Read the Original Article at http://www.informationweek.com/news/showArticle.jhtml?articleID=214600338
In the perfume business, it's new products, like last year's U.S. launch of Kate, a fragrance Coty branded for supermodel Kate Moss, that can make or break a company's year. But big hits also can lead to big problems.
When a product takes off, Coty must respond quickly to keep shelves full. But its ability to ramp up is dependent upon glass, packaging, and other suppliers."If we can't meet demand ... it annoys the retailers, the consumers lose interest, and we lose sales," says Dave Berry, CIO at Coty, whose other brands include Jennifer Lopez, Kenneth Cole, and Vera Wang.
Empty shelves are the scourge of manufacturing and retail. Just look at the annual shortages of the Christmas season's hottest toys or the rain checks stores must write regularly on sale items. At any given time, 7% of all U.S. retail products are out of stock, says AMR Research analyst Lora Cecere. Goods on promotion are out of stock more than 15% of the time.
That's why manufacturers and retailers are pushing for the next breakthroughs in demand forecasting, what has emerged as the discipline of "demand-signal management." Instead of just relying on internal data such as order and shipment records, manufacturers are analyzing weekly and even daily point-of-sale data from retailers so they can better see what's selling where. This sort of timely, detailed data lets manufacturers spot trends much sooner by region, product, retailer, and even individual store.
Predicting demand is particularly hard in a recession. Just last week, home improvement chains Lowe's and Home Depot came out with falling sales, forcing Lowe's to rein back its expansion plans. In November, after same-store sales fell 7.6% in one month, Best Buy CEO Brad Anderson described "rapid, seismic changes in consumer behavior." Intel isn't even forecasting revenue this quarter because global economic uncertainty makes it "particularly difficult to predict product demand."
While the need for demand management is playing out most dramatically among manufacturers and retailers, its use is spreading to other industries. Raw materials suppliers and logistics companies are applying it at the beginning of the supply chain, monitoring real-time consumption and inventory data at manufacturing plants rather than relying solely on orders. Utilities are doing finer-grained analysis of consumption patterns by tapping gas and electric meter data previously used just for billing. Even health care providers are exploring the use of detailed, real-time patient information rather than just admissions stats to better forecast occupancy levels and set requisite staffing.
But it's in retail and consumer goods where such demand analysis could quickly play a much more important role in separating winners from losers. In an economic downturn, people buy more based on discounts and promotions, so stock-outs on advertised goods could be even costlier to companies. The most sophisticated users of demand data stand to gain a competitive edge through the recession and into the upturn.
Handling demand-signal data presents the same problems real-time data causes in any industry: how to access and integrate high volumes of data, and then combine and analyze it alongside historical information.
With the advent of highly scalable data warehouses, low-latency integration techniques, and faster, deeper query and analysis capabilities, the technology is finally here, at a price most can afford. And with easier-to-use business intelligence tools, manufacturers and retailers are pushing analytic tools into the hands of front-line decision makers, most often field sales and marketing people involved in planning, merchandising, and supply chain management.
The payoff from early efforts by Coty, Goodyear, and Kimberly-Clark has been more accurate forecasting, higher on-shelf availability, and more effective promotions. With faster and more detailed insight into demand, manufacturers as a whole can ratchet up revenue by 2% to 7%, says AMR's Cecere.
Wal-Mart Sets The Trend
Wal-Mart has been moving toward demand-signal analysis for more than a decade--and pushing its suppliers that way as well. Wal-Mart certainly has the leverage: It accounts for 28% to 40% of the sales of most consumer product manufacturers, according to AMR. Wal-Mart introduced its Web-based Retail Link data-access site in 1996 and five years later stopped supplying its data to independent market research firms. The goal was to force manufacturers to rely on Retail Link's detailed, store-level data to do a better job of supplying its stores.
To make the most of Retail Link data, Coty built a Wal-Mart-specific data warehouse five years ago, and it started using Information Builders' WebFocus business intelligence software to develop reports and forecasts. The frequency and granularity of the data and reports have steadily improved to the point where Coty now does weekly forecasts for individual Wal-Mart stores.
Over the last two years, Coty has pushed the responsibility for developing accurate forecasts down to its salespeople. Field-level forecasting makes for more accurate and responsive planning, says CIO Berry, who credits an analytics application from vendor CAS with making it easier for salespeople who are new to BI to analyze point-of-sale data and develop forecasts.
But even a decade into Wal-Mart's push, most companies are just starting to make full use of Retail Link data.
Wal-Mart "dumped the data on suppliers and said, 'OK, now we're expecting you to improve service levels and reduce inventory because you have this information,' but many of them weren't prepared to handle all that data," says Sandy Markin, a director in SAP's supply chain management division. Even now, though, only leading-edge companies use point-of-sale data for supply chain management, Markin says.
Another obstacle to broad adoption of demand-signal analysis has been the lack of standardization beyond Retail Link. Coty gets point-of-sale data from the likes of CVS, Target, and Walgreens, but each uses a different format than Retail Link. "The timeliness, accuracy, and depth of the data also varies from retailer to retailer, so it's tough to bring it into a data warehouse," says Berry.
