Skafos.ai

CASE STUDY

We need to find better ways to optimize for the customer and provide them a more helpful way of interacting with lengthy store catalogues rather than just letting them loose on giant product lists in hopes that they find what they’re looking for.


— Luke MacFarlan,
Chief Product Officer, Skafos

Skafos.ai

Company: Skafos.ai
Industry: Technology, e-commerce

The Customer:

As a provider of personalization software to e-commerce businesses, Skafos develops and implements a range of tools that help enable customers to identify what they want to purchase through interactive, visually-driven experiences that empower shoppers to key in on the perfect product selection much quicker than with traditional e-commerce offerings like broad product catalogues and endless page scrolling.

The Challenge:

The world’s first AI-powered product determination platform, Skafos is faced with the task of handling constantly fluctuating data sets of product collections that can be in the millions. As it learns more within a store’s product portfolio and customer interaction samples, Skafos needed to find a way to index that data for quick returns in its real-time tools, but also save it for long term use.

The Solution:

Redis Enterprise Cloud was a clear and ideal solution to support Skafos’s microservices architecture now and for the projected scaling and growth in years to come. Skafos was able to leverage Redis Enterprise Cloud as a cache for capturing response data on retail websites, as a database to back up those records, and additionally as a complementary tool with RediSearch to support heavy document searching in real-time for consumers.

The Benefits:

Conversion rates for retailers have ranged from 30% to 165% in total uptick since deployment of the Skafos platform. With added requirements for availability and uptime redundancy as Skafos takes on a greater load of customers and product information sets, Redis Enterprise Cloud has given the Skafos team confidence that the deployments of their product are future-proofed and capable of growth for years to come.

“Our backend is basically subsetting the products using RediSearch in real-time on the fly. Plus, we capture that result, too, so then we end up using it to help itself go faster each time around.”
— Tyler Hutcherson, Senior Machine Learning Engineer, Skafos


Michael Prichard, CEO and Founder, Skafos
Michael Prichard,
CEO and Founder, Skafos

Michael Prichard had one intention in mind when he started Skafos.ai: bring a little more humanity into the online shopping experience. To do that, Prichard wanted to bring more of the conversation and interaction one might experience in a traditional setting, like with a shopkeeper at the local mall or in a big box retailer, into the world of e-commerce.

Traditionally in retail, an individual consumer had no choice but to stop by a local retail outlet if they wanted a new piece of clothing or an upgraded kitchen appliance. Often in these environments, sales associates work with the shopper to understand their desires and hopefully help them land on the product(s) that they would purchase.

Skafos wants to take buyers down that path again, but more genuinely than ever in an e-commerce environment.

Delivering the Products That Fit Best

Products That Fit Best

The flagship Skafos platform is a plug-and-play recommendation engine that brings intent-based suggestions to consumers through an interactive tool rather than a static data flow based on predetermined factors. In other words, the tool allows the consumer to have a say in what recommendations show up for them on a retailer’s e-commerce store rather than the long-established method of trying to predict their desires simply based on their original landing page without knowing anything about their intent.

Along with the conventional ability to filter between categories and other criteria, Skafos’ technology presents a consumer with a series of product images, to which they can give either a thumbs up (“like”) or thumbs down (“dislike”). With each selection, the tool, in real-time, flips tiles in the grid to show more complementary products based on their responses to weed out undesirable products and bring up more options that will get you closer to what they coveted from the start. 

“We begin to track the shopping journey as the buyer interacts with products on the merchant’s site,” explained Prichard, CEO and Founder of Skafos. “What we’re saying is, ‘Okay, we get an idea that you like a product, maybe with this shape, this color, et cetera, so let’s show you some different options that have those variables and see if we can get you closer to what you want.’”


Plenty More Than Just a Cache

Retailers have an incredibly high and ever changing number of product SKUs. For one of Skafos’s partners, Lights Online, that figure surpasses 22,000 in the system at any given moment. For others, it could be millions. Add in the deeper product information like naming, pricing, associated tags and descriptions, and more, and the Skafos platform needs to be able to comb through massive data sets for just one of their partners in order to deliver recommendations in real-time for a consumer that doesn’t have patience for a “Loading…” screen.

