Search for a report

News

Explore the Technology’s disruptive expansion amid COVID-19 in our conversation with Analytics2GO

interview with Analytics2Go on the Technologys disruptive expansion amid COVID-19

Analytics2Go (A2Go) was established in 2017, with the aim to ease the transition for companies into the next generation of data-driven business operations. They are a motivated group of seasoned data scientists, management consultants, and engineers that address many difficulties with data and AI deployment and adoption.


Inkwood: 
Recently, L’Oréal Canada has used artificial intelligence and augmented reality for consumer engagement during the COVID-19, with its live shopping video efforts. Can you tell us about such interesting/disruptive use cases of your AI platform/services by some of your clients during the pandemic?

Analytics2GoA2Go focuses on using AI to optimize and customize both pricing and purchase recommendations in a number of different industries and verticals. Both pricing and purchase recommendations affect the customer’s experience, whether it is B2C or B2B, and both directly affect revenue, especially in the ‘new normal’ where much of business is being done online.  Below are three use cases that we believe are disruptive to traditional pricing and purchase recommendation methodologies increasing revenue even during the pandemic.

Dynamic Pricing for Wholesale Food Distributors:

FoodCo, a $500 million independent food distributor, faced a challenging environment amid the COVID-19 pandemic. Sales plummeted initially as their customers—primarily bars, restaurants, and convenience stores were hit by closures and customers shifted to stay-at-home lifestyles, including more online purchases. After about three months, revenues had recovered significantly to about 80% of pre-COVID-19 levels, but the positive trend seemed to fizzle out at that point. FoodCo’s executive team needed new ideas to improve revenue and profits as soon as possible.

One crucial strategic alternative they considered was to improve pricing strategies and techniques. FoodCo had attempted to re-engineer pricing in the past with the help of pricing experts and pricing software but had been dissatisfied by a lack of tangible results. They recognized that their operating context was sufficiently complex to make a systematic approach to pricing quite hard to design and implement.

Altogether, FoodCo served more than 50,000 customers and 2,000 sales reps across a relatively broad and diverse regional area with significant differences in population density, consumer spending, and economic performance. On average, its sales reps responded between 25,000 and 40,000 requests for quotations every week. The process was semi-automated using industry-standard approaches to establish list prices by segments, relying on sales incentives and the expertise of the sales force to optimize quoted prices.

Sales reps were incented to minimize discounts but were prone to offer maximum discounts to avoid losing orders or customers. There was a small, internal pricing group that attempted to optimize the pricing matrix and act on deeper discounts for big buys. Still, the administrative burden left very little time for analysis or planning, and everyone was convinced that money was being left on the table.

A2Go was asked to assist FoodCo in applying an AI-based approach to price optimization. Working with FoodCo’s executives and subject-matter experts, A2Go configured a two-engine, machine-learning solution that succeeded in optimizing pricing and increasing the level of automation in the end-to-end process. With the new solution, source costs for center-of-plate products like salmon were predicted more accurately so that the base prices offered more closely aligned to actual landed costs at the time of sale. Besides, AI was also used to classify each of FoodCo‘s 50,000 customers into one of approximately 1,000 clusters of customers and provide optimized base price lists for each of the clusters.

The solution had an immediate impact. Even before sales improvements could be identified, the sales reps observed how easy it was to process quotations. From their perspective, each customer had its own base price list. Customer pricing was provided in a familiar format as a base price from which sales reps could base their discounts. One advantage of improving an existing process with AI is that the tangible effect of better insights and better pricing recommendations is felt immediately and is measurable. Based on A/B testing, FoodCo increased both revenues and margins by more than 3%, even in the middle of a pandemic.

FoodCo is now extending price optimization throughout all product categories so that AI pricing techniques can increase revenue across all customer segments. Extension of the solution across all segments, including higher gross-margin non-perishables, is expected to yield even higher revenue and profit gains. Now that pricing has been AI-enabled, FoodCo is investigating the possibility of applying AI to demand prediction so that purchase quantities and timing can be more closely synchronized to actual demand patterns, reducing both stock-out and spoilage.

On-line pricing for eCommerce Retailers- (xAI):

The pandemic has accelerated the inevitable shift to online purchasing.  Amazon, for example, saw an 80% increase in Q2 2020 sales. As more and more companies emphasize digital channels for selling their products, customers experience the convenience of shopping from home as well as low prices that are guaranteed by the transparency of pricing across the internet.  It is easy for consumers to check prices between websites, and increasingly, there are “aggregation websites” that provide instant price comparisons and links back to the website chosen by the customer.

