Optimizing Your Site Search
Optimizing Your Site Search: Know-How for Online Shops and E-Commerce Businesses
- What does “site search” mean and what does it entail?
- How exactly does a site search work?
- What are realistic goals for a site search?
- How can the search experience be designed from the customer's perspective?
- What criteria does a professional site search fulfill?
- How is a search-friendly data feed structured?
- How can high product data quality be achieved?
- Which KPIs can be used to measure the quality of a site search?
- How can ROI of your online shop be quickly achieved?
- What control options should a site search offer retailers?
- What analysis and testing options should a site search offer retailers?
- What types of integration are available for your online shop?
- What special requirements do B2B shops have for a site search?
- Open source or standard solution: which is better for your site search?
- How can AI improve the customer search experience in e-commerce?
1. What does “site search” mean and what does it entail?
Site search is one of the most central functionalities of online shops and is often referred to as the “internal search” of a website. It consists of a product search within the shop and the products or items found. A well-implemented search function fulfills the search needs of users, minimizes bounce rate and, above all, increases the conversion rate. An analysis of the site search provides valuable information on visitor behavior.
2. How exactly does a site search work?
During a product search, the customer’s search query (also know just as a query) is compared against the index of the site search. The search function then provides or returns a list of matching products to the online shop. The ranking, or the order in which the articles are displayed, is also a function performed by the site search.
The search index is created from the available product data and its purpose is to accelerate the speed of the search. Because the index contains references to relevant items, it eliminates the need for the site search to search through all the underlying data.
A query is what the customer enters into the search field and what is processed and answered by the search server. Individual clicks, for example on filters or categories, also generate queries to the search server.
Whether individual terms, entire word chains or even questions, customers enter anything into the search field – often making spelling mistakes and typos. The site search needs to be able to process whatever customers may type. To achieve this, there are different technologies and concepts that are designed for e-commerce.
The Levenshtein distance method, for example, determines the similarity of terms based on the number of character changes needed to get from one term to another. This method is well suited for websites such as Wikipedia, where the search is specifically designed to match search queries with heading and body text and can compensate for inaccuracies such as spelling errors and typos within the words.
This method, however, is inadequate for online shops as it is not able to correctly evaluate complex queries such as “swivel chair for the office” instead of “office swivel chair. For long-tail queries like “red shoes for men from nike 42” the Levenshtein distance method does not provide the most relevant results – maybe even zero results. Therefore, retailers lose valuable sales, since very specific queries are usually made by visitors who are intent on buying and niche products with a high margin are often sought after.
Worldmatch ® technology, on the other hand, can correctly identify word components in all possible combinations and match them in a error-tolerant and language-independent manner. The similarity of terms and phrases is compared phonetically – according to the way they sound – just as a human being does.
3. What are realistic goals for a site search?
Online shop visitors who are interested in purchasing prefer websites where they can find what they are looking for quickly and easily. In order to improve the search experience for customers and generate a measurable increase in the conversion rate, it is important to make decisive improvements to your site search. With a successful search function, you can achieve the following goals:
Suggest function for quick orientation
By displaying fitting suggestions, online customers are guided directly to items they can purchase. Links to other content such as blog articles, videos or store opening hours can also be displayed.
Increase in mobile conversion
The smaller the display, the more important the relevance of the search results. Shop visitors should be offered the usability they expect from modern mobile commerce.
Semantic profit optimization
The Semantic Enhancer recognizes what customers really want – those who are looking for a monitor want to buy a monitor, not products with a monitor.
More offline sales through location-based results
Many shoppers research products online before purchasing the item in-store. Location-based results allow you to automatically show online visitors where their nearest store is located along with availability and prices.
Exploiting the hidden sales potential in long-tail
Understanding complex queries and increasing sales of brand-strong niche products - because site searches tend to be conducted using combinations of terms and thus very specific searches. For example, the traffic generated by long-tail search queries is usually relatively higher than that generated by individual search terms. At the same time, the more specific the search query, the higher the conversion.
It is easier to get to the top of the ranking for terms that are hardly ever used by anyone as keywords. Furthermore, a regional reference can be made in the long tail. A user looking for a shop in his city enters the name of the place in the search and thus falls into the Long Tail.
