The Intelligent Food Chain

Welcome to The Intelligent Food Chain - a series of news and opinion pieces where we explore how business intelligence (BI) can be used in the food value chain.

Please come back often to review our bi-weekly updates about BI and such core processes as inventory management, customer service, and assortment optimization to name but a few.

A Fortune 1000 food manufacturer in our area started its implementation of enterprise resource planning (ERP) software in 2002.  It was a massive undertaking, with an internal project team of almost fifty supplemented by external consultants from one of the Big Four consulting firms.  The initial implementation project cost more than $100 million over several years.

Recently I had the opportunity to attend a presentation about demand planning made by a representative of that company.  He related that, in the demand planning function, they:

• Use the ERP system, its advanced planning and optimization ( APO ) module, and its data warehousing module
• Consider the integration between these three modules poor relative to their needs
• Do the bulk of their forecasting and demand planning work by extracting ERP and APO data into a complex series of Access databases and Excel spreadsheets
• Still are using the version of the APO module they implemented five years ago, which now is several software versions obsolete
• Base their forecasting on their own shipment data rather than retail consumer off-take
• Receive point-of-sale data from three retail customers (supermarket chains), but don’t really use it in their demand management process

From an enterprise-wide view, five years after initial implementation this company:

• Continues to struggle with having timely and accurate master data
• Has trouble integrating acquisitions into the ERP system and into its planning processes
• Is criticized by Wall Street analysts for its level of working capital, but has yet to be successful increasing inventory turns

As I drove home from the presentation, the thought that kept resonating in my mind was, “They spent more than $100 million to implement ERP, and five years later they’re still doing their demand planning using spreadsheets and Microsoft Access.”  Yet this company is not a “poor performer” and its situation is not a reflection on the ERP software it selected.  Its experience is fairly typical of Fortune 1000 global food manufacturers.  Many spend comparable sums of money on ERP and achieve similar results.  These results aren’t what executives were promised when they approved the investment of tens of millions of dollars for an integrated enterprise-wide system.  For the vast majority of food companies (and, indeed, all companies), the process-streamlining, revenue-enhancing, cost-reducing dream of ERP systems remains unrealized.

Why is success with enterprise-wide systems like ERP so hard to achieve?  There are several reasons:

• The benefits were oversold (the dream was too big).
• The amount of process change required to fully leverage the capabilities of the new software was underestimated.
• Using a new enterprise system requires wide-spread cultural change, which always is difficult.
• Commercial-off-the-shelf (COTS) – that is, packaged software – ERP systems are extremely complex, because they are designed to be able to be used for a wide variety of industries, manufacturing methods, business models, company sizes, and so on.  Any one company uses only a small fraction of the functionality of the software, but determining that subset and configuring the software for it can be quite challenging.
• The initial implementation typically is so expensive and so exhausting that no funds and little enthusiasm remain for the continuous process improvement that make the software and the processes enabled by it perform at a high level.

What does this discussion of ERP and its failure to live up to expectations have to do with business intelligence (BI)?  For food manufacturers, the ERP system is one of the most important data sources for BI.  And BI to leverage the ERP data to improve decision making is one of the next steps that often isn’t done due to funding exhaustion or diminished enthusiasm.  What should food companies in this situation do?  Here are some approaches to consider:

• Get your ERP system, and especially its master data, stable first.  Running BI against poor quality data is like building a house on quicksand.
• Start with small, focused BI projects that deliver immediate business benefit.
• Be patient.  User fatigue with process and system changes cannot be wished away.  It is better to do less (build fewer, smaller BI applications) and have what you implement be widely and effectively used.
• Take the broader view that shortcomings of the ERP system actually might be opportunities for BI -- to integrate data, facilitate analysis, present information more visually, and more.

The challenges of effectively using ERP systems are real and will not be completely resolved any time soon.  As BI champions, we need to recognize these challenges and factor them into our thinking and planning, but they need not prevent us from accomplishing big things with BI.

American families spend 9.8% of their income on food, one of the lowest percentages around the world.  For years, growth in personal income and low inflation in food prices have made food a bargain for U.S. consumers, enabling them to grow consumption of new categories and premium-quality products while building affinity with their favorite brands.  Consumer product companies have responded with greater variety, more frequent promotions and marketing programs aimed at reinforcing brand loyalty.   But things have changed quickly in the last few months, and brand loyalty now stands on shaky ground.  

