Following on from my review of NOSTO recently, I decided to write a detailed review of Klevu; a leading eCommerce search engine for mid-market and enterprise online retailers. I first became aware of Klevu in ~August 2015, when I met Founder Nilay Oza in New York. Following this meeting, I ended up consulting for Klevu and still advise them on various elements of product development and platform integrations today.

So, let’s start with exactly what Klevu is and what their average customer looks like. Klevu is a search (and now category merchandising) suite that excels in three core areas (in my opinion), which are:

  • Automated influence on the catalog and context of results – Klevu is hugely focused on natural language processing, which forms a big part of their proposition and also machine learning, which provides actual retail-focused context around results (by optimising results based on clicks and sales). This is a big thing when you compare Klevu to a lot of their competitors, as there’s not the same level of influence around popularity.
  • Merchandising of results – Klevu allows admin users to control which products are most prominent in results at different levels, which I think is hugely important. I’ll go into this in more detail later on.
  • Platform integration – Klevu have built strong integrations with their primary platforms (Magento, Shopify Plus and BigCommerce) which allow users to get set up relatively quickly and easily. Again, this is a big pro against some of their competitors and the integration is often also deeper (e.g. being able to preserve your SRLP template in Magento).

Klevu’s average customer is likely to be a retailer turning over more than £1m online with an interest in creating a better UX and improving conversion rates. Common types of users for Klevu include:

  • Relatively complex B2C retailers – e.g. Toolstop, Made.com, Cox & Cox, Builder Depot, Richer Sounds
  • Fashion retailers – e.g. Sunspel, Victoria Beckham, Cluse, Me + Em, Agent Provocateur, Helly Hansen
  • High volume retailers – e.g. PUMA, Liverpool FC, Made.com, Zulily, Color Pop
  • B2B retailers – Alimed, LeCot, Ingredients Online, Trippline, Amphenol, Mid South AG, MVE
  • Simple retailers wanting to provide a better UX – e.g. Bulletproof, Skinny Dip, Bjorn Borg, Nails Inc

Here are some of the key features and selling points of Klevu:

Customisable JavaScript overlay / auto-complete interface

Currently, Klevu has two default overlay templates, which can be customised (via HTML and CSS directly or Klevu can implement JS customisations). The two default layouts are a grid view option with filtering in the overlay or a list view layout, which is faster and cleaner. Below are some examples of customised versions of the JavaScript overlays.

Made.com – Magento 1 (products, keyword suggestions and hard-coded content links)

Pai Skincare – Shopify Plus (products, keyword suggestions and content results)

Cox & Cox – Magento 2 (products, keyword suggestions, categories and hard-coded content links)

Me + Em – headless implementation (products, keyword suggestions + hard-coded content links)

Klevu are currently in BETA with their new JS library v2.0, which allows developers to completely customise their JavaScript overlay and even add in custom result types. This is more in-line with how you’d work with something like Algolia and it’s a big move in the right direction for more complex implementations. 

Currently, I really like how simple the templated approach is and it allows for a good level of customisation, however, some larger retailers want to be able to completely tailor overlays for different devices and do more with it. From what I’ve seen so far, this is really nice, but I think the templated approach is still a good option for most standard use-cases.

Keyword-level, rule-based and global merchandising / boosting

Often, the main driver behind using a third-party search solution is to be able to merchandise results, with very few eCommerce platforms offering this capability (e.g. Shopify Plus or Magento). Klevu allows for three types of boosting, these are:

  • Global boosting – this is based on assigning a global boosting score to a product, which then boosts that product across all relevant terms. This is something that very few other search solutions seem to offer, but it’s really useful for dictating the prominence of a product. This score also then automatically goes up and down based on performance.
  • Rule-based boosting – Klevu also allows for rule-based boosting, also based on a score. A good example use case could be boosting all men’s products by 100, which would ensure that men’s products would make the affected products more prominent for generic terms (e.g. a branded term with no specified gender). This can be really useful and the rules are based on product attributes, which could include custom values, such as profit margin or season.
  • Keyword-level boosting – Klevu then also allows users to manually assign hero SKUs at a keyword level, which would supersede the boosting scores applied globally and the machine learning data. This is managed via a drag and drop interface.

Klevu also allows for different types of results to be indexed, including blog posts and content pages. 