If you optimize for the long tail, you should design the corresponding text in great detail and make sure that many different combinations of search terms as well as thematically relevant terms and synonyms occur. Longer texts also result in more combinations. However, there is the disadvantage that theoretically there are infinitely many possible combinations. Therefore, the primary focus should not be on optimizing for individual or specific long tails, but rather that the content covers the topic of the website as broadly as possible and thus also covers many long tail search queries.
Through a language-independent core technology, any number of international shops can be set up. Search, suggest, and all other key functions of the online shop remain available in any language. FACT-Finder's Worldmatch® core can even take into account the visual similarities of Chinese characters and process all dialects simultaneously. Worldmatch® enables you to compare characters from different languages at the same time - in an error-tolerant manner. For example, you can compare Latin characters with Arabic or Japanese characters in a duplicate match.
Error-tolerant site search
A good site search is also able to offer relevant product suggestions even in the case of spelling mistakes and typos.
4. How can the search experience be designed from the customer's perspective?
Users decide within the first few seconds whether the online shop carries the desired product. In most cases, the potential customer uses the search box on the main page for this purpose.
In addition to being easy to find, good legibility of the font, especially font size, is a basic requirement of the search box. It is essential to recommend meaningful product suggestions or categories to the customers as soon as they enter the first few letters in the search box. Make sure that important information such as price, a short description, or size, is immediately visible. A good search function is also characterized by the range of functions: search suggestions, spelling error tolerance and suitable filters are extremely relevant for user support.
Mobile devices are often used for purchase preparation, research, and information retrieval. Smartphone users often shop on the go and if the usability does not meet their expectations, patience runs out faster than on desktops, which leads to giving up. This is why products and information must be easy to find. The smaller the display, the more important the relevance of the search results. Smartphone traffic is, on average, higher than desktop traffic – but the conversion rate on smartphones is four times lower. Is it therefore important to offer online visitors good usability to make the journey to the shopping cart more attractive.
4.1 What needs to be considered when positioning the search function?
Do not make users search for the search box
For maximum effect, position the internal search function at the top center or alternatively at the top right of the page, clearly differentiated by color, has proven to be the optimum for successful e-commerce websites. It is also recommended that the search box follows the scrolling of your visitors.
Offer a comfortable input option
By default, the cursor should already be in the search box and provide enough space for at least 27 characters.
Make the search button clear
Place the search button to the right of the search box. The most common icon for the search button is a magnifying glass, or alternatively, text such as "Search".
The search box or the standard magnifying glass symbol should be clearly visible on the home page and ready for use. Do not hide the search box in the menu.
Mobile site search should be touch-friendly and users should be able to easily tap the search button, as well as suggested items in the suggestion window, with their finger.
On a zero-results page, in addition to searching in the header, you should display the complete search box again in the main content area, in close proximity to the zero-results message. After all, you want to encourage the visitor to make another attempt, or at least stay on the site, despite their unsuccessful search.
4.2 How should results and filters be displayed?
Results can be ordered by relevance, but bestsellers and product availability should also be considered.
Precision through filters
If the selection of results is too large, the user will be confused rather than helped. You should therefore allow online visitors to narrow down the results of their search and filter by brand, manufacturer, price, color, for example.
Only offer filters that are critical for most users; too many filters can quickly make the search confusing. The more products you offer in a category, the more detailed the filter options should be.
The relevance of specific criteria, such as monitor size or color, varies from product to product. A professional search function displays suitable filters for each search query depending on the article, availability and product group.
Placement of filters
Place the most relevant filters at the top for greater clarity. Search functions can adapt dynamically to visitor behavior and move frequently selected filters to the top. This makes it even easier to refine search results. Placing filters at the top is more noticeable so that even inexperienced users will not miss them.
Avoid double scrolling. If the entire page can be scrolled, the filters should not have a separate scrolling option. In an accordion menu, this can be prevented by showing filter options when it is expanded.
In addition to relevant results for a search query, you can also recommend other similar products or display "Popular items" on the search results page.
If the search was unsuccessful, you should help the user. For example, by offering them tips for better matching search terms.
How should I display the filter function?
Depending on the type of filter, sliders and checkboxes have proven to be helpful.
When a new filter is selected, the product overview should automatically update itself without reloading the entire page. This prevents the page from “jumping up” and the user loosing their orientation on the page.
The loading time for a website on smartphones is often longer than on a PC. This makes live updating less suitable for mobile devices. Rather, use multiple selection for mobile devices with a corresponding button that remains visible while the customer selects their filters. This way, the customer won’t get annoyed by constant updating and reloading of the site, which takes time.
