TAGS: AI CX Merchandising Online Shopping Retail
Optimise onsite search with part 5 of our comprehensive guide.
In part 3 and part 4, we highlighted ways for retailers to help shoppers do great searches. In the final blog post of this series, we will focus on how to make it easier to navigate even the largest search results pages.
An area that is often overlooked when thinking about optimising search is what happens when the right results have been returned to the shopper. Retailers should aim to help customers refine and browse search results to continue the conversation. This follows the same rationale behind helping shoppers express their needs: search results become more relevant the more context we can ask or gather from them. The reason is simple. Even if a retailer has done everything well so far, it is not unusual to end up with hundreds or even thousands of equally relevant results. With numbers like this, it's impossible for shoppers to review every single product.
When studying shopper behaviour for these kinds of search results, we notice a steep decline in attention to products further down the page, which translates to much lower click rates of any result beyond the first dozen or so. The image below shows how quickly click-through rates drop during a typical fashion search.
If relevant products are not immediately shown, retailers run the risk of disappointing shoppers. To surface key products, there are two options: help the customer remove irrelevant results or use product ordering and page structure to push relevant products. We will look at both options in turn.
A natural way to continue the search conversation is to ask the shopper to further specify what they are interested in. For this reason, filters or facets have long been an integral part of almost every search interface.
Getting filters right is not always easy. Firstly, filters should be arranged by their ability to remove irrelevant results. Generally, the most important filters are those that shoppers don't want to remove when browsing results. Gender filters in fashion are great examples - it is unlikely that a shopper needs to change the selection in one browsing session. Depending on the exact business, category and brand filters typically fall into this category too. During sales periods, when stock becomes fragmented quickly, promoting the size filter can also become important to avoid disappointing shoppers.
While it may sound appealing to re-order facets automatically based on context and shopper information, this should be done carefully. It is important to keep some level of consistency to prevent confusion with ever-changing main navigational elements.
The next challenge is deciding which options to show first in each filter. A general rule of thumb should be to follow any natural order or screening mechanism shoppers would typically use.
For example, size facets are typically sorted from small to large. If there are multiple size ranges they tend to be grouped like at Missguided.
For brand facets, where no clear order can be defined, using alphabetic sorting like at Selfridges helps shoppers because they can easily spot where in the value list they need to look. That way there is no ambiguity.
If there is no natural order, using popularity or the number of values as proxy for popularity is generally the best option.
Traditional faceted navigation faces another challenge - one that is often overlooked. On desktop, facets go easily unnoticed because product imagery is designed to draw attention away from navigational elements. On mobile, facets are often hidden altogether. These two factors lead to shoppers browsing results rather than filtering them.
To help shoppers recognise the value of filters we can draw attention to them by positioning them more prominently and using imagery.
Screwfix positions key filters above the results and uses pictograms to clearly mark these as navigational elements, compared to products.
These key filters don't necessarily need to be category filters. Thinking about the multitude of different customer journeys, it is often possible to define different personas that may end up on a particular result and offer relevant options for most of them.
For example, Debenhams offers a split by occasion (formal or casual), by brand or style of shoe.
Search relevance should be the most crucial factor in search rankings. There is little use in promoting trending or high margin products if they only partially match what a shopper was looking for.
Among the most frequent cases of partial search relevance are accessories. Shoppers looking for 'iphone', for example, are most likely interested in the phone, not in cases or cables. In this instance, demoting all accessories can be beneficial.
Identifying what is relevant can be a challenge in itself. Using clues from what type of products other shoppers bought after searching for a specific term can be helpful to identify what should be considered relevant. Looking at search refinements (which searches were started from a search results page without clicking any results first) often gives a good indication of what shoppers were interested in.
If the shopper's interests are known, search results can be re-ordered to show the most relevant results. Ted Baker shows a different product order depending on whether a shopper is interested in menswear or womenswear. Generally, boosting relevant products is preferred over filtering out potentially irrelevant products.
What exactly to promote can depend heavily on the retail sector. In sectors with high repurchase rates, such as groceries, pushing recently bought products in the result set can make shopping more convenient.
In other sectors, the opposite might be true. When selling appliances, it's probably best not to promote more refrigerators to someone who bought one last week.
We also need to consider the case when we know nothing or very little about the shopper. In this instance we can fall back on global popularity and trends. They are often good indicators for personal preferences. Generally, it's recommended to mix knowledge of personal preferences with trends across the customer base to prevent filter bubbles and statistical noise generated by having only few data points about customers.
When deciding which products to push, it's important to ask yourself what the KPIs of the merchandising team are. When running strategic product ranking projects with retailers I find that the answer is rarely just conversion rate. Often, there is some element of profit or margin involved.
If that's the case then how important a product is to the business should be the final ingredient for a good search ranking. While it is always good to give the highest priority to shopper interests, sometimes they are not aligned well with those of the business. During end of season sales it may be critically important to shift as much stock of a particular product as possible, but at these times shoppers also tend to be more forgiving if they need to scroll a little further.
Finally, there is an option to combine the filter and the ranking approaches and combine them into one. On search results pages retailers should try to show different interpretations of the same request. A straightforward way to do that is to show one interpretation per row. When customers search for generic words like 'blazer', ASOS shows a mix of its own and other brands' menswear and womenswear.
Just like highlighting key facets, this can help shoppers understand that the search result set is diverse and that using filters can make results more relevant.
At the start of this series of blogs we spoke about why emulating in-store conversations is so important to online retailers. Over the past five weeks, we have examined how best to optimise onsite search to make this possible.
Our aim with the series has been to provide retailers with an armoury of practical tips that can be easily implemented.