NoSQL Caching and Search Capabilities

The pre-standard normalized set of data that has foreign keys and primary keys in the database tables and the ability to interact with the data is virtually unlimited when you’re looking for a website speed-increasing eCommerce solution. But whenever it comes to the search results in indexing, we flatten this data out for a particular scenario. This is true whether it's autocomplete searching, being able to show search results grid or list view, or being able to run a product selection wizard. 

These can all be taken as inputs into their UI. The data is flat, so every time there is an update to the product data, every time there is a change, we can reindex that particular subset of the data that's changed and been updated. Then we can basically keep updating the search results that the end-user is getting. Search results in this cache are always kept up to date in real-time so that as the user searches for these complex, taxing searches, they're not hitting the system’s database or infrastructure very heavily. They're just hitting on these cashed sets of data. 

Feedback Loops

The other thing about the architecture that's very important — and commonly overlooked — is feeding back into the system and having a feedback loop. An eCommerce business needs to be able to have a feedback architecture wherein different users’ patterns are analyzed and measured. How far do they get in the search process when they view a particular item, and how much time was spent on the detail page? Did they put it in their cart and then abandon it, and how many actually come back and buy? It’s important to analyze the funnel to see what aspects of the entire process drove them to that final purchase. 

From a data perspective, this type of eCommerce business intelligence provides data that a site can accumulate over time and then feed back into our search results to provide better autocompletes and synonyms when someone uses a particular keyword. All of this data is information that users are giving a business for free. It's very important from an eCommerce architecture perspective that we have all these fields in categories, products, tags, and meta descriptions labeled properly so that we can architect the data intelligently. 

Following Up

Once the search is launched, it's very important that the eCommerce architecture developer provides some level of data analysis within the application itself so that a business can go back and intelligently adjust the results. Aspects such as bounce rate, time looking at a product, and abandoned carts can help improve the search for the next customer. 

For example, if the search showed something very high in a search result but the business notices that there's a high bounce rate for this item, parameters can be set up so that the vast amount of user information can override initial results. This can be user-specific as questions are asked, such as: 

• What type of customer are they? 
• How are they navigating the site? 
• What type of persona do they have? 
• Are they a visitor or a returning customer? 
• Are they primarily a window shopper, and how can what’s presented to them make them purchase? 

The eCommerce architecture may need to determine what persona the user has and then apply an analytics feedback loop to be able to show them a category or product type customized for them. 

Find the Right eCommerce Solutions for Search

This level of sophistication can provide a major lift for companies looking to improve their eCommerce business model. Yes, it will take significant effort to incorporate into a system, but doing so can boost sales and bring customers back again and again. 

Clarity recommends looking for a custom eCommerce platform that has these concepts already built versus having to build them from scratch. Even the largest SaaS platforms can’t provide search like ours.  They simply don't have this level of sophistication built into their applications and can’t deliver this level of capability. Contact us to see examples of what we can do. 

Within an eCommerce architecture and framework that allows strong search capabilities, it's very important to consider how a business website is physically working when it searches for and retrieves user queries. This has to address business intelligence for eCommerce and logic, as well as the physical mechanisms for presenting the data. One important thing that Clarity Ventures has noticed in our 15 years is that presenting the data with a direct database query is not going to be effective at scale. This holds true for filters on custom fields and attributes dealing with hundreds of thousands — and sometimes tens-of-millions — of products and their category associations. Direct database deliveries simply won’t supply a fast enough result if we’re not using some form of NoSQL caching. 

From an eCommerce architecture perspective, there must be some form of NoSQL caching; it's just a fundamental requirement for big sets of data. Whenever doing a search, users expect the search results to come back in under a second. In order to perform at that level, an eCommerce developer must have the data cached and a very flat set of architecture where the system can get the data based on the user’s inputs. All this must happen within a few hundredths of a millisecond. NoSQL is excellent at being able to take a significant number of inputs and bring back the data from a set of millions of permutations, then effectively present the results in the near-instantaneous fashion that people have come to expect from top B2B eCommerce platforms.  

We strongly recommend incorporating these types of infrastructure components, and that's why we use programs such as Elasticsearch and Redis for our search results within our eCommerce framework and architecture. Both are great applications using the concept of caching and NoSQL.