What is the priority of importance to my product discovery journey?

Within the last 5 years there has been a change in the eCommerce space that has led product discovery technology road-maps to shift towards a consolidation strategy. In the case of Klevu, a best-in-breed search solution, this has meant branching out into merchandising and recommendations and the same can be observed for almost every other product discovery player in the market.

The reasons why a merchant would choose to consolidate in this way are clear – one account manager, one invoice/renewal, one technology for their team to learn, and most importantly a unified onsite product discovery journey driven by a single platform. This approach simplifies operations.

The market’s response to this consolidation offering has been mixed, some have opted to stick to a best-in-breed approach whilst others like Paul Smith have explored consolidation. And, whilst most product discovery solutions in the market will now in some way position their solution as ‘the best at everything’, the reality is each will have varying degrees of strength across search functionality, effective merchandising strategies, or personalized product recommendations and the modern buyer will be in tune to this.

This means that decision-makers are left asking the question ‘On balance, what is the best product discovery technology for my business?’ and part of answering this question involves also asking ‘On balance, what is the priority of importance to my product discovery journey – search functionality, digital merchandising, or personalized product recommendations’?

This article will attempt to unpick this question and ultimately make a case for why the importance of search should not be understated. Of course, every business has its nuances (i.e. if you are working with a subscription platform selling a handful of products, these cases may not apply) however for the vast majority of merchants (who would consider consolidation in the first place), these arguments can all be applied.

Managing search queries is notably more challenging

I have worked in sales within product discovery for 6 years, and the biggest pain prospective clients face with product discovery is the fact that search is impossible to manage with their incumbent technology.

The purpose of any search solution is to match a customer’s search query to a relevant SKU within a catalog. If you go to your Google Analytics and check your or your client’s monthly search usage and compare the respective number of SKUs within that catalog, you should start to get an idea of how impossible this becomes to manually manage without a best-in-class search solution. It simply isn’t feasible for most businesses to manually optimize every single search query to the relevant product(s) in a fast-paced eCommerce environment.

Search is vastly more unmanageable in contrast with browser journeys where products are already conveniently organized within a comparably small number of category pages and to a lesser degree product recommendations where you’d need to manually optimize recommendation banners on the homepage, product page, and cart page. To summarize this point, the nature of search is such that your business’ search volume and the number of SKUs within your catalog make search significantly more unmanageable than other aspects of product discovery.

Brand reputation and search accuracy

In recent years it has become apparent that more so than ever C-suite executives, directors, CEOs, and business owners all take a particular interest in their business’s product discovery journey. We’ll often speak to eCommerce Managers who will highlight that they will receive messages from senior management asking ‘When I search for this product, I get no search results, why?’ or ‘If I search for x product, y product is not appearing, why?’

Given the focus in the eCommerce space is largely commercial i.e. increasing conversion, you would be forgiven for assuming that this level of interest from senior management is purely driven by this same set of interests and values. However, when digging beneath the surface in discovery calls, what I often learn is that this interest in search largely relates to the brand’s identity and perceived credibility.

Business owners in particular feel that a sub-standard search solution is letting down their customers, leaving a bad taste in their mouths and thus having a knock-on effect on their brand’s reputation. So why is this so in the case of search more than other aspects of product discovery?

In an era of search where consumers expect precision similar to Google and Amazon, any shortfall in search functionality is painstakingly obvious. If a simple typo breaks a search query entirely, it’s obvious. If an Iphone is the first product that appears when a customer searches for an ‘Iphone Case’, it’s obvious. These are obvious flaws that a customer is not used to and will observe, be frustrated with, and therefore have an impact on the brand’s identity and credibility.

To use the age-old physical shop analogy, my most frustrating experiences in physical stores have been when in a rush to locate & buy a specific product (i.e. site-search) and a customer service representative has been unable to help me do so when I am certain the product is within the store. In my head, I say to myself ‘I am a game customer, ready to give you my money, help me find this product and I will hand it over’. The high-intent and urgent nature of my shopper persona is the very thing that causes my frustration and subsequent impact on my perception of that brand.

Enhancing conversion rates via search optimization

Due to the aforementioned high-intent nature of customers that use site search, there is also a commercial argument to be made for focusing on these shopper journeys. As part of the sales process at Klevu, we will request a prospective client to share several metrics from their Google Analytics to ascertain whether they are a good commercial fit for Klevu. Two of the metrics we request are conversion rate & search conversion rate.

In every single case, the conversion rate for customers who use site search is higher and in most cases 2-4x higher than the non-search customers. To model this, a 10% CVR uplift on a 1% CVR for non-search customers will increase this CVR by 0.1%. However, a 10% CVR uplift on a 3% CVR for search customers will increase this CVR by 0.3%, which is a 3x greater increase than for the non-search customers. Due to the scale of uplift for search CVR, focusing on optimizing these search sessions is a more efficient use of time, resources, and cost than the non-search sessions.

The second part of the argument is simply that the increase in CVR and sales from optimizing search can be observed as being significantly higher than other aspects of product discovery. Having worked previously at a market-leading merchandising platform early on in my career, one of our biggest challenges was getting case studies that included CVR uplift-proof points.

Whilst our client’s merchandisers loved features that allowed them to curate category pages in a way that made them feel empowered, there was a question mark over whether it was adding commercial value to their business and this led to a gulf in case studies that included credible proof points. In contrast, at Klevu we have a wealth of case studies to choose from (34 currently) to reference in our sales conversations and use as benchmarks in estimating CVR uplifts for prospective clients.

Decoding the complexity of search: How best-in-class search solutions outperform the others

One of the advantages of starting out as a search solution is the fact that search is undoubtedly more complex than other aspects of product discovery. Why? This is largely down to the Natural Language Processing element that is far more comprehensive in search than merchandising and recommendations. In relation to search, there are three Natural Language Processing variables that can vary from client to client:

• Data Processing i.e. product data going in (SKU structure i.e. attributes, attribute values)
• Customer search queries 
• How a search algorithm handles the customer’s query and processed product data

What to a customer feels like a seamless search experience is actually driven by highly a complex search algorithm formed from multiple AI functions working together in conjunction i.e.:

• One function will correct a misspelling
• One function will enrich relevant synonyms for the merchant (i.e. linking the term ‘Sneaker to Trainer’ or ‘Dark blue and ‘Navy’) 
• One function will automatically generate relevant search suggestions for the shopper as they start typing
• One function will identify the noun (intent) within a search query and product title to ensure search relevancy

Layered on top of this NLP algorithm is Klevu’s machine learning algorithm for eCommerce which considers behavioral data (clicks & checkouts), a 1-2-1 personalization engine that uses first-party cookie tracking to personalize search results as well as any ‘searchandising’ strategies set up for the client. You’ll also see these types of algorithms used in most merchandising and recommendation platforms. However, it is the complexity of Natural Language Processing algorithms that is unique to site search that gives established search providers and their clients the edge.

In addition to the complexity of search algorithms, it is the expertise of those at established search providers that provide further added value. At Klevu, Klevu’s CTO & Co-Founder Niraj Aswani has been a subject matter expert in Natural Language Processing applications since studying his PhD, and has dedicated his career to applying this expertise in an eCommerce context and sharing his knowledge within the team at Klevu who can then use knowledge to support our clients needs.

Donald Russell learned when trialing a new search provider in the market that trusting new search providers can have a huge impact on search performance and sales. Donald Russell recoups lost revenue within 1-week of switching back to Klevu from Nosto search. At the time, their Customer Success Manager and I were discussing this and he said bluntly, ‘Everyone thinks they can just build a search solution, it isn’t that straightforward when you understand how complex it is’.