Introduction

The rise of E-Commerce (mobile) search

In the last years, E-commerce search has been growing in importance, partially because of the growth of mobile commerce. While 2021 brought a 10% to 20% revenue growth for many e-shops, many experienced a growth of 80% to 100% of revenue generated by mobile customers using the search of the e-shop.

Search is now one of the key pillar of e-commerce as visits using the search bring 54% of the total online revenue, of which 57% is done on a mobile device, according to our benchmarks.

This should not come as a surprise, as on a mobile screen, navigating through menus and pages to find products is time consuming and therefore search offers a better alternative.

Customer expectations for good UX is quickly growing

On many leading e-shops, the search experience has improved a lot recently, especially on the mobile, raising customer expectations to new heights.

Typing on a mobile is more difficult than on a desktop, and now a fast search preview completing the search while the customer is typing and suggesting relevant products is expected.

Example 1: Good textual with narrow down arrows

For example, many e-shops like http://otto.de or http://zalando.de are completing the user search using a right arrow on the mobile enabling a fast completion and narrowing event before going to the search result page.

As an example here, on http://otto.de , typing 3 letters ('her') then clicking twice on the arrows (therefore staying on preview without changing pages) and then clicking a suggestion brings the user to only 8 products, which is easy to go through on the mobile without the need to filter.

Unfortunately, while this can bring quickly to a small list of results, it is not easy to change the search criteria in case the suggested products were not what the customer expected (the user need to open the search bar again, restart the search, or delete the last words, etc.

If the user would simply open the filters (as on the screen-shot here), no option could be changed, even if the user clearly indicated a size, a brand and a product type.

This is why Boxalino Winning Interactions developed Semantic Filters which are automatically set as filters which can be changed, so the user doesn’t need to restart his search to modify his search criteria.

[comment for Michael: this actually doesn’t work, but going through the user experience, I realized this would make a lot of sense to do to show the power of the semantic filters, it could be done quite easily for Flaschenpost which would be a good example case for it, I think, to be discussed]

Example 2: Personalized from prior searches

Many customers need to repeat their prior search to find again the products they were looking for before.

As an example here, on http://ikea.de , searching for “bett” can be easily repeated when coming back on the e-shop at a later time.

However, the user might be interested to do not exactly the same search, but another related search to find something different, which is unfortunately not possible in this example.

Also, as per the screen-shot on the left, the user might have selected several filtering criteria to find the right products and such filtering would need to be redone to find the same products again.

This is why Boxalino Winning Interactions developed Personalized Search Suggestions showing related searches that other customers who made the same search also made, to help the customers explore related products.

Boxalino Winning Interactions also developed the possibility to remember not only the prior searches, but the prior filtering selection as well, suggesting searches like
“bett mit Farbe:rot, Grösse:200/220m”.

[comment for Michael: this actually doesn’t work (the second part) but going through the user experience, I realized this would make a lot of sense to do and while not super easy (needs Simon) would not be very hard either]

Example 3: Full Screen Merchandising start

On many e-shops like http://shop-apotheke.de the simple fact to click on the search bars open a full screen mode, which maximizes the available size above the keyboard on the mobile and which is used to suggest the favorite searches and products directly, before the user starts typing.

Unfortunately, the display here is not optimal (big empty space on the right) and not personalized (it could have personalized banner campaign as in blue).

This is why Boxalino Winning Interactions helps you optimize and personalized the content appearing on starting search preview, because for many e-shops this display is 1.8x more viewed than the banners on the home page.

A better Search with Boxalino Winning Interactions

Boxalino Winning Interactions is the most advanced technology to fully benefit from the values of the cases described before and to go beyond their limitations.

For example, on our client’s web-site PerfectHair.ch, the leading swiss e-shop for hair and beauty product, it all starts with the first click on the search bar.

Let’s start with the mobile.

The search bar expands in full screen-view and the customer is presented with 3 sections of content, each clearly visible even as the mobile keyboard is open:

• `Oft gesucht`, which presents the current search trends on the e-shop

• `Suchvorschläge`, which retarget searches recently done by the customer and completes them with other searches often done by visitors doing the same searches

• and `Empfohlen für Dich` which provides personalized product suggestions based on the prior clicks and purchases of the customer.

In this first step, before even typing the first letter, many customers will already fine what they wanted!

Then, when the user types the first letter (let’s say 'b'), a list of meaningful and non redundant suggestions are directly appearing.

