shelf-level collaboration Archives

At the CSCMP Annual Conference a couple of weeks ago, I attended the best supply chain presentation I’ve been to in years. It was given by Ron Volpe, Customer VP – Supply Chain at Kraft Foods, and Jon Golovin, CEO of Retail Solutions Inc. Kraft is using solutions from Retail Solutions to better leverage its downstream data. The company has a truly audacious goal for inventory reductions based on using this data.

I’d like to share some comments made by Jon about the maturation of shelf-level collaboration, with some supporting examples based on Kraft Foods’ experiences.

A few years ago, not all retailers were willing to share downstream data. Some actually charged manufacturers for it! Today, according to Jon, every major retailer in North America, except for Publix, is willing to share downstream data, or is at least piloting the value of sharing it. Nevertheless, the quality of the data, and the amount of data categories shared, is usually not at the same level as the data provided by Target and Walmart.

Jon described the maturity phases we have gone through to better leverage demand signal data. In the first generation, Walmart and Target built massive data warehouses that contained data about a manufacturer’s sales by store and SKU, as well as their inventory (at the SKU level) at both the retail store and warehouse levels, and other sets of data that allowed consumer packaged goods (CPG) companies to see what their sell through was.

The retailers gave the data for free, but it was up to the manufacturers to figure out the use cases. Kraft, for example, leveraged this data to see distribution gaps and send more inventory to where it was needed. Some CPG companies were successful in using this data for certain use cases, many others were not. First generation processes were based on sharing data, not collaboration.

The second maturity phase involved retailer/manufacturer collaboration. The goal was to make the data actionable by joint processes called playbooks. Jon did not get specific here, but based on past research, I know that one playbook example involves detecting out-of-stock situations around a particular promotion at the store level. It is not enough to detect that a store shelf does not have inventory on it; a root cause analysis followed by actions by either the retailer (pull stock from the back room) or the manufacturer (reallocate inventory and increases the shipment size to a particular retail warehouse) may be needed to correct it. Jon spoke of the ability to develop as many as 50 playbooks.

In this second generation maturity phase, you don’t start with the data; you start with the use case. For example, if the goal is to make sure that a retailer does not buy too much inventory to support a promotion, you start by asking “What data do we need?” You would then work backwards to create analytics, alerts, and collaborative playbooks.

Jon also made the point that some retailers that were late to the game of sharing downstream data are now actually more advanced in collaboratively developing and acting upon these playbooks than the retail titans that invented the game. Delhaize was singled out, for example, as a retailer that would accept out-of-stock promotional alerts and whose merchandisers would actually act upon those alerts.

Third generation demand-driven companies seek to become more real-time enterprises by leveraging sell-through data to better make what people are actually buying. I’ve written this topic before (see Del Monte Foods and Shelf-Level Collaboration). Today, this is state of the art and it is the direction Kraft is headed toward.

Jon’s vision of fourth generation retail/CPG collaboration is retailers and manufacturers operating with one joint supply chain, instead of each operating their own. If a manufacturer and retailer each operate their own, then they each keep their own pools of safety stock. That really should not be necessary. Kraft also has a vision for how CPG manufacturers could better collaborate to serve a particular metropolitan market. Debbie Lentz of Kraft, who has been speaking of this vision for some time, has been elevated to the Senior Vice President level.

When I did a deep dive into shelf-level collaboration two years ago, I concluded that this would be one of the next big things that could allow our field to advance. However, I concluded most companies were probably years away from really being able to act upon this vision. That was why I was so excited by Ron and Jon’s presentation. Things have apparently moved far faster than I thought possible.

I recently spoke with Mike Spindler, the CEO of ShelfSnap, about the company’s entry into the “shelf-level collaboration” arena.

(Quick aside: As I mentioned in a previous posting, I prefer the term “Shelf-level Collaboration Solutions” over Demand Signal Repository (DSR) products because I believe it’s both more accurate and evocative.)

ShelfSnap is in start-up mode, and although the company has conducted some interesting pilots for large consumer goods companies, it has not yet been selected as a long-term partner by these companies. Nevertheless, I think ShelfSnap offers an interesting solution that fills a void when it comes to shelf-level collaboration.

Right now, DSRs are populated mainly with POS, syndicated data, store-level inventory data, and shipments from the retail DC. One form of shelf-level collaboration occurs when a DSR merchandising application sends out an alert saying that an SKU at a particular store is out of stock at the shelf, or that it is selling at a much slower rate than expected. These alerts are based on inference. The application does not know that a store has not restocked the shelf; it makes that prediction based on math. The math can be very simple: this product is jumping off the shelves at other stores but it’s not moving at this store; we know that inventory was shipped to the store, so this must mean that stockers have not moved the product out of the backroom yet. The math gets a lot more complex when it comes to predicting things like inaccurate perpetual inventory levels.

Mike’s point is that these are inferences. If you tell a Walmart store manager that a shelf needs to be restocked in the hope that he will assign someone to do the job, you better be right. You don’t want to generate false positives and send labor to the shelf to fix a problem that does not exist. Even if only 10 percent of your alerts are false positives, you could become “the boy who cried wolf” too often. So, some consumer goods companies use the merchandising alerts to reroute their third party merchandisers to visit a particular store to make sure the alert is correct.

But Mike points out that this does not fully solve the problem. Mike began his career at Nielsen, where he used to ask a group of his managers to separately go to a store and check to see if the planogram was correct for a particular category. In this store, Product A should have three facings and be in the middle of the top shelf, Product B should have one facing and be on the far right hand side of the bottom shelf, and so on. Mike said the managers never came back with the same data.

The ShelfSnap solution is based on workers using a digital camera to take a picture of the store shelf, and then the solution’s patent-pending Image Recognition Technology analyzes color, package size, and label design to identify the correct product. These pictures are fed into the ShelfSnap Analytical Engine that then generates quantitative analytical results: What shelf was the product on? How many linear feet did the facing have? Was the product out of stock? This data can then be turned into prioritized merchandising exceptions. They also provide analysis of new product cut-ins, displays, and point-of-purchase display reporting.

ShelfSnap Analytics (Source: ShelfSnap; click to enlarge)

I basically see the ShelfSnap solution as providing a practical way for manufacturers and retailers to use planogram data for shelf-level collaboration, or for retailers themselves to have more effective task management at the store. I also see it as improving the alerts generated by existing DSR solutions. Say a DSR is predicting that the perpetual inventory level at a particular store is incorrect (the store says it has 12 of the item, but the DSR application thinks it only has 9). This math is based on lower sales than expected for that item at that store. But if the product is supposed to have three facings on the prime shelf but it only has one facing on the bottom, the product location could explain the variance in sales rather than an incorrect inventory level.

In a recent posting, Adrian Gonzalez quoted Jim Stengel, the former global marketing officer at Procter & Gamble, on the greatest challenges facing marketers in the consumer goods industry. To paraphrase slightly, Jim asked how consumer goods manufacturers could add “digital” to “their branding model?” How can they “work with retailers to have maximum impact on value creation?” This is an application that contributes in both of those areas.

Kraft Foods Leverages Downstream Data to Drive Supply Chain Collaboration

Filling a Void in Shelf-level Collaboration

Archives

Login

A complete analysis of logistics trends, technologies, and services

Archives

Login