the original phrase.
  4. Click the Add parameter button if you want to add more synonyms.

Preview


In the next step you can try out your engine.

Image presents three bars set on the white background. The first bar presents four headers with marked Preview header. The second bar contains search two search engines. The third one presents two separate squares, one square shows a list of filters, the other one three images with products and information about them.

  1. Type in the search engine any phrase to see the results you get from particular words.
  2. If you want to see the product recommendation for a specific customer, paste the UUID of the customer.
  3. You can see the visual representation of the product recommendations returned by the search engine.
  4. Check the attributes which are visible while the product is displayed.

Image presents the xml code with a list of product attributes.

Settings


The last step in the Settings section involves selection of attributes that will appear in the search results and language of search engine.

Integration

Image presents two bars set on a white background. The first bar presents four headers with marked Integration header. The second bar presents a dropdown list consisting of attributes

  1. To define attributes by which search results are filtered, in the Filterable attributes click the Add attribute.

    Note: Filters exclude results based on initial criteria and do not change with each search.
  2. On the pop-up, in the Custom attributes field you can select attributes that contain . and * in names.

  3. To add attributes which value is presented as a range, select them in the Custom ranged attributes dropdown list.

  4. Confirm your choice by clicking OK.

  5. To define the facets that the users of the search engine can select while searching, in the Facetable attributes, click the Add attribute button.

    Note: Facets exclude using the properties from returned results and change with each search.
  6. Choose the way you want to select the attributes:

    • To select single attributes from the dropdown list, on the pop-up click Standard.
    • To select all attributes that start with a particular word, click Prefix.
  7. Depending on the chosen method of attribute selection, select attributes from the dropdown list (for Standard) or type the name of the attribute (for Prefix).

  8. Confirm your choice by clicking OK.

  9. To choose the attributes of products that will be displayed in search results, in Included attributes click the Add attribute button.

  10. Choose the way you want to select the attributes:

    • To select single attributes from the dropdown list, on the pop-up click Standard.
    • To select all attributes that start with a particular word, click Prefix.
  11. Depending on the chosen method of attribute selection, select attributes from the dropdown list (for Standard) or type the name of the attribute (for Prefix).

  12. Confirm your choice by clicking OK.

Language

Image presents a dropdown list with languages

Choose the language of the search engine. Make sure that the selected language is compatible with the language used in the product feed. When those two languages match, the results of the search queries will be accurate.

Statistics


The statistics module of the AI search engine allows you to break down the search results of the users into data categories such as filter usage, popular searches, etc. This way you can assess whether the search experience available on your website is data-driven.

The statistics module is built into the feature, you can find it in Assets -> AI Search Engine -> Statistics

Global statistics

description

  1. KPI - Key performance indicators will reveal more information about the conversion rate, click through rate of the products found through search engine, etc.

    • CTR - (Click-through rate) it is the ratio of search results that ended up being clicked.
    • Conversions - It is the effectiveness indicator that shows how many times a unique product was bought after clicking it in the search results during last 5 days.
    • Conversion rate - This ratio is counted through dividing the number of conversions by the sum of conversions and all clicks after which the product was not bought.
    • Revenue - Revenue generated from search result clicks during last 5 days.
    • Avg. Click Position - The average position on the list of the clicked product.
    • Avg. Order Value - The average value of an ordered product after clicking the search result.
  2. Total - This is the place where you can check the general statistics for search engine.

    • Searches - The total number of searches.
    • Clicks - The total number of clicks in the search results.
    • Suggestions used - The number of times when a suggestion was made for the search results.
    • No results - The number of searches without any results.
  3. You can use a calendar to narrow the statistics to the period that interests you the most.

description

  1. Most popular searches - This is the list of the most frequently searched queries.
  2. You can use a calendar to narrow the statistics to the period that interests you the most.
  3. Query - This column contains the list of all popular queries searched in the chosen period.
  4. Count - This column shows the number of times a particular query was searched.

You can check the details of each query which is available in the form of a chart. Click the arrow next to the name of the query.

  1. Totals - You can check the clicks, searches and suggestions used while searching for a chosen query.
  2. Click Through and Conversion rates - You can check the click-through and conversion rates for a query.
  3. Revenue - It is the total amount of income generated from the most popular searched queries.

description

  1. Popular searches - This is the list of the most frequently searched queries that don’t have any results.
  2. You can use a calendar to narrow the statistics to the period that interests you the most.
  3. Query - This column contains the list of all popular queries searched in the chosen period.
  4. Count - This column shows the number of times a particular phrase was searched.

You can check the details of each query available in the form of a chart. Click the arrow next to the name of the query.

  1. Daily stats - This chart presents the number of times a certain term was searched.

You can narrow the analyzed period by hovering the mouse cursor over the edges (as presented on the picture above) and slide the edges to the desired dates.

Filters

description

  1. Popular filters - This section presents the list of the most frequently used filters available in the search engine.
  2. You can use a calendar to narrow the statistics to the period that interests you the most.
  3. Filter - This column contains the list of filter names used by website visitors.
  4. Count - This column shows the number of times a particular filter was used.

Click the arrow next to the filter name to see the details.

😕

We are sorry to hear that

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.

😉

Awesome!

Thank you for helping improve out documentation. If you need help or have any questions, please consider contacting support.