Bill Bishop, Brick meets Click
You may not know Gary Saarenvirta, but if you are in retail today you should take the opportunity to benefit from his knowledge and passion. Gary is on a mission to bring the powerful information technology capabilities he learned in the aerospace industry and at IBM to retailing. The result is Daisy Intelligence Corp., a software-as-service provider that’s making it possible for grocery retailers (and other businesses) of all sizes to use artificial intelligence tools for optimized decision-making in order to increase profits.
As retailers and merchandisers, your business is selling, and the competition is fiercer than ever. In the future, growth will mostly likely stem from the ability to use your data in new ways, and the power of A.I. services like those offered by Daisy can improve your business results, despite what will prove to be a yet another challenging year ahead.
This post is part of our Spotlight series and is sponsored by Daisy Intelligence Corp.
BMC: Most retailers are familiar with predictive analytics, how is artificial intelligence different?
Gary Saarenvirta: When we discuss traditional predictive analytics – like regression and other forecasting tools – we are talking about solutions that look to the past for answers. This tends to evaluate only subsets of data to draw inferences about historical activity. The problem with this is approach is that it ignores the related actions and ripple effects. The “predictions” it produces are defined entirely on events that have occurred in the past. Any “learning” that occurs happens only when new experiments are conducted, i.e., you can learn only as fast as time.
Artificial Intelligence, on the other hand, creates a model of the retail environment that connects a retailer’s actions (promotions, prices, inventory, purchasing, real estate) to market results, taking into account the “ripple effects” caused by the actions. This allows A.I. to evaluate future outcomes, even if there is no historical precedent – in effect, A.I. can simulate the future. First, it uses 100 percent of the retailer’s historical data to learn what it can – it then takes the analysis to the next level by simulating a mix of previously known and new, untried actions to find the many different ways the future may unfold, choosing the optimal sequence of actions that achieve the best long-term outcome.
As a consequence of this ability to simulate, an A.I. system can learn years of retail in one day, limited only by computing power. A.I. directly answers the question, “What actions should I take to maximize long term revenue/profit?” The A.I. approach is comprehensive and a balance of historical and future-focused.
While A.I. isn’t new, the good news is that in recent years, these A.I. capabilities are now accessible to businesses of all sizes – not just big corporations and government entities – because the power of computing has increased dramatically at the same time the cost has come down.
BMC: So how does Daisy use A.I. produce/generate results for retailers?
GS: Promotion optimization is one of many areas where A.I. offers significant and immediate improvements to profitability. Daisy uses A.I. to help retailers:
Select the right products for promotion
Set the right price for promoted products
Forecast demand (sales)
Promotions are complicated. For example, when you put Coke on promotion, you can expect to sell more Coke, but then there are ripple effects. For example, while selling more Coke increases the sales of related products like salty snacks, it’s also likely to cannibalize Pepsi sales for that week, and probably next week’s Coke sales due to forward buying.
So, where do you stand in the end?
Is the increase in related (non-promoted) products large enough to cover the cost of the promotion?
How big is the drag from cannibalization and forward buying?
What’s the impact on other categories?
Daisy can use A.I. to simulate all of the ripple effects, so you can look at the whole picture and you’re able to gauge the long-term benefits and consequences, not just the immediate effect. In contrast, traditional analytics tend to be more narrowly focused and concentrated on the short term – it gives you the answer that sells the most of the promoted product for the next week.
We’ve used these capabilities to create a metric we call Net Promotional Effect or “NPE.” The NPE captures and accounts for all of those plusses and minuses, and then allows us to rank every SKU in the store based on its contribution to NPE.
BMC: What makes this NPE metric distinctive from other metrics retailer currently use?
GS:The NPE enables retailers to analyze results beyond the direct effect on weekly sales numbers and determine how all of the effects of the promotions are contributing (or not) to achieving the overall goals of the business. In effect, measuring NPE lets you look more broadly to see whether a promotion is taking you to a loss leader or an increase in basket size, profitability, sales, or margin.