Multiple selection (multi-select)
With desktop applications, you will rarely find the possibility of multi-selection followed by a "Apply filter" button. Multi-select allows customers to click on different filter functions without the results being automatically updated. To activate the selected filters, the "Apply filter" button must first be clicked.
Show number of products
For example, within the filter for “brand”, show customers how many products are offered per each selection so they know how many they can expect.
If no results can be displayed, you should offer online visitors a good way around this e.g. "If you remove [this filter] you will get results again".
4.3 What needs to be considered when it comes to loading speed?
Fast loading times are an important quality of online shops for many users. Slow shops have a higher bounce rate - if a site loads for longer than a second, the probability of users leaving increases. At three seconds, the bounce rate increases by about a third.
Google also prefers fast-loading websites. Since 2010, loading speed has been part of the Google algorithm for determining ranking.
Loading speed not only affects the number of page visits, but also revenue. If a user had to wait too long the first time, he/she will quickly leave the site, without converting, and likely without returning. The result of this problem is that the conversion rate via mobile devices is significantly lower than desktops, even though the number of Internet uses via the smartphone is already higher than via the desktop.
Mobile websites are often packed with too many functions that offer no added value when used on the go. Mobile versions of online stores should be streamlined in order to perform better.
5. What criteria does a professional site search fulfill?
Never having to search for the site search
The input field of the site search should be placed prominently on your website so that it can be found at first glance. Make sure that you always focus on usability when designing your website.
Smart suggestions with the help of the typeahead feature
Typeahead is an additional feature of the site search that increases the user-friendliness of the search by displaying the correct phrase to the customer as soon as the first letters are entered. An optimized site search leads visitors directly and seamlessly to the ideal target page.
Editorial control with the help of a blacklist
Just as your can use the search to direct users to a specific page, you can also use it to keep users away from other content. When you create a blacklist for your site search, you can list words and phrases that are not considered or displayed in the typehead or in the search results.
Text correction and phonetic search to counter typos
Site searches should be able to detect and correct spelling errors and display the correct phrase. There is nothing more annoying to the user than getting a "No results" message because of a typo. In this situation, a phonetic search is also useful, as it also helps against typos by recognizing words that sound alike.
Targeted narrowing of results
In addition to searching for terms, the selection of certain characteristics of the product sought should also lead to the goal. Each search result is assigned certain properties, which are summarized in facets.
Learn from users
In order to continuously optimize the search, it is essential to systematically analyze the search behavior of users and to draw concrete conclusions.
5.1 How is a “search-friendly” data feed structured?
A lack of product data is common which makes it difficult to find your products and leads to lost sales. So what should a data file look like so that your shop functions work optimally and your visibility on external platforms is as high as possible? An e-commerce-optimized data feed has four dimensions:
Whether its search, filter navigation or a recommendation engine, every shop function only works as well as the data from which it obtains its information. The more complete the attribute fields of your product data are, the easier it is for customers to find the right product.
Only when attributes are available in standardized descriptions, formats and units of measurement can they be used to filter search results and for personalization.
Simple product information is not always sufficient for many shop items to be easily findable. With additional keywords, also called tags, you help make your products easily findable under alternative and colloquial terms.
The grouping of products must be easy to understand, consistent and customer-oriented.
In order for a search function to produce optimal results, it must first be set up with your product information. To do this, you need the product data in an orderly structure.
General points to be considered when creating a data feed file:
- Create the export file in CSV format and separate the fields with separators.
- Make absolutely sure that none of the separators used appear in the article data itself. Otherwise the field structure will be interpreted incorrectly during import, the field contents will be mixed up and the corresponding data record cannot be imported.
- Always create the file with UTF-8 encoding.
- As much data as necessary, as little as possible.
- Create each product (record) in a new line. The product data must not contain line breaks.
- Each line must contain the same number of fields, regardless of whether a field is filled or not.
5.2 How can high product data quality be achieved?
The more precisely a product can be matched to a search query, the more relevant the generated product recommendation is. Data must be correct, clear, consistent, complete, understandable and up to date. In order to further improve the quality of the data feed, special attention should be given to structure and content.