It has never been more important for individual e-commerce retailers to maintain up-to-date pricing at all times to maximize margins, and also never been harder to do. One of our customers is a medium-sized furniture retailer with approximately 240,000 different furniture and home goods products available on their website at all times. Nearly all of the products are the same as or very similar to other products offered online by their competitors. These competitors are free to change prices once/week or once/day or even many times per day, if they want. Under these circumstances, it is very difficult to maintain optimal price levels manually or with the help of traditional pricing software that does not consider external factors that often dramatically affect customers’ purchasing behavior and preferences.

Here’s why. The sheer number of items makes it impossible to price manually and complicated to apply automation to pricing that considers more than just competitors’ prices. Manually, a category manager cannot respond quickly and optimally to changes in competitor prices. If automation software is applied, the first go-to approach is to establish pricing rules that can be applied automatically given specific ‘If X, then Y’ conditions. However, these rules have to be continually updated as consumer behavior changes and unexpected external factors arise.  Finally, competitors are always working to “outprice” your company, so any ‘rules’ used must be sophisticated enough that they cannot easily be deciphered.

Conventional AI, using historical sales data and competitor prices, overcomes some of the problems of automation, but category managers for online retailers have been reluctant to adopt these AI solutions, because the price recommendations are usually made without any transparency provided to the logic on individual recommendations. They are reluctant to make price changes when they don’t understand the rationale behind the recommendations that are made. This problem, sometimes referred to as “black-box” solutions, has been a barrier to the acceptance of AI across industries for all kinds of solutions. In fact, even where the AI can be demonstrated to be right a high percentage of the time, individual category managers worry about the big mistakes that can occur and, therefore, resist implementation.

A2Go’s approach to optimizing prices for online retail overcomes these problems with the provision of “Explainable AI” (xAI). Our xAI solution is called Price-Right AI for e-commerce. A2Go uses a design platform from a company called Rulex that uses AI to define optimal pricing decision rules for online retailers. These decision rules take advantage of the power of AI to uncover business rules that cannot be detected using excel spreadsheets or intuition. For our online furniture retail customer, AI determined decision rules were differentiated broadly across all 240,000 SKUs and, when necessary, rules are determined for subcategories of SKUs that followed their own rules. The decision rules reflected the level of current competitor pricing as well as the company’s history selling against competitive products. This is important, because of the right price level vis-à-vis a competitive product is not always the same by product or over time and also depends on the relative inventory position for that product.

Category managers like using Price-Right AI for a couple of important reasons. First of all, because the decision rules (and not the AI directly) are used to recommend optimal prices. Therefore, the category managers can review the rationale behind the pricing recommendations. Normally, the approach illuminates the decision rules when a cursor is placed over the proposed price and clicked. The category manager can see a serious of if-then statements that were used to define the recommended price. This feature overcomes most resistance to adoption.

In addition, the conversion of AI’s insights into decision rules facilitates the more frequent and more flexible use of the decision rules per SKU. The AI used to determine the decision rules for any given online retailer, in general, does not need to be run more often than once every 24 hours. The frequency of using AI to recalibrate rules is dependent on the size and complexity of the business. In fact, for many businesses, once per week has proven to be a good frequency. Once a set of decision rules are determined, the prices can be run many times per day, because the computing time and resources to run the decision rules are not significant when compared to the computing time and resources required to run the underlying AI algorithms. Sometimes the AI solution needs to run for 12 hours or more. The decision rules, by contrast, can be run for one product instantly or all products in a matter of minutes, not hours.

Price-Right for e-commerce is the perfect combination of AI-enabled insights, recommendation transparency, and ease of use to make it an easy decision for retailers who want to optimize online pricing while retaining control and insight into the pricing process.

Product Recommendation for Quick-service Restaurants (QSRs):

The pandemic has significantly reduced foot traffic at stores and reduced the number of in-person visits and purchase transactions. Many restaurants, including QSRs, have observed long-lasting shift to digital sales channels for both pick-up and delivery services. QSRs now have to improve customer experience during the ordering and delivery process to retain customers and improve profits.

Problem: QSRs do not have a way to personalize each transaction (ticket) either for in-person or digital channel orders. Customers in all channels have come to expect personalization of their shopping experiences. They expect brands to recommend products and experiences that they would be interested in and not ‘one-size-fits-all’ offers. For in-person orders, QSR operators have struggled to get cashiers to process customer orders quickly, accurately, and recommend a highly appropriate purchase recommendation to add to the order. For digital orders, QSRs are relatively inexperienced and have not yet developed ways to personalize product recommendations in real-time for each order.