This is all about improvement and efficiencies – A.I. makes it possible for a retailer to increase NPE with no additional investment in price or margins, and in some cases it actually allows you to increase NPE with less margin investment. It also makes the grinding, week-after-week process of promotion planning far more efficient, so more energy can be put into marketing and merchandising instead of trying to find the right data and pull useful information out of it.
BMC: What differentiates Daisy in the minds of the retailers you work with?
GS: Unlike a lot of analytics solutions, there’s no hardware to purchase, no software licensing agreement (and updates) to pay for, no complicated system integration, and no personnel training required. We take the retailer’s sales history and we do the work. We take the responsibility, instead of putting it back on the retailer.
The most important deliverable Daisy produces is a set of detailed recommendations that will grow the business’s profits and sales. In terms of what makes us different, the retailers we work with especially appreciate two things:
First, our business model. We’re a software service business – this means we do the heavy lifting, and if we don’t deliver, then you don’t keep paying us.
Second, the cost is reasonable – a fair charge based on the value we create.
BMC: What is your background? What motivated you to start Daisy Intelligence?
GS: Early in my career, I learned about computational fluid dynamics and massively parallel computing applied to massive volumes of data with millions of variables while studying for a Master’s degree in Aerospace. I was exposed to sophisticated machine learning techniques like neural networks, classification trees, and association rule mining at IBM.
When I moved over to business, I was shocked to see how little use was being made of the math and science I was familiar with. I saw a gap. Businesses – especially in environments that are hugely complex and changing all the time like retail – needed better tools to solve their problems. I started Daisy to help everyone realize the promise of information technology to make the world more efficient, lower the cost of living for everyone, and along the way make businesses more profitable.
People like to joke that something “isn’t rocket science,” but when it comes to meeting challenges like evaluating the true ROI of promotion choices, it is actually a lot like rocket science. The thing is, now retailers of all sizes can benefit (and profit) from applying the same kind of analysis methods that rocket scientists use.
BMC: Customer-centricity is a big focus for retailers these days. How can Daisy help them?
GS: My view on this is a little bit on the maverick side. I believe customer-centricity actually starts with products. Transaction data is actually the most direct way customers “talk” to retailers. The reason a customer goes to the store starts with the product, and the key is giving them the products they want to buy.
It’s the interplay between these two – products and customers – that’s the key to customer-centricity. It’s got to be both, and Daisy’s A.I. capabilities marry these two essential elements. That transaction data is what tells the retailer a) what products customers want to buy, and b) the patterns in which they prefer to purchase them.
BMC: Daisy’s impact is particularly powerful in fresh and perishables. Why do you place so much emphasis on these departments?
GS: Perishables are important, mainly because they trigger the weekly visits to the store – typically much more so than non-perishable products. For the customer, produce and meat are usually the core reasons for the trip, so it makes sense to build a promotional plan around those categories.
Perishables also give retailers more freedom to do what’s right for the customer, so to speak,because they aren’t affected as much by supplier payments that are earned if a retailer promotes in a certain way, even if it actually runs counter to the retailer’s self-interest. We help retailers achieve better results with the right mix of promotions – with or without supplier payments – and perishables are a key part of this.
BMC: Aside from improving promotional effectiveness, what other challenges do you see A.I. helping grocers to overcome in the next year or two?
GS: We’re working on developing A.I. solutions to improve profitability in three additional areas:
Tailoring the assortment: Identifying which items to offer and how much space to allocate to each category and department.
Improving the merchandising layout by improving the positioning of products in the traffic flow, determining the best distance between complementary products, and/or where key categories should be in the traffic pattern.
Evaluating real estate site selection.
BMC: If you are retailer or merchandiser, growth can come in 2018 from using your data in new ways and the power of A.I. services like those Daisy offers can help you improve business results