Typo Tolerance plays a crucial role in making a well-performing search function. Site search or eCommerce search is the most common way online customers discover products. It's such a game-changer for a positive shopping experience that up to 99% of shoppers are likely to revisit a retail website if its search function performs up to expectations.
The problem is the conventional eCommerce search doesn't deal with typos very well while spelling errors are widespread. That's why Boost takes Typo Tolerance seriously. Here is a simple explanation of how our Typo Tolerance algorithm works.
What Is Typo Tolerance?
Typo Tolerance allows the site search engine to understand and correct misspelled words or phrases and still provide accurate search results. It's the ability to recognize spelling errors and auto-correct them so searchers don't have to type the search query again. Therefore, Typo Tolerance is also referred to as auto-correction or spellcheck.
Common Cases Of Typos In eCommerce Search
There are countless possible typos we can make. After researching and watching the search pattern on thousands of eCommerce stores, we find out the seven most popular spelling errors as follows:
- Missing letter: “dress” → “dres”
- Extraneous letter: “dress” → “dreses”
- Letter swap: “dress” → “derss”
- Letter substitution: “dress” → “dreas”
- Word splitting (adding space): “frypan” → “fry pan”
- Word combining (removing space): “frying pan” → “fryingpan”
- Combination of the above, for example, missing letter + letter swap: “vehicle” → “vecile”
In eCommerce search, this feature can significantly improve the customer experience by reducing frustration and helping customers find what they are looking for quickly and easily.
Why is that so?
Why Does Your Site Search Need Typo Tolerance?
To understand the importance of Typo Tolerance in eCommerce search, let’s walk through the basics of a search engine.
Search relevance algorithms are built from the concept of “exact matching". After a shopper enters a query, the engine will scan through all searchable sources such as product name, vendor, description, and other available content (blog and pages, for example). If it bumps into something that looks exactly like the entered query, it will return that “something" in the search results.
No results for misspelt words cause frustration to customers because they are used to the auto-correction in Google search.
There are two main reasons for this. First is conceptual mistakes when customers need to recall the word or phrase because it's uncommon (for example a complicated vendor name). As a result, they just make a guess and input what they think is correct. The second reason for typos is a slip of the fingers. When using mobile devices or non-standard keyboards to search on a website, customers are prone to spelling mistakes. We all agree that mobile commerce is in bloom; therefore, misspellings are becoming more common.
The inability to handle typos in search queries means your site search engine won't be able to generate accurate results for a quarter of searchers on an online store. Such a great loss since searchers can convert much better than regular visitors.
Impact of an unsuccessful search. (Google Cloud for Retail)
Typo Tolerance or auto-correction is implemented to solve this problem. It uses a different approach called “fuzzy matching". With “fuzzy matching", the search engine looks for something that looks not only precisely like but also similar (with a slight difference) to the entered query.
Still, the fuzzy search can be problematic as well. When the search engine becomes too “fuzzy", results with medium to significant differences can also be accepted. This makes the search results inaccurate and far from what the searchers expect.
At Boost, we have been working on an optimal solution to combine “exact matching" and “fuzzy matching" and implement it to our Typo Tolerance.
Inside Typo Tolerance Of Boost AI Search & Discovery
How Boost’s search engine handles a typo
There are two factors in our Typo Tolerance algorithm to help deal with typos more efficiently: Edit distance and Term frequency.
This is the number of steps needed to transform one word into another. For example, to change “dres” into “dress", we need one step, adding an “s" at the end → This is one edit distance. However, to make “vecile” “vehicle", we need to add an “h" between “e" and “c", and swap “l" and “e" at the end → This is two edit distance.
Why is Edit distance important?
Boost's algorithm for Typo Tolerance takes Edit distance into account to avoid too much fuzzy in search results. The larger the edit distance, the fuzzier the search results.
By default, search keywords with up to 4 characters can have a 1 edit distance, and more than 4-character queries can have up to 2 edit distances. The misspelt “shrt”, for example, will get results returned for “short”, “shirt”, but not “shoe", “skirt". The misspelt “shrts" will get results returned for “shorts", “shirts", “skirts", “shoes".
Note: This is a simple example to explain edit distance. In fact, our search algorithms combine Typo Tolerance logic with precision search, partial search, etc, and the store data so the results can vary.
This is the optimal logic we concluded from in-depth research and observation to prevent irrelevant results.
Edit distance is also a consideration for ranking when Term frequency is the same among all possible corrected spellings. The closer a spell-correct candidate is to the original query, the higher its ranking will be. This helps identify which term is closest to what the searcher entered.
Term frequency refers to how often a possible correction appears in the store data. Our Typo Tolerance algorithm prioritizes the term that appears more frequently in the store's catalog as the probable candidate for spell correction.
For example, if your customer enters the query "shrts", and the possible corrected spellings are "shirts", "shorts", “skirts", "shoes". The algorithm will correct the query to "skirts" if it appears more often in your catalog than the rest. If “shorts" and “skirts" have the same frequency. Returned results for “shorts" will have a higher ranking since they need one Edit distance while “skirts" require two Edit distance.
Again, this is just a simple explanation. Our algorithms employ a number of different search logics to enhance relevancy, so the results can be different in reality.
Edit distance and Term frequency reinforce Typo Tolerance logic in Boost AI Search & Discovery and are used to handle all the common typing mistakes with efficiency.
Moreover, unlike Shopify's storefront search which requires the first 4 letters of a search term to be entered correctly for typo tolerance to take effect, our algorithms work like magic even when customers make mistake right at the beginning.
The store on the left hand side is using the default search by Shopify and it can't handle “nermarid" (a typo of “mermaid"). On the righ hand side is a store using Boost AI Search & Discovery and it returns accurate results for “kickle" (a typo of “pickle")
AI-powered search add-on
Above is how the Typo Tolerance algorithm of Boost AI Search & Discovery works in the Basic package.
Boost users can take a further step to make the logic even better with Natural language processing (NLP) and Machine learning (ML) when they use an AI-powered search add-on. NLP and ML powers allow the search algorithms to be more adaptive to each store. This will constantly learn, train, and improve the models through the interactions between shoppers and web pages. Thus, the returned results of a search are getting more and more accurate.
You can try the Boost app for FREE within 14 days. The AI search add-on also comes with a 14-day free trial. No credit card needed.
Don't hesitate to install Boost AI Search & Discovery and see the difference it can make.
Boost’s Typo Tolerance in use
“frypan" → “frynpa". Two edit distance (missing letter and extraneous letter). Store Green Pan
“wooden" → “wooedn". One edit distance (letter swap). Store Yes Bebe
“print top" → “printtop". One edit distance (removing space). Store Barco
“win mau" → “win mai". Two edit distance (adding space and letter substitution). Store Premier Darts
In today's highly competitive eCommerce landscape, it's important to stand out from the crowd. By offering Typo Tolerance in search, eCommerce websites can differentiate themselves from competitors and provide a better customer experience