The retail industry is an environment of Big Data just waiting to be exploited. It’s all there, from data related to real-world store sales and traffic, to customer loyalty engagement and social media sentiment, to supplier transaction information and even sensor findings from devices that monitor things like product temperatures. But one problem is that the Prescriptive Analytics informed by this data usually manifests in reports whose recommendations aren’t always immediately obvious or easily translatable to the staffers best positioned to act on them, nor are they generally closely followed up on.
“People are busy and they have to focus on their day-to-day jobs, so actions may be communicated to them by email but then it stops there,” says Guy Yehiav, CEO at Prescriptive Analytics vendor Profitect. “It’s never communicated back what happened.”
Yehiav says the company, whose roots are as a consulting firm for the retail sector, heard its clients voice such complaints on a regular basis. That led to its interest in the idea of delivering Prescriptive Analytics in a more streamlined way. “Instead of sending people dumb reports, we wanted to be able to tell them what’s inside the report and what to do about it in simple, plain language,” he says. Retail managers, after all, are hired because they know how to manage staff, not for how well they read reports – yet they’re drowned with them, he says.
Also key was that customers be able to implement the technology quickly and scale it easily.
Profitect’s Take on Prescriptive Analytics
The company’s Cloud-based solution relies on three pillars in its aim to accomplish its ends. Both structured and unstructured data are ingested quickly into its Cloud from across the retailer’s supply chain and other systems, in a matter of days, he says. Profitect solutions are embedded in eight different modules that segment the data to fit retailers’ various Prescriptive Analytics needs, such as Planning and Buying, where forecasts, allocations, and raw materials data are critical, or Inventory, where on-shelf availability, shrink, waste, and movement are high priority data.
The second pillar, he explains, revolves around using pattern recognition and Machine Learning algorithms to find the “golden nuggets” of information that identify where the business needs to improve. “The idea is that most businesses are going the right way in most respects, but let’s identify how to improve their averages by finding anomalies and feeding them into the averages so that businesses improve on a continuous basis,” he says. The process happens automatically, rather than having analysts create reports based upon the data and interpret behavior patterns and recommendations from it.
That brings things to the third pillar: Actionability. No more sifting through reports to find what to do, but rather prescribing and sending out actions to users in clear, descriptive language as part of sophisticated workflows – and making it possible for those users to easily communicate feedback on actions taken, as well.
For example, one big box retailer, according to the company, used the Profitect Omni-channel module to help with the unrecognized net margin erosion it was experiencing due to having higher costs to serve customers who returned products at a high rate and then received re-deliveries at the company’s cost. By implementing the module, the retailers could determine when it wasn’t profitable for it to bear the cost of return. The module automatically sent alerts to the retailer’s Order Management System recommending profitable shipping arrangements and pay alternatives for return options for these customers in real time.
Retail-Related Industries and Beyond
Most recently Profitect signed on as a customer supermarket giant King Kullen Grocery Co. Inc. The company, which has 35 Long Island King Kullen stores and five Wild by Nature natural food stores, in June selected Profitect’s Sales module to initially focus on inventory shrink, fraud-related activities, and point-of sale (POS) efficiency. The firm’s SVP of Finance & Administration, Jim Flynn, is quoted in a press release about the news as noting that one reason King Kullen selected Profitect was “for its superior ability to provide actionable information.”
The retail vertical has been a strong place to start, but other industries that connect with it, such as consumer packaged goods (CPG) and distribution channels, also have similar challenges of making analytics actionable and so can leverage its model of Prescriptive Analytics to their benefit, as well. Yehiav cites as an example how a beer brand distributor could benefit by using its technology. Say the distributor has 90,000 points of consumption in the area it serves, from bars to restaurants to mom-and-pop shops to large supermarkets. Perhaps a few hundred drivers deliver product to these sites daily or a couple of times a week. Sending these drivers reports with analysis of consumption patterns of the customers they supply probably isn’t going to help anyone; each driver’s job is to be a good representative to the customer, not a Data Scientist.
Using Profitect and its unsupervised Machine Learning cluster analysis algorithms for on-shelf availability, however, the company could discover patterns in how its clients in different clusters order and buy beer. It may come to the conclusion, for instance, that there is an estimated 97% probability of a particular small grocer being out of stock of the brand’s light beer. Once Profitect has identified the probability of it being out of stock, it will deliver to the driver who services that location “a descriptive insight stating that Store X has very light on-shelf availability of this beer and they didn’t order more from you,” Yehiav explains.
It then will deliver a clear action for the driver to take, such as asking the owner if he or she needs to order more, and alerting the party that the item is on the truck now to provide the service if needed. When the owner realizes the mistake and accepts the driver’s offer to provide a case, the driver can provide feedback to the system in the simple form, say, of a thumbs-up, to indicate that an additional case was sold.
Such information crowd-sourced, as it were, from multiple drivers can be analyzed so that the company learns more about what actually works in the field. It may realize, for example, that there is a high probability that sales will increase just by having drivers speak with owners about on-shelf availability. Measuring outcomes like these can support constant improvement of practices and growth in revenue and profits.
Other action items can be included as well, such as counting how many cases of that brand of beer are on the shelf compared to a competitor’s, or taking a picture of the shelf to upload to the system. The latter, he says, points to “the cross-organizational nature of retail and CPG.” The picture can go to brand marketing showing that a competitor has added a cooler to its display for greater attention, or to legal showing that the shelf height where the brand is stocked isn’t at customer eye-level, as per the contract, so that compliance issue can be addressed.
Profitect’s technology can stretch beyond retail-affiliated industries, though. For instance, it’s currently engaged in a pilot revolving around production scheduling for a chemical company; the company is using the system to collect and analyze reactor data from internal sensors that record system temperatures, process times, and more to find root causes of quality issues. To that end, it’s using the same pattern and Machine Learning algorithms found in Profitect’s sales and buying module. “The Internet of Things will help us get into new verticals,” Yehiav says. “It’s great for companies but it creates so much more data and they can’t analyze it in reports anymore.”
Companies will always have a need for analysts to do the work of creating high-level reports. But Profitect will be valuable when it comes to creating the on-the-ground reports in the hope of driving actions. “We empower reports with actions,” he says.
Businesses themselves can take the lead in empowering actions using its technology toolkit and platform for slicing and dicing data differently, leveraging its Machine Learning algorithms for purposes beyond what it currently builds in. “They can do a lot of things with the platform by themselves and we share what we see among other customers also,” Yehiav says. “It’s a continuous improvement platform and we learn from customers and they learn from each other.”