In such a case, it is too early to know what type of products the customer wants, and therefore the textual suggestion (which are the most helpful at this stage) play a dominant role in the limited display, occupying most of the space with 6 suggestions, even if the suggestions of products is already hinted on with the first one appearing before the keyboard on the bottom of the screen.

The customer can now chose between selecting one of the suggestions by clicking on the term (which goes to the search page), clicking on the arrow, which fills the word while updating and staying in the search preview or keep typing.

Let’s imagine the customer will now continue typing.

On a mobile keyboard, it is very easy to click on the letter next to the one we wanted, and this is what is happening to our customer typing ‘babx’ instead of ‘baby’.

As this is a clear and common mistake, the system corrects the spelling seamlessly and directly like if the mistake was never made.

While it seems now quite clear that the word ‘baby’ is limiting the product possibilities, there are still many many possible products and therefore the textual suggestions are still fully dominant in the display.

The customer decides now to click on the arrow of the first result ‘BaByliss' …

As 219 products are available in the brand BaByliss and therefore it is still not the time to focus on showing directly product, even if it is very possible the customer is interested in one of the top matching products of that brand.

As the user clicked on the right arrow which updates the search preview, the keyboard has also now been hidden, giving the full display of the screen to the user and showing the top 4 product as well as the product counts directly on the screen.

The customer has now transition from typing to clicking, after initiating the scope (219 out of over 40’000 products) to navigate within by typing only (and wrong) 4 characters…

…

After clicking the arrow of `babyliss curl` which reduces the selection to 34 products, the customer clicks on the arrow of `babyliss curl secret`.

…

…

The results are now down to 11 and the system switches mode , focusing on showing products instead of textual suggestions.

Indeed, the 6 products shown here represent more than 90% of the sales out of the 11 products of the search results.

It is very, very likely that one of these products is the one the customer is looking for.

But it is not always require to have a long search query like ‘babyliss curl secret’ for the system to switch mode.

For example, 3 letters might be enough, like ‘ola’ which clearly predict the brand Olaplex, with only 17 products, of which 6 represent also the vast majority of the sales.

Going from over 50’000 products down to a selection of 6 with only 3 letters and without even changing page once, this is possible with Boxalino Winning Interactions!

On the desktop, the larger display enables many elements to be displayed at the same moment, which the Boxalino Winning Interaction smart search fully exploits, as in this example for ‘color’, showing most common searches, brands, product lines, categories and matching products, all in one structured and clean overview.

Underlying numbers (to be removed / changed):

• Desktop Conversion Rate is typically 2x higher than Mobile Conversion Rate

• But the difference is only 1.5x considering visitors using the search

• 54% of e-shops transactions done by visitors using the search

• 57% of these transactions where done on a mobile device

• 1.8x more click on the search bar VERSUS visits to the home page

• 3 average search results page views / session
  VERSUS
  30 average search autocomplete preview / session (10x)

• 62% of visitors using the search click on suggestions from the search autocomplete preview
  

• 24% of visitors using the search find directly their product from the search autocomplete preview
  might be a bad interpretation
  On a well designed autocomplete preview, close to a quarter is buying directly from there instead of going to the search result pages, which shows how well accepted this better user experience is
  

• 28% : session conversion rate of visitors clicking on suggestions from the search autocomplete preview, this is often over 7x higher than the average session conversion rate of the e-shop
  

• 25% of searches (after spelling correction) can be redirected automatically to smart brand or a category page

• 81% of searches rely on Winning Interactions Automated Semantic Filtering** Self-Learning
  

• 11% of searches rely on Winning Interactions Automated Weighted Synonyms*** Self-Learning
  

• 81% : Prediction Accuracy**** in Winning Interactions Automated Query Suggestion
  

• 2.5% of search queries automatically corrected for their spelling
  

• PH - nov-dec 2021 versus nov-dec 2020

  desktop revenue : +11%
  mobile revenue : +43%

  +23% increase of mobile revenue not using the search

  +83% increase of mobile revenue using the search
  

** The Winning Interactions Artificial Intelligence learns automatically interpret the meaning of the keywords of the search of the users to to change the search results or redirect them to the appropriate page

*** The Winning Interactions Artificial Intelligence learns automatically synonyms which are relevant for the e-shop based on an open thesaurus and the search online behaviors of the clients on the e-shop

**** calculated on the frequency of textual suggestion matching the next typed character by the user