Rubikloud Delivers on the Promise of Omnichannel

By Hessie Jones   |  May 26, 2018
Hessie Jones is CogWord's Reporter from Toronto, and is a CoFounder of Salsa AI

It was about a decade ago that this idea of Omnichannel began to take hold. Traditional retailers saw it as a "concept" as their predominantly bricks and mortars presence continued to be the focus channel to serve their customers. Fast forward to 2018 and Machine Learning is bringing this concept into fruition. By definition,

“Omnichannel promises to bring retailers closer to their customers by meeting heightened expectations of shopping immediacy, convenience, and ubiquity. It also recognizes an inevitable reality: shoppers increasingly demand a retail experience that transcends the barriers between shopping online and in store.”

I had the pleasure of meeting with CEO Kerry Liu and Manager of Data Science, Kanchana Padmanabhan of Rubikloud, a Toronto-based software company that has come the closest to delivering on this promise today.

“The World’s Leading Machine Learning Platform for Retail”

This is the company’s value proposition. Rubikloud is an advanced machine learning platform for retail that delivers in-market recommendations for promotions and loyalty marketing. The impact: optimized personalization that delivers the right product with the right message to the right customer, on the right channel, at the right time.

Customer Centricity

Now with 100 employees, with over a dozen retail clients worldwide, and a VC raise of $45MM, Rubikloud is on its way to revolutionizing retail.

The Beginning: Convergence that Changed Enterprise Software Forever

Five years earlier, Kerry exited a start-up following an acquisition. At that time, he noticed 3 significant events which helped form the impetus of Rubikloud:

  • CLOUD: Big 3 - Amazon, Google and Microsoft were making massive investments in Elastic computing, providing cloud access and enabling computing scalability for developers.
  • DATA FRAMES: The rapid deployment as well as adoption of open source and various data frameworks.
  • AI: AI was starting to become more pervasive in the consumer space powering things like Facebook’s newsfeed, Netflix’s recommendation engine and Amazon’s dynamic pricing.

These 3 trends were powering internal AI systems and they were converging into a new form of enterprise software. A new company was formed on this very thesis:

We founded Rubikloud to address these 3 challenges with one vertical initially, and as we expanded, move into new verticals. We decided to focus on the world retail industry, i.e., traditional retailers with diverse end user experiences in critical parts of the world where they have hit market maturity.

Kerry’s belief was that traditional retailers were not dead. They just needed technology to help them transform and thrive.

Reconciling the Customer Journey Offline to Online

Measuring the impact of a customer purchase to determine propensities has been difficult because up until now, the mechanisms to focus on one customer at a time and his/her purchases at each purchase point online or offline have largely been fragmented because of disparate systems and databases.

Rubikloud’s retail customers have been collecting this data for a very long time and in many cases, they’ve been stuck in multiple legacy systems and nothing has been done with them.

Reconciling insights from the data and matching it to the goals of the business cannot be done overnight. However, by understanding the goals and the challenges of each business stakeholder, i.e., the Merchandising and Marketing teams, Rubikloud has fundamentally enabled these teams to be more precise in their forecasts, more seamless in their executions, and has created stronger systems for performance.

As Kerry notes, the merchandising team needs to know how much inventory to order under hundreds of thousands of SKUs, at what price to set, and how much to allocate for each of the channels and stores.

Rubkiloud’s Price and Promotion Manager will give the retailer store level allocations, online level allocations. It will also tell the retailer what promotional tactic to use for that inventory and how often they need to replenish the inventory in the middle of a promotion as well.

Promotion Manager

All those variables are computed by the system in an automated way and sent by the system to various touch points along the way whether it be the supply chain system, store operations or ERP… It is designed to solve one of the most computationally difficult problems retailers face today: how much inventory, at what price point, what specific SKUs are moving, in what region, and most importantly, what customer segments are impacted by those SKUs, by that price point and inventory volume.

The value of these insights is destined to attract the larger CPG players, who have been traditionally unable to control or access the investment impact of their marketing or promotional efforts. Exposure to retail data has largely been elusive in the past. Inferred data sources like Nielsen provide CPGs information on price, market share, sell through. Rubikloud, in comparison captures hundreds of data points with respect to SKUs as they relate to customer segment, promotion tactic, purchase, price and location over time.

The State of Retail Data

Data is the new currency. It is and will be the driver to make decisions in-market, in near-real time. However, given the general challenges in transforming data today for analysis (some industry estimates claim 80% of the model development efforts are in the cleaning phase) coupled with absence of usage, aggregation and analysis of the data among large retailers, data transformation can take between 1-6 months. Kerry noted this is highly dependent not only on the state of the data and their systems, but also on the requirements dictated by the retail objectives.