Managing duplicates, postal cleansing, adding information – almost every process step in a cleansing project is based on the comparison of data. Matching technology therefore plays a key role in determining whether a data quality initiative was successful or not. With the help of special algorithms, it is possible to replicate the similarities like a human would, and match your data in an error-tolerant and language-independent way: duplicates are found, additional data is correctly added. Both of these are optimal conditions that ensure high data quality is sustained.
Thesaurus and preprocessor allow you to define synonyms, antonyms and words that are substituted ahead of the search process.
Optimize with the Thesaurus
Antonyms are used to remove results that have no relevance to the search term.
Synonyms are used to include terms with the same or similar meaning to the search result.
Optimize with a preprocessor
A preprocessor directs the entered search term to a specified term. This allows search terms to be matched to the product data. For example, a search for "down jacket" will yield poor results because within the database, the products as referred to as "jacket with down". A professional search function will find all possible jackets, including down jackets.
5.3 Which KPIs can be used to measure the quality of a site search?
Search CTR (Click Through Rate)
There are several ways visitors can get to a product detail page, for example through a standard search or through a suggest function. The search CTR shows the relation between the displayed possibilities to search on the website (visits) and the number of visits with clicks on the search function.
Search CTR = (Number of visits using search − Activity × 100) ÷ Number of total visits
The suggest function suggests products to the visitor by auto-completing the search query. Compared to the standard search, the suggest function only displays search suggestions for which results are available. This way, every visitors who uses the suggestions has both a shortened customer journey and check-out.
Suggest rate = (Number of visits using suggest − Activity × 100) ÷ Number of total visits
Standard search rate
Compared to the suggest, the majority of shop visitors use the familiar standard search.
Standard search rate = (Number of visits using standard search rate − Activity × 100) ÷ Number of total visits
Zero results rate
When a visitor uses the standard search function of an online shop, it may happen that the product they are looking for is not in stock or not offered – leading the customer to a zero results page. This is especially frustrating to visitors who come have across the product elsewhere on the site. It encourages visitors to abandon their visit.
Zero results rate = (Number of zero search results × 100) ÷ Number of searches − Activity
Search exit rate
If the visitor does not find the product they are looking for, they often abandon not only the search, but the website overall.
Search exit rate = (Number of search exits × 100) ÷ Number of searches − Activity
Search Filter Rate
If search filters are used, the visitor probably has a specific goal and hopes to find the desired product quickly.
Search filter rate = (Number of searches – Activity with filters × 100) ÷ Number of searches − Activity
Searches per user
This KPI shows the average number of searches once visitors have used the search, which is a good indicator of search quality.
Searches per user = (Number of searches × 100) ÷ Number of searches − Activity
Search conversion rate
This shows how effective the individual changes in search optimization were in the end.
Search conversion rate = (Number of conversions using search – Activity × 100) ÷ Number of visits − Activity
Search turnover rate
In the end, all improvements should result in your visitors generating more sales through search optimization and the Search Turnover Rate shows you how high your onsite search is as a percentage of total sales.
Search turnover rate = (Turnover using search – Activity × 100) ÷ Total turnover
5.4 How can the ROI of your online shop be quickly achieved?
When calculating the ROI, it is crucial how high the conversion rate is at each stage of the sales funnel. Conversion optimization helps you identify strengths and weaknesses within your website. This way, you have a better overview of ROI and increase your success with optimizations.
While optimizing, the whole website is considered. It is possible to analyze at which point visitors leave the site and then optimize it. A well-planned online shop will be more useful for your visitors and have lower bounce rates.
It is necessary to adjust content, images, and frame elements to improve the usability of your website. Performance improvements reduce the cost per conversion and boost the ROI to increase the long-term value of the entire online presence.
6. What control options should a site search offer retailers?
Transparency instead of a black box: a good search function is able to show you why a product is positioned where it is within the search results.
Ranking rules play an important role in every search. If you do not use any ranking rules, the results list is sorted according to relevance or, if the relevance is equal, in the order of the database entries. This general something you don’t want, but more importantly, what customers don’t want.
The order of the products in a search must meet two criteria: that of your customers, according to new items, attractive offers, or important brands, as well as yours, according to sales frequency, margin, or importance.
Ranking rules can be used with search functions at the product level. For example, you can feature private labels first and highlight best-selling products by displaying them at the top. This strategic positioning can also be used to relocate products to the bottom, for example items that are out of stock or those that don’t have product images.