After seeing typical changes between 2 and 3 percent annually in food prices, consumers will see an increase of 5 to 6 percent this year.   In addition, personal incomes likely will not increase any time soon.  As a result, buying patterns are shifting toward lower-priced brands. And lower-income shoppers are not the only ones trading down from their favorite brands: a recent survey by IRI found nearly one-third of high-income shoppers said they bought more private-label products during the second quarter, up from about 20% in the first quarter of this year.

As consumers pay more attention to their grocery receipts than to their favorite brands, consumer product companies are revisiting their marketing mix plans.  For instance, Procter & Gamble has changed its marketing approach to focus more on in-store promotions.  Other companies are focusing more on category management.

For companies to be successful in their brand strategies and cope with shifting buying patterns, they need to become more aggressive in measuring the effectiveness of their marketing programs.  By incorporating measures such as return on program investment, consumer product companies will be more capable of allocating dollars to the right accounts, channels and products that minimize erosion of consumers’ loyalty to their brands.

In addition, companies seeking to retool their category management and shopper marketing programs need to be more proficient at exploiting point-of-sale data.   Only by looking at store level and customer segment detailed transactional data will they understand how to tailor promotions to each cluster, market or segment.

Business intelligence (BI) solutions provide the blueprint and technology to measure effectiveness and integrate data from multiple sources, both internal and external. Today’s rapid changes in shoppers’ buying patterns, call for marketing mix plans that can change at the same pace.  BI provides the means to make those changes.

Through 2007 and the first half of 2008, food manufacturers faced steep price increases for commodities like corn, wheat, and soybeans.  Food prices in general increased, as manufacturers attempted to recover the commodity cost increases in the pricing of their packaged products.  Now commodity prices have declined significantly from their summer highs.  And the deteriorating state of the U. S. economy is causing consumers to be particularly price conscious.  The convergence of these two events forces food manufacturers to pay close attention to their pricing: if they raised prices when their commodity costs went up, should they lower prices now that commodity prices have declined, and consumers are price conscious?

Economists use the term “price elasticity of demand” to describe the relationship between the quantity of a product that consumers will purchase and the price they pay for it: in general, the higher the price, the less quantity they will buy.  (Answering “How much less?” is where the mathematics comes in.)  Understanding the price elasticity of demand is one of the means to a very important end: determining the price for each product at which the food manufacturer maximizes its profits.  For food products sold at retail through grocery stores and supermarkets, the relevant price is the price paid by the consumer at check-out, which is a function of both manufacturer and retailer actions.  This means that, in order to understand the price elasticity of demand, the manufacturer must analyze point-of-sale (POS) data.  Many food manufacturers are building specialized data bases called “demand signal repositories” to contain POS data and enable price elasticity of demand and other types of analysis.

Understanding the price elasticity of demand is the first step in selecting the price that will maximize profits.  The second step for the food manufacturer is to understand its own cost structure, particularly which of its costs are fixed and which are variable.  Reducing the price in order to drive more volume makes the most sense when the majority of the cost is fixed, and therefore spreading it over a larger sales volume results in less cost per unit.  Some fixed costs (such as those associated with manufacturing capacity) typically are fixed over a certain range of volume with “step function” changes in cost at the endpoints of that range; both the volume ranges and the associated cost function must be well understood in order to find the price/volume combination at which profit is maximized.

Both of these types of analysis – price elasticity of demand, and fixed versus variable cost – require sophisticated mathematical techniques to be applied to large volumes of detailed data.  They are well-suited to business intelligence solutions because BI applications can acquire the necessary data from multiple internal and external sources, organize it, manipulate that data using the appropriate algorithms, enable analysis across multiple dimensions (product, plant, customer, time, and so forth), present the resulting information in ways that are easy to understand, and support simulation of “what if” scenarios.  Effective pricing is critical to optimizing profit; many food manufacturers could benefit from business intelligence to strengthen their capabilities in pricing.

The Wall Street Journal (September 10, 2008) included an article describing how some retailers are using “workforce-management” or “human-capital management” software to better schedule the workers in their stores.  These systems calculate performance metrics (average sales per hour, dollars per transaction, and so on) for each worker, and schedule the most productive workers to work the busiest hours.