According to Kanchana Padmanabhan, Data Scientist, good data is key. Rubikloud ingests all the client data – transactional, promotional, and other related data sources. There are differences among retailers that can lead to scalability challenges: like the different schema, legacy systems, different technology stacks and data volume differences. Onboarding the data, and especially validating them becomes critically important.

Over time, across many use cases and clients, Rubikloud has gotten better at data handling, managing and processing. They have learned and have become adept at asking the right questions up front before the data enters their systems. Kanchana adds:

We also have a stable product so we know what model we’re building. We know the solutions we’re offering, and more importantly, we know what data we need.

Kanchana notes they can process a variety of formats: from surveys, to advertising, email, transaction, CRM, ERP and social media etc. In some of these cases, marketers are encouraged to data enter the information consistently and correctly in order to ensure data usefulness.

In addition to primary information, Kanchana states they also overlay demographics, competitive pricing information, product images and descriptions. While social media has value, today that data is messy and trying to map transactions from social media is difficult… which leads us to the elephant in the room…

Can Customer Centricity and Privacy Co-Exist?

Considering the European General Data Protection Regulation (GDPR), are smart systems like Rubikloud threatened? Tackling the holy grail of personalization through the Lifestyle Manager will require increasing context to determine the customer’s intent to buy, in an ideal environment and circumstance. However, Kerry emphasizes,

We don’t get personally identifiable information (PII): We don’t get your name, your email, credit card number and phone number. The reason: it doesn’t help us. Knowing PII doesn’t help our model at all. We prefer clients to give us customers in a hashed ID...We are going down to the 1-1 customer level but we don’t know the person’s name or PII.

This hashed ID will be sent back to the marketing automation stack via Adobe Campaign, or Salesforce Marketing Cloud and will be matched to customer information within the client systems, outside of the purview of Rubikloud and its systems.

For many marketers, the use of any and all available data for analysis and segmentation has been par for the course. With AI, the search for patterns among thousands/millions of data attributes across the entire customer base, does not require, as per Kerry, the need for personal information to build an effective model.

While PII today is a strong pillar within the privacy debate today, the impending inferences from the metadata where AI finds those patterns to define context and predicts customer propensities will surely be a topic for debate. More on this as GDPR takes flight.

The Future of Retail Means Getting Back to Basics

Kerry believes retail is growing as a whole. Customer expectations have changed significantly. Measuring a retailer’s technical competence is also an earmark of a brand’s evolution. Customers want relevant recommendations; they like Amazon’s dynamic pricing; and they expect the experience to be a little more familiar every single time they interact with the retailer.

We think retailers aren’t dying but they have a very short window to get technology correct. They can’t waste time on big legacy projects or makeshift innovation projects anymore. They have to address either the consumer problems or the product problems.

Kerry noted that retailers are overcomplicating the problem. The challenge is how they deploy solutions in a productized way. We are living in a time where AI experimentation is rampant. Rubikloud as moved into full production and they have learned how to build a holistic solution that can work for the industry.

You buy software that has packaged all the use cases and inputs and you work with the company that knows the nuances of your business. In our view, you buy software to address CORE problems that can be solved.

This means that the retailers will go back to being retailers. For the buyers, they will remove the financial analysis from their job descriptions. Instead, they will apply their core competencies in price negotiations, resourcing product, and experimenting with new products and new customers.

By adopting technologies it allows clients to do what they do best. In Rubikloud’s case, it’s definitely influencing this mantra.

Kerry Liu is the Co-Founder & CEO at Rubikloud, where he leads three important functions: people, sales, and technology disruption. In his role, Kerry works to manage and maintain a thriving company culture that recruits the best and brightest in the industry, while also maintaining relationships with global retailers and investors. Kerry is passionate about machine learning and big data, and enjoys providing enterprise retailers with the tools and knowledge needed to enhance their overall business goals.
Kanchana Padmanabhan is the Manager of Data Science at Rubikloud. Kanchana received her PhD from North Carolina State University in Raleigh, NC with a focus on Graph Mining and Analytics, Computational Biology, and Network Modeling. Prior to Rubikloud, Kanchana spent 3 years at Sysomos, a social media analytics company for research projects that use machine learning, network analysis and NLP techniques in conjunction with big data technologies such as Hadoop, MapReduce, and Spark. She currently  focuses her time on solving big data problems in retail. She has 4 patent filings and co-edited a book titled "Practical Graph Mining with R“ CRC Press.