“Retailers are starting to embrace this idea that there’s this extremely messy component to the online shopping experience,” according to Luke MacFarlan, Chief Product Officer at Skafos. “We need to find better ways to optimize for the customer and provide them a more helpful way of interacting with lengthy store catalogues rather than just letting them loose on giant product lists in hopes that they find what they’re looking for.”

Skafos has been an open source Redis user since its origins, and as their tech stack evolved, transitioned to Redis Enterprise Cloud for three critical components:

  1. Skafos utilizes a microservices architecture, which applies Redis Enterprise Cloud as a key component of the system. Each individual service is designed to fit a particular use and does not need to run everything in the Skafos portfolio of offerings, however Redis Enterprise Cloud enables those dozens of services to communicate active data and pieces of the architecture in a way that can scale efficiently and make up for latency overhead brought on by the ongoing inter-service communications.

    “To be able to start scaffolding out microservice communication and all of these things within the architecture, Redis Enterprise Cloud made the most sense and was the tool that the technology leadership confidently decided was the best fit,” noted Tyler Hutcherson, Senior Machine Learning Engineer at Skafos.
  2. Instead of processing every consumer selection through ML or algorithms, Skafos caches all results through each of its recommendation engines for a period of time to help process data smoother and faster. The team also uses caching on Redis Enterprise Cloud with API activity that it connects through various e-commerce vendors.

    For example, if a Skafos retail partner has a product go on sale or a new collection is launched, it triggers a process. The record gets re-indexed by leveraging the store’s own API to fetch data periodically and at the alert of webhook events to make sure the attributes in the data can stay up to date.
  3. At the heart of Skafos is its own dynamic recommendation capabilities, and RediSearch helps to power that. Skafos stores all of a retailer’s products as a hash and then relies on RediSearch to index and ultimately search through that data in real-time to limit results to a given price range, or narrow down by certain tags or product identifiers, as a couple of use case instances.

The Guiding Light of the Buyer’s Journey

Skafos is increasingly proving out a value-add to retailers that shows all of the potential gains they can obtain by bringing on this new style of recommendation engine. With sights set on increased conversion rates, deeper buyer engagement, and higher revenue per visit, there have been early wins for the Skafos team in choosing Redis Enterprise Cloud as a critical component of their platform to deliver clear benefits to their retail partners.

Lights Online saw an increase in conversion of 165% in its first three weeks after implementing the Skafos technology. The interactive product discovery tool, in this case called “Fan Finder,” also helped Lights Online observe a 119% boost in user engagement and also tracked that after its initial rollout, products bought through the “Fan Finder” mechanism were 22% less likely to be returned as compared to traditional purchases through the online store.

“Our backend is basically subsetting the products using RediSearch in real-time on the fly,” noted Hutcherson. “Plus, we capture that result, too, so then we end up using it to help itself go faster each time around.”

Everything that the Skafos engine learns from each consumer interaction is pulled from a stored hash and is then indexed with RediSearch. With most of a retailer’s products staying static on the backend, there remains a great deal of data that ends up staying on Redis Enterprise Cloud as long term storage and not just as a real-time cache. Skafos additionally relies on Google Cloud as a provider and storage host to optimally host all of this data plus any other unstructured data they handle for their retail partners.


Bringing Back the Shopping Experience

The Skafos team is strategically continuing on a path favoring a microservices architecture to regularly iterate faster and be flexible with their customer’s demands. With more product features and scaling ambitions on the way, Skafos has successfully transitioned its systems completely off of a legacy platform and onto its new tech stack that heavily incorporates Redis Enterprise Cloud to support this growth.

While retailers and e-commerce outlets remain steadfast in their search for higher conversion rates, Skafos is working to figure out an ideal shopping funnel that can drive people to purchase events, but also allow customers to guide their own journeys again. With that ambition, Skafos requires the ability to better understand the insights and full picture of the online shopping industry.

“One recent idea in retail circles is that brick and mortar stores are for shopping and e-commerce is for buying. So we’re trying to bring shopping to e-commerce,” concluded MacFarlan. “We think a huge portion of that is a consumer’s own agency and being able to help themselves self-direct their own journeys through a store. That means taking some of the new types of information and data that we can generate and delivering that back to customers and retailers alike in a productive real-time way.”