Is your search spellcheck costing you $3M in lost revenue

Is your search spellcheck costing you $3M in lost revenue?

According to a 2016 WebLinc report, over 30% of online shoppers use the search feature on e-commerce sites. The report also suggests that shoppers who used the site search showed a 216% increase in conversion rate and a 21% increase in average order value.

But often, businesses are not able to convert these shoppers with high intent to purchase, owing to lack of a good site search solution that allows for human errors such as spelling mistakes.

A site search solution that doesn’t take user errors into account will return either irrelevant or no results at all, leading to high search exits. A survey suggests that “73% of customers will leave a site in less than two minutes if they can’t find what they’re looking for.”

Misspelling is a much wider problem than what most e-commerce companies anticipate

Unbxd’s data indicates that about 25% of all e-commerce site search queries are misspelled.

From our research at Unbxd, we find that shoppers tend to misspell the simplest of words. We studied how the most frequently searched for products on a major fashion retailer’s site were misspelled.

The results are shocking:

“Jumpsuit” was misspelled in 216 ways and “bodysuit” was misspelled in 223 ways.

Is your search spellcheck costing you $3M in lost revenue?
Ecommerce Autocorrect

The above-misspelled results are just for simple, single words. The complexity exponentially increases for long-tail queries.

Manual mapping can’t fix misspellings at scale

Usually, most e-commerce sites manually map the misspelled words to the right products. But this means hours of manual effort and intervention. While this might be a great solution to solving small scale problems, it is highly ineffective if you have a large catalog with thousands of products, in which people search using complex compositions of words and symbols.

Even with a large team to handle the manual mapping of incorrect queries to the right search results, businesses can only fix 30%-40% of misspelled queries. The remaining is easily lost out on, resulting in high customer attrition and lost sales.

A smart, automated system solves this at scale.

But care has to be taken in choosing the system. Most popular spellcheck solutions in the market do only a text-pattern matching. They check the misspellings against the basic English corpus and product catalog. They then figure out how many letters are misspelled and match them to the closest possible alternatives.

While this is a quick fix to manual mapping, more often than not, it has huge implications in that they offer wrong solutions. The most popular spellcheck solutions offer accuracy of less than 60%.

A smart spellcheck solution can give you 30% more accuracy than a generic, more commonly available solution.

For instance, the Unbxd spellcheck solution computes the closest possible alternatives for common misspellings against an extensive, context-aware and industry-specific corpus, apart from the basic English corpus and product catalog. Therefore, it offers 30% more accurate results than a common spellcheck solution.

For example, for a fashion retailer, the system understands that “skort” is an apparel category and not a misspelled variation of “skirt.” Therefore, it returns appropriate results for skorts, instead of incorrectly accounting the search query as erroneous.

Is your search spellcheck costing you $3M in lost revenue?

While most spellcheck systems cannot effectively identify more than two misspelled letters in the same word, the Unbxd solution offers 90% accuracy through its ability to correct up to four misspelled letters in the same word. Whereas correcting one misspelled letter in a word is quite easy, the complexity increases as the number of misspelled words go up.

For example: Consider “jumpsuit.” Its most commonly misspelled version is “jumpsuite,” which has one misspelled letter. Most spellcheck solutions can correct this mistake and show relevant results.

Now consider an unlikely spelling mistake in the same search query, say “jmpsuts.” It has three misspelled letters. Most spellcheck solutions cannot correct this mistake, much less show relevant results.

The Unbxd spellcheck solution is smart enough to correct this and return relevant results. Here is a real-world example:

The Unbxd spellcheck solution takes into account a shopper’s vertical and expected keystroke errors to accurately identify the right user intent from the misspelled shopper query.

Similarly, the system takes into account the surrounding keys of a misspelled letter in a word and improves upon the suggestions for correct alternatives. Most users misspell a search query by pressing an adjacent key to a letter by mistake. This is particularly true for mobile shoppers owing to the”fat finger” syndrome, and therefore, spelling mistakes due to QWERTY key positions are more likely to occur.

Spellcheck might only be one site search feature. However, it has a huge impact on your revenue.

Fixing the spellcheck feature could add up to $3M to a typical retailer’s revenue

Site searchers are 2-3 times more likely to convert. Improving the spellcheck solution to account for human errors returns highly relevant results for the 25% of the shoppers who misspell search queries. From Unbxd data, we find that these additional conversions could add a potential $3M to the revenue for a typical retailer.

On the other hand, if the spellcheck feature is not fixed or improved, it could cost businesses more than $3M potential loss in revenue, not discounting high customer attrition and low brand loyalty.

(We shall delve deeper into the analysis and working of Unbxd’s spellcheck solution in the second part of this series.)
To know how much additional revenue Unbxd’s spellcheck feature can add to your topline, contact us at

More To Explore

Learning To Rank with eCommerce data

Learning to rank with eCommerce data

What, Why and How of Learning to Rank: Learning to Rank (LETOR) is used in the information retrieval (IR) class of problems, as ranking related