Handling of complex product data

Another big selling point of Klevu is that it’s built to natively handle complexities around products, so for example:

  • Configurable products – Klevu would natively show all simple products in search, but users can also opt to only show the configurable product and then inherit the image of the most relevant simple product (e.g. a red shirt for the keyword “red shirt”). Klevu also provides the option of listing the number of purchase options and lots of users have added swatches in the product cards also (requires customisation). These principles are applicable for variants in Shopify too.
  • Grouped products – Klevu also natively support grouped products, showing only the grouped product as an example with a from price. You can also show range pricing against grouped products, but this would require customisation.
  • Customer-specific pricing and availability – with Magento, Klevu natively supports pricing and availability set against customer groups. This allows for customer sprinting pricing, discounts etc and is a common requirement for B2B retailers. This can also be achieved in other platforms or bespoke implementations but would require customisation.
  • Product labels – Klevu can replicate product labels or overlays on product cards, as per a PLP template. These can be achieved via assigning tags or product attributes.
  • Search by SKU – Klevu allows for SKU-level search natively and can apply the same principle for other key attributes.

Search Reporting

Klevu are in the process of rebuilding their reporting capabilities, but the current set of reports gives you visibility over the fundamental areas. Currently, Klevu has reports split out for:

  • Keyword usage
  • Product clicks
  • Geographic data
  • Product sales (by keyword)
  • Error queries / keywords with no results 

These reports can be looked at for different date periods and exported to CSV etc. There’s then separate reporting available for the category merchandising solution, focused more on categories and products driving clicks and revenue.

Personalised search suggestions

Personalisation is something that Klevu is gradually building into the core search product, with a view to soon being able to refine products based on a user’s affinity to brands, categories or products. Right now, Klevu’s main use of personalisation comes in the form of product recommendations, which are displayed to users in the following ways:

Upon activating search – Klevu shows high performing product recommendations when a user first uses the search function. These products are then personalised as the user interacts with products from search. 

For error queries – Klevu provides a set of personalised product recommendations as featured products when a user searches for something and gets no results back. These are personalised based on the user’s behaviour with search. 

Personalisation is likely to be a big part of Klevu’s future, but it’s important to get it right before building it into the core algorithm. It also needs to be tested to ensure it’s as effective as the other algorithms. Beyond this, Klevu will then be looking to build personalisation into the category navigation product, but this is particularly challenging with a platform like Magento where you’re relying heavily on caching. 

The personalised product recommendations mentioned above are available in the premium plus and enterprise packages.

Machine learning and NLP

One of the main differentiators between Klevu and other search providers is their NLP and machine learning, both of which have been built for eCommerce only.

The machine learning side is focused on automated optimisation based on the products that are being clicked and purchased the most – providing almost real-time boosting of items that are gaining traction / performing well. This is an important part of their proposition, as other providers rely solely on relevance (based on product data) and require a lot more manual effort around merchandising.

The roots of Klevu are focused on natural language processing (NLP), with CTO Niraj a real expert in this area. Klevu’s use of NLP is designed to extract more context from queries and then enrich product catalog alongside this – giving them an ability to provide relevant results for longer tail and less direct queries. I’ve seen some really good examples of this in the past – particularly for retailers that get a lot of detailed, generic queries and retailers with more complex products.

Support and on-boarding

Another big selling point for Klevu is their support and how involved they generally are with integrations. Often this isn’t needed, however when it is, very few technology vendors provide the level of reactive support that Klevu do.

In terms of onboarding, this is an area where Klevu are investing in now and they have two people in their customer success team, which is partly focused on onboarding. This is an area they’re planning on building out over the next few months.

Integration

One of the main pros of Klevu against competitors is the ease of integration, particularly on Magento. Although very few integrations with Magento are straightforward, Klevu do have a good, well-supported module (for both Magento 1 and Magento 2) and they reduce the amount of development work needed by allowing users to preserve their templates.  This basically means that Klevu can use the existing Magento templates, rather than build new ones hosted by Klevu.

The module allows for a relatively straightforward sync of product data and also provides options attributes that should be passed over etc. The templates overlay approach is also straightforward and customisations can be handled by any front-end developer.

The integrations with Shopify Plus and BigCommerce are similar but without the ability to preserve templates. With these platforms, users would style the product grid which would be served via JavaScript. The same principles around general setup and the overlay apply.

The other approach, which is used for custom platforms etc, is to use Klevu’s APIs. This is relatively straightforward still and more detail can be found here.

Pricing

Klevu starts from €499 per month (premium package) but then goes up based on the size of the store and features. In my experience, the average £10m retailer will pay around £1,000 per month for Klevu, which would include support, onboarding and would give you all search features. 

Klevu’s category merchandising product (category navigation) is then additional, with that generally being around the same ball-park. 

—

If you have any questions about Klevu or search optimisation, please feel free to drop me an email or add to the comments below.