7. What analysis and testing options should a site search offer retailers?
A/B tests are relevant for all companies that are active online. With A/B testing you can put all your measures for increasing sales to the test. Based on the results, you can determine the most effective ways to improve your online shop and provide data-based proof of how much revenue your decisions can generate. This creates ultimate visibility for e-commerce managers, CRO experts and on-site marketers.
8. What types of integration are available for your online shop?
Web Components – ROI in record time
- Front-end adaptations in no time
Whether you want to integrate new shop functions, configure existing ones, or set up A/B tests, you can now do all this in record time. It has never been easier to react to new requirements and trends.
- Updates without advanced planning
You will only need to make minimal adjustments for future new version releases of your search function. And when you update your shop system, the Web Components ensure that all functions of the search solution keep working without restriction.
- Compatible with all common browsers
Web components display your search function directly in the browser and can be styled according to your look and feel. They function regardless of which shop system and which web technologies you use.
Web Components are not yet readily available in every browser, but you can upgrade this functionality using polyfills. This can result in performance that does not match the native implementation and can cause an additional burden for the user, since the respective script has to be loaded and executed.
REST API – the web standard
REST API is always independent of the type of platform or languages used as it always adapts to the type of syntax or platform used. This gives a great deal of freedom when changing or testing new environments within a development.
A disadvantage of REST APIs can be found in the lack of standardization, which can lead to potential misunderstandings.
9. What special requirements do B2B shops have for a site search?
Online visitors consider a long, cumbersome product search to be a nuisance, especially in B2B where the product range is comprehensive but the time of the buyer is limited. In fact, for 74% of B2B buyers, a sophisticated site search is the most important shop functionality.
B2B eCommerce is all about customer centricity, so personalizing your shop is of critical importance. This is where the search function makes a valuable contribution. Using is a Machine Learning module that can personalize the search results based on the tracked user data. Through this, your customers receive a tailor-made and highly relevant product selection.
Particularly in the B2B industry, long-term and strong customer relationships are crucial to success. A personalized environment that features customer-specific catalogues, offers and prices as well as customer-related product ranges, is necessary. A B2B online shop that offers a personalized experienced that is well thought-out and structured, can contribute significantly to customer loyalty, increase the duration of visits and is rated as relevant by Google. For 68% of business customers, the display of individualized prices is one of the most important shop functionalities.
10. Open source or standard solution: which is better for your site search?
Which option to choose depends on how much time, costs, and resources a retailer is willing to invest.
Contrary to what the term "open source" might suggest, in-house developments usually incur higher costs than third-party solutions – certainly if the internal site search has to meet a certain standard. If time and costs are not an issue, an open source search like Solr can be set up so well that it can compete with the best solutions on the market.
Solr – strong in full text search, but without customization it is inadequate for shops
Most solutions based on Solr or Elasticsearch use the Levenshtein algorithm introduced in the 1960s. This is a proven method that uses a database query to determine similarities between individual words. It is ideal for full-text searches, such as Wikipedia. However, since the method works word by word and does not look at the entire search phrase, the Levenshtein method shows weaknesses in search queries that consist of several words.
To counter this, the scoring of the search has to be adjusted with the help of plugins. The same applies to the adaptation of the search to other languages.
FACT-Finder – e-commerce search ready for use out-of-the-box
With FACT-Finder’s search solution, it's the opposite: when it comes to full-text searches, it is inferior to an optimally adjusted Solr, but scores significantly better than other searches when it comes to international e-commerce. A/B tests show that FACT-Finder can increase sales by 10% to 15%, compared to a well-adjusted Solr.
According to the provider, the built-in FACT®-Algorithm recognizes similarities in the same way a human does, based on different criteria.
FACT-Finder does not reduce the word to its stem and try to find exact matches. Instead, it simply searches the query as it is and assigns relevance to all products in the data feed. There is no such thing as "not found", at most "nothing found that is relevant enough according to user settings". FACT-Finder also has no problems with searches with multiple words. FACT-Finder is especially powerful when searching for long-tail terms, compared to other search providers. And because long-tail can be very lengthy (consider the amount of typos you can make), this is one of the reasons for the increase in sales generated by FACT-Finder.
11.How can AI improve the customer search experience in e-commerce?
AI is able to provide customers with customized offers. This allows them to make more precise purchases and saves them a lot of time leading to a more efficient purchase process and increasing customer satisfaction.