As you might expect, the use of such systems is controversial, and many workers dislike them.  The more interesting aspect of these systems is that they are an example of leveraging information to improve decisions that matter.  As an expert quoted in the article states, “The biggest controllable cost in retail is people.”  Optimally scheduling the workers (how many people to have on duty, and which ones, during each part of the work-day and work-week) not only is key to controlling costs, but also greatly affects customers’ service experience.  And that’s the essence of the problem/opportunity: determining the optimal staffing that minimizes labor cost while maximizing the customer experience.

The companies mentioned in the WSJ article aren’t food retailers, but what’s true for AnnTaylor Stores, Limited Brands, and Gap Inc., also is true for supermarkets: labor costs for in-store personnel are a significant cost, and proper staffing drives the customer experience.  Supermarkets know that not having enough checkers on duty during a surge of customer traffic means long check-out waiting lines, grumpy customers, and some customers who will defect to a competitor who has shorter lines and faster check-out.

Workforce-management systems demonstrate a fundamental element of the power of BI: to transform data into useful information that enables better/optimal decisions to be made about an important aspect of the business.

We who work in the food industry are accustomed to its unique requirements, and often don’t realize how demanding they truly are.  Take traceability and the ability to do product recall.  As food manufacturers, we must have the capability to take a single code-dated package in the hands of a consumer and trace it back through retail (the supermarket), the supermarket distribution center and/or distributor or wholesaler distribution center, our own distribution center(s), our own plant and/or contract manufacturer, through to a specific lot from a specific vendor, potentially for each ingredient in the recipe.  We must accomplish this trace quickly, potentially in a matter of hours.  And then, once we’ve identified the problem, we must explode our trace back down the supply chain to identify every case of potentially affected product at every stage of the supply chain, all the way back to the retail shelf.  While the likelihood of having to execute a product recall is remote, the consequences of failure are huge, possibly including putting us out of business.

It’s true: other industries, such as pharmaceuticals, have as stringent or more stringent product recall capability requirements.  But pharmaceuticals have both high prices (often several dollars per pill) and high margins to absorb the cost of traceability systems and processes.  We don’t have the luxury of their cost structure.

The portion of traceability within our own walls is fairly straightforward.  Even though code dating of finished goods and lot traceability of ingredients isn’t innately designed into most ERP systems, it usually can be accomplished using lot numbering or serialization functionality, perhaps modified.  The challenge comes from the inventory that resides or already has flowed through our downstream partners in the supply chain: carriers, distributors/wholesalers, and retailers/customers.  Because of the importance of managing inventory age (i.e., FIFO to reduce the risk of aged obsolescence), all food supply chain participants track code dates in their inventory systems.  But each one of these many participants might use a different system and have implemented code date tracking in a different way.

Obtaining the data we need to do product recall from our supply chain partners can be a challenge, but equally challenging is what to do with it, once we have it.  We must be able to combine differently-defined and differently-structured data from multiple sources to provide a complete picture of the potentially-affected inventory.  Luckily, integration and standardization of data from multiple disparate sources to provide a summarized view of a business situation is one of the things business intelligence/data warehousing (BI/DW) applications do well, making lot traceability and product recall a food manufacturer imperative well-suited to BI/DW.  It’s a capability that we can’t afford not to have, and BI/DW can help us get it.

An important aspect of increasing revenue is the day-to-day struggle to increase distribution of your products: to secure placement of more items, obtain more facings, and defend what you already have from competitors.  Account reps spend a lot of time working to increase distribution.  But how do they know exactly what to propose to the buyer?

Let’s say that you are selling a line of cake mixes.  The line contains twelve varieties, of which this account currently carries four.  You want to convince the buyer to take another two varieties.  Which two of the eight should you propose?  The obvious answer: the two, that, when combined with the existing four, will yield the most unit sales, revenue, and profit for both you and the account.  But determining the specific answer to that question for that account can be analytically challenging.  It definitely is not something we want the account rep doing on the back of an envelope while in the customer’s lobby, waiting for his/her appointment with the buyer.

In some product categories, the relative ranking of varieties is well established.  In ice cream, vanilla, chocolate, and strawberry have a strong lock on the top three flavor spots.  But other categories are less well structured.  Cake mix, for example, has regional variation, with pound cake being more popular in the South than in any other region of the country.

The optimal number and selection of varieties or flavors for a specific account depends on a number of factors:

•           The number of varieties in our product line

•           Which varieties the account already carries

•           The relationships (substitutability, complementariness, etc.) among varieties (If two of the four varieties the account already carries are chocolate-based, say Devil’s Food and Chocolate Fudge, then perhaps pitching another one (German Chocolate) is not the way to go.)