Solution: We know that personalizing the ordering process requires information about not only the order but also about the context in which the order is being made. For example, a customer in Atlanta, GA, who orders from a QSR on a 95-degree day at 1p on July 4th, may be more inclined to add ice cream to their order than fries given those circumstances. Our AI-enabled purchase recommendation solution (xSell) considers real-time data streams that provide the needed context. The calendar, weather, local events, and store location of each order are provided in our solution, and are used in combination with the Point of Sale data to calculate a purchase recommendation ‘that particular’ customer is most likely to accept, thereby increasing their ticket total. Our solution does not require any personal information from the customer. However, if personal information is available through a loyalty program, for example, the purchase recommendation calculation can be much more accurate.

For in-person orders (counter and drive-thru), xSell suggests a product to the salesperson on the screen before the sale is finalized. This improves the typical approach of offering the same item to each customer. For all digital channels (kiosk, online, store specific app), our AI-enabled xSell delivers a recommendation automatically on orders before payment is made. Our QSR customers who have deployed our AI-enabled purchase recommendation solution see an average sales increase of over 5% and up to 15% improvement in profits.


Inkwood: 
The adoption of AI in healthcare is gaining more traction nowadays. How do your services disrupt the market?

Analytics2GoAnalytics2Go delivers AI products to specific business process pain points in particular verticals. Our solutions are delivered as a service from the cloud and designed to integrate, deploy, and operate invisibly within the existing infrastructure. Each solution consists of reusable components that are configured to fit a customer’s needs and tech stack. The value of this approach is manifold, especially in the current economic environment. Delivery from the cloud seamlessly delivers a solution without the costs and time required to build out a data science team, vet out dozens of AI tools to handle data, algorithms, visualization tools, and more. It also allows all Analytics2Go customers to run their AI solutions on our platform and ‘pay as they go,’ which effectively shares the cost of the infrastructure and services among all customers. We address the many requirements such as data storage, compute time, customization of dashboards, etc., which otherwise all come with individual price tags and time for RFQs from multiple vendors to pull together a complete solution. Our customers can operate one AI solution or many at the same platform that provides all the data, security, compliance, integration capabilities, speed, and dashboards required. We provide tools needed to go from data to answers with the bonus of having decades of domain expertise in the areas where we focus. Business domain expertise is often an overlooked ingredient to successful AI solutions, and is important to assure your solutions are exactly what you need to succeed. Our approach and our specific solutions and services disrupt the way AI solutions have traditionally been delivered. Instead of a DIY model, we use an AI-as-a-Service model.

One observation we have made since the massive shutdowns due to COVID-19 is that companies have been willing to purchase AI-based solutions that address very targeted problems within their business if they aim to increase revenue. All other solution types are being put on the back burner. Off the table for all but the biggest of companies are company-wide AI digital transformation initiatives that require long-term change management and continual investment to maintain.

There is a need for AI-driven solutions to help companies on tight budgets to compete and remain relevant in a world where many businesses are being acquired by the prominent players or going out of business due to plummeting revenue from the COVID-19 fallout. Our targeted AI pricing and purchase recommendation solutions address immediate needs in these companies.


Inkwood: 
Besides healthcare, which other industries, according to you, can prove as lucrative investment destinations and why?

Analytics2GoAI adoption in nearly all industries and verticals will prove to be lucrative over time. Lucrative financially but also in the wealth of benefits it can bring in the form of predictions and prescribed actions for our future.

From retail to environmental conservation, data is at the core of understanding the true ‘drivers’ of markets, policies, and human behaviors. The use of data to uncover otherwise inaccessible insights remains relatively new to most industries, with the financial and retail/marketing industries being the most advanced in the use of AI in daily operations. Another area that is growing rapidly is the warehouse and manufacturing robotics. These sectors are driven by a need to improve efficiency in processes and reducing workforce-related expenses in order to compete with the Amazons and Apples of the world.

One of the main barriers to the acceptance and adoption of AI is the reluctance of workforces to trust what AI is telling them. This is particularly true in the healthcare industry for good reasons but is also true in many business environments, such as supply chain logistics, pricing strategy, and other areas that have traditionally been driven by experience and intuition. AI solutions that can be ‘explained’ are often required in many industries, so an AI solution that provides that feature should be considered a lucrative investment. Our dynamic pricing solutions are explainable to pricing teams and therefore adopted readily by customers.

To identify industries where AI will have the most impact, it is wise to understand where AI can be applied and beneficial. There are many problems in healthcare, business, or the environment that cannot be solved by grabbing data and running it through algorithms. Identifying a viable AI use case is the most reliable indicator of AI success in any given industry. There has to be a problem to solve and tools to solve it. In our experience, working with third party AI service providers with the expertise in the application of AI to a specific industry is the straightest line to success and, therefore, to a lucrative investment.


Inkwood: 
Lack of a skilled workforce has been a prominent restraining force for the AI market. What steps must be taken to tackle such challenges?