In the field of e-commerce there are countless applications that can be optimized with AI. Currently, usability and product recommendations are very popular because they are comparatively easy to implement and have a high potential to increase the conversion rate of an online shop that uses this type of AI.
Retailers who know their customers can provide them with better service. Because there are many different types of customers, it is not so easy to know them all. This is where AI steps in to analyze each customer and assign them to a category. Online retailers can then create a unique offer for each customer category that the AI can then display to the customer.
With growing amounts of data, intelligent data processing is becoming a key subject in the e-commerce industry. Especially in customer experience, where users ultimately decide how and where they navigate, data mining and Machine Learning are valuable tools to help businesses get real value from their own data. This enables shop managers to actively react to the individual behavior of their shop visitors and to recognize which product recommendations and which content is suitable for the individual user. AI helps shop managers optimize their online shop in a purely objective and data-based manner, without losing time with data analysis.
11.1 Which AI technologies can be used today for site search?
Text analyses make a significant contribution to setting the shop system apart from the competition. An intelligent search must be able to recognize the user’s intentions and determine suitable product suggestions – even if the user can not clearly name them in the search. The customer's search query is examined by the AI and enhanced by semantic and contextual information. Using the enhanced search query and other similar queries analyzed by AI, the customer is shown a highly customized list of products.
Search functions are usually structured in such a way that they only filter for certain keywords. But what if the user does not know exactly what he wants? When entering a search term such as "clothing for golfing", not every search function can provide appropriate suggestions as the topic is too broad. A chatbot, which captures the input through a semantic analysis, can ask specific questions and make suggestions.
A chatbot also provides customers with an opportunity to ask questions throughout the ordering process. Since customers receive an immediate answer, they are less likely to cancel the order. A chatbot is a mixture of semantic analysis and automatic text generation so therefore it should be clear beforehand which problems the chatbot is supposed to assist customers with and communicate this to users.
Modern systems are also able to recognize the mood of the user and adjust the conversation behavior accordingly.
Today's Machine Learning methods are able to use visitor’s click history to determine the preferences and interests of individual customers and tailor the offers specifically to them. This could be specific brands, colors, categories, price ranges or even special product descriptions to which the customer will more likely respond to.
Buyers often have to procure large quantities of the same products and therefore prefer to avoid the time-consuming process of recalling each individual C-part. In this case, it is an advantage if an online shop suggests the needed products right from the start, so all that the buyer has to do is drag and drop the products into their shopping cart.
Machine Learning tools utilize tracking data and evaluate a customer’s buying patterns to predict which products are most relevant to the buyer during that specific time.
FACT-Finder's Recommendation Engine analyses frequently occurring product and category relationships. It examines purchase-relevant data from the online shop's purchase or click history, and determines the products, categories, and attributes that match the product. The results are intelligent recommendations that can automatically appear on product detail pages, the home page, or in the shopping cart.
This is a function that detects connections between search queries and subsequent customer purchases. The information obtained is automatically fed back into the search results. With every click on product detail pages, every shopping cart content and every completed sale, the quality of results increases – as well as customer satisfaction.
11.2 Which innovations are on the verge of a breakthrough in e-commerce?
In the coming years, there will be developments that will completely change the way people shop.
While home assistants like Alexa are very popular, almost no one uses them to make new purchases because it is much easier to compare and select an item that you can see.
Voice commerce, however, is already a useful tool for reordering familiar items ("Alexa, buy detergent") - especially in combination with Predictive Basket technology.
If your industry is one where customers order the same products repeatedly, such as food, pharmacy, drugstore or B2B, AI can enable you to transform the purchasing process. With the newest AI technology of the Predictive Basket, you can accurately predict what customers will buy in their current session. The AI calculates the suggestions based on the behavior of the individual customer and on the behavior of all other customers – this way it learns the customer's buying patterns and adapts to it most precisely.
Visual search is also based on neural networks, but the technology is not yet as reliable as speech recognition. However, once its development is sufficiently advanced, visual search will take a firm place in the market. After all, it is very practical for customers to simply photograph an object they are interested in and the system inform them which product it could be or suggest similar products in the product range.
An accurate semantic search understands entire sentences like "I'm looking for a flight over Christmas with my young daughter to a beautiful beach". To do this, it uses a network of words and terms and their references to one another. However, many customers do not want to type as much – many enter only a single word. Therefore semantics is currently used for other purposes.