•           Number of items and facings for the category for this account

•           The variety lineups of our competitors, and the extent to which they match ours

•           The number of competitive items carried by this account, and which ones they are

(If we’re selling cake mix, and the account already carries Yellow cake mix from both Duncan Hines and Betty Crocker, perhaps we should pitch Angel Food rather than our Yellow.)

Business Intelligence (BI) can help us optimize our assortment with each account.  Whether all we have is our own shipment data (good), or we have POS data for our items sold by this account (better), or we have POS data for all items in this category for all accounts (best), BI can support fact-based understanding of which numbers and combinations of which varieties maximize unit sales, revenue, and profit.  “Our other accounts that carry six varieties, including Angel Food, have x% more unit sales per week than those that only carry four varieties” is a compelling argument that BI can enable us to make to the buyer.

Individual account category management is a data intensive and analytically complex undertaking, making it a good candidate for a BI application.

Much has been said in the media recently about how the economy is affecting consumers’ shopping habits for food.  Grocery retailers are experiencing significant paycheck-to-paycheck seasonality in sales patterns, and are responding by varying assortments throughout the month: more discretionary and indulgent products after payday, and low-cost staples like rice and beans before payday.  What do these within-month assortment changes mean for upstream players in the food value chain, like wholesalers and manufacturers?  They mean smaller order quantities, more frequent orders, less time to remedy retail out-of-stocks, increased risk of returns: in short, more change, more complexity, and more stress on the supply chain.

What must wholesalers and manufacturers do to cope with frequently changing assortments at retail?  Fundamentally, they must uptempo the process by which they sense and respond to demand.  The manufacturer of an indulgence product that is “two weeks on, two weeks off” at retail cannot survive with a monthly replenishment process.  It must analyze both retail off-take from POS data and store-level inventory in order to supply just the right amount of product at the beginning of the two-week “on shelf” period so that the inventory all sells through before the two-week “off shelf” period begins.  Too little, and the product will be out-of-stock during a portion of the already limited selling period; too much, and the retailer might return the remainder.

The more frequently and quickly a planning process must be executed, the more important it is for it to be automated.  If a “sense and respond” replenishment process must be completed multiple times per week, there’s no time for semi-manual data acquisition, manual data cleansing, manual manipulation using spreadsheets, custom data integration by analysts and planners, or similar methods still ubiquitous in even large sophisticated food companies.  Accurate, integrated, useful, actionable information must be automatically presented to planners so that they can spend their time solving the business problem (right quantity of product to the retail shelf at the right time) rather than laboriously turning the crank on the process.  And that’s what a well-designed business intelligence (BI) application for replenishment planning can do: distill mountains of POS and store-level inventory data into information the planner can use.

For today’s piece we’ll start with a historical perspective, but first a quick definition of BI is in order.  While the words might vary a bit, there is a general consensus that BI is a business process improvement and performance management tool that leverages transactional business information from companies’ transactional systems to feed analytical tools that provide better understanding of complex business situations and company options, thereby allowing the company to improve the effectiveness and efficiency of the core processes that drive its business results.  A great example of BI use in the food industry Kroger’s use of loyalty card data (transactional information) to segment shoppers based on their buying behavior (analysis) in order to develop customized offers for the individualized needs of various shopper segments and better target its promotional offers toward its most profitable customers (increased revenues and profits).

As to the historical perspective, food industry companies in particular -- and supply chain experts in general -- have been talking for years about leveraging demand data to create a more efficient and responsive supply chain.  And some companies have been doing just that, i.e. using point-of-sale (POS) data to drive store replenishment.  That being said, the food industry has been slow to change when it comes to leveraging demand information, even though retailers are increasing competing on the basis of optimizing the retail shelf (increased focus on preventing out-of-stocks, category management, customized assortments, and other efforts) and thus moving from the old supply push industry model to a demand pull industry model.  The extent to which this shopper-centric model will be adopted by all participants in the food value chain remains to be seen, but there is no doubt that better use of demand data (transactional information) can feed sophisticated analytical applications (analysis) across core business processes such as assortment optimization, pricing, category management, distribution, inventory management, and supplier management (increased revenues, decreased costs, increased profits).  In other words, BI can make a difference in the food value chain, just as it has in many other industries.

To see past comments from The Intelligent Food Chain, please send us a note at FoodIndustrySolutions@DecisionPath.com.