Analytics2GoThere is no doubt that there is a shortage of skilled data scientists, engineers, and programmers available to solve problems in all the areas where AI can benefit healthcare, business, and other industries. To compound that shortage, there are additional realities that narrow the pool of the available skilled AI workforce:

  • The reality is that all data scientists and engineers are not created equal. We are talking about specialization. Similar to scientists in the area of biomedical research, specialization is important. A trained Ph. D. in the area of immunology cannot transition easily to productive research in neurobiology. The experimental tools are similar between the two areas, but the domain knowledge is quite different. It is not impossible but transitioning comes with time and cost considerations. Similarly, data scientists train in specific areas. They all use math/algorithms and data as their tools to experiment and solve problems, but they also require domain knowledge. Training, tools, and domain knowledge are the ingredients of any successful AI initiative.
  • Another reality is that often the most highly sought after data scientists, engineers, and programmers prefer to work for the technology giants of the world. Who can blame them? Salaries are good, co-workers are highly qualified, and they will be developing well-funded, cutting edge AI technology. This makes it difficult for other industries to hire from the same pool of skilled workers.

So what is to be done? We believe there is motion now in higher education showing a shift toward more people entering technology related degree programs, which will result in larger numbers of qualified, albeit mostly inexperienced, data scientists, engineers, etc. Even better, there are also many programs being developed that bypass the traditional college path and instead are technology-focused degrees. In fact, Google is developing certification programs that will build up the pool of qualified, skilled AI workers in the U.S and elsewhere. Many of the certifications offered by Google require a certain level of prior work experience and internship work to complete, which should result in skilled individuals with domain knowledge entering the workforce at a steady pace in the near future.

We hope that more tech certification programs are developed in parallel with Google’s. Also, many academic institutions are now offering data science degrees online, making it possible for ‘re-skilling’ the workforce from traditional, obsolete jobs to digital jobs of the future.


Inkwood: 
What are your views on AI, Machine, Learning, Blockchain, and cognitive computing? How will these technologies span out over the next few years?

Analytics2GoWe see AI, machine learning, blockchain, and cognitive computing as one continuum. They are all part of the future of a data-driven world. They are all part of the skills requirements for a future workforce, and they are all valuable components of digital healthcare, business, education, and much more. The short-term difficulties with all of these components will be for individual organizations, big and small, to pull them together into a single comprehensive AI operating system that improves processes and outcomes. We believe that in a few years, AI providers will not be talking about these individual components as just technologies but what they, as a whole, can automate, solve and improve for every industry, government, or public sector.


Inkwood: 
How have you adapted your product/service offerings during the COVID-19 crisis, and what is your strategy for the post-COVID-19 period?

Analytics2GoWe operate in multiple countries; therefore, when the pandemic hit and supply chains and businesses were shut down, our business was impacted like most others. Companies stopped buying ‘non-essential’ products and services, which meant any AI initiatives and services geared toward longer-term improvements were postponed. Before the pandemic, Analytics2go provided ‘assemble to order,’ and custom AI solutions for demand prediction, price optimization, and purchase recommendations. We focused on CPG, QSR, Online Retail, and Supply Chain Management. Since our approach to delivering AI solutions was somewhat more time-intensive, Analytics2Go decided to shift to provide AI-driven business products that could be integrated and deployed rapidly. We took our most mature AI components and developed them into three products that can be rapidly integrated and deployed within a customer’s ecosystem.

Our transition to providing targeted AI solutions in the ‘product’ form has been a blessing in disguise. Customers have been extremely open to learning and purchasing AI solutions targeted to specific business processes in specific verticals that are easy to use, cloud-based, and provide AI as a service. Our Price-Right solutions for wholesale distributors and e-commerce and our QSR solution suite (purchase recommendation, promotion optimization, and more) are geared toward a rapid impact on revenue and profits. Revenue and profit are the top concerns for most businesses in the current situation. They will remain even in post-COVID times as businesses will be in recovery mode for some time.

In addition, we have built our solutions to be integrated into existing systems, which allows customers to bypass any infrastructure build-out or hiring requirements and the costs associated with them and go straight to having a functional solution. Our unintrusive AI solutions are adopted readily by the end-users since they are built to be as invisible to the end user as possible. This approach is similar to companies like Amazon or Google, where their workforce uses AI-driven technology daily to move inventory around and price products – but often have no awareness that it is even there.

Our post-COVID-19 outlook is good. We will continue developing AI products targeted to specific business processes and are invisible yet explainable, when needed, to the end-users. Our products will be targeted to mid-sized companies who will benefit from our data science, our domain expertise that goes into our products, and our delivery of AI as a Service. We will focus on mid-sized companies simply because these companies compete for a skilled data science workforce and affordable AI solutions.