Goodyear's Long Road
A demand-signal initiative started by Goodyear Tire & Rubber three years ago shows how hard it can be to get complete, standardized point-of-sale data from retailers. It had a head start, since it has 600 company-owned stores and 1,200 Goodyear-affiliated independent dealerships. But it had to work out data-supply arrangements with scores of distributors and national retailers, including Sears and Discount Tire. Its goal this year is to have demand-level data for at least 90% of its products sold at retail, up from about 70% last year.
"We're getting fairly granular data down to the product code, day sold, and store location," says John Wright, Goodyear's manager of business intelligence. The data helps track new product launches and the effectiveness of advertising campaigns across the country.
Retailers send data to Goodyear daily or weekly via flat-file transfers or EDI. Wright's team loads it into a Teradata data warehouse that also stores transactional information from Goodyear's SAP ERP system, and the team uses IBM Cognos Cubes and ReportNet to analyze and report on sales trends.
More than 100 channel, brand, and category managers use the data for forecasting and planning, and sales reps access basic reports through a CRM system. Goodyear also has developed scorecards for retailers so they can see how specific products are selling by region and how they compare with dealer averages on a variety of metrics.
Food Lion will add suppliers to its portal, Prothero says, till it hits a point of diminishing returns
It's been a long road for Goodyear, but this data lets the company see spikes or dips in demand more quickly than it would based on orders and shipments, or broader data on industry shipment totals. "Those reports lag behind by weeks," Wright says, "and they aren't very granular."
Taste Of Success
As retailers see the benefits, they're sharing point-of-sale data and encouraging suppliers to do fine-grained analysis. The 1,300-store Food Lion supermarket chain, for example, launched a portal called Vendor Pulse for nine of its biggest suppliers in 2007, and last year it expanded the program to more than 80 of the company's 2,000-plus suppliers. It will keep adding manufacturers this year "until we see diminishing returns based on the vendor's volume and presence within our stores," says Troy Prothero, the company's supply chain manager.
Food Lion started Vendor Pulse with two main goals: avoid stock-outs and, at the other end of the supply chain, reduce returns of goods that went past their expiration dates. The grocer also wanted to improve logistics planning and reduce losses by helping spot "phantom inventory" that stores think they have but don't because of theft, warehouse mistakes, scanning errors, or other reasons.
10% stock-outs? not good enough for Pike.
In September, Kimberly-Clark used Vendor Pulse to measure a Scott towels and Cottonelle bath tissue promotion. The analysis revealed stock-out rates as high as 18%. Not good.
Before it had Vendor Pulse data, Kimberly-Clark knew what it had shipped to Food Lion, but "we couldn't see what was happening in each store," says Greg Pike, Kimberly-Clark's team leader for Delhaize USA, the parent of Food Lion supermarkets. Now, with Vendor Pulse, "we can see the in-stock metrics by store on a day-to-day basis."
After adjusting stocking forecasts for each store based on the Vendor Pulse analysis, Kimberly-Clark reran the same promotion in October. Stock-out rates averaged less than 10% and sales increased a whopping 167%, Pike says. The company hopes to do better on future promotions by doing more timely and sophisticated reads of the data. Obviously, the data flags that a store has run out of product when sales go to zero that day. "But we're now looking at sales on preceding days to better predict what's happening," Pike says.
The challenges of transmitting, integrating, and analyzing point-of-sale data on a daily basis can be monumental. The Vendor Pulse portal runs on a software-as-a-service platform from Retail Solutions, so Food Lion doesn't have to manage the secure connections and data integration needed for suppliers to access it. Retail Solutions also provides an OLAP-style analysis tool. Retail Solutions says it has more than 25 retail customers for its demand-signal management service, accessed by 250 suppliers.
Vision Chain, Shiloh, and TrueDemand are among the other specialized vendors in this market (see "Software Giants Join Specialists In Demand Management" ). However, Teradata, SAP, and Oracle all have announced demand-analysis offerings within the last year, further evidence that demand-signal analysis is headed for the mainstream.
The approach a company takes to demand-signal analysis depends in part on the industry it's in and the availability of retail point-of-sale data. Food Lion could turn to a vendor with a SaaS offering developed for consumer packaged goods; Goodyear had to build its demand-signal repository in-house, as it's very industry specific.
Neither option is cheap.
Maintaining a system in-house could easily require several additional full-time employees--people who make upward of $100,000 a year--just to handle the data collection and integration. Add to that the cost of a larger-scale data warehouse (ranging from $50,000 to $100,000 per additional terabyte of capacity required), the cost of additional BI software, up-front infrastructure costs, and perhaps salaries for a few analysts, and the investment easily runs into six or seven figures per year, depending on the frequency, volume, and number of data sources.
Before reinventing the wheel, companies should explore data standardization initiatives, such as GS1's Global Data Synchronization Network, and commercial software and online service alternatives, such as Retail Solutions, Vision Chain, and DemandTec, which can help with retail data integration or complete outsourcing of the demand-signal repository.
SaaS or hosted off-the-shelf software options are likely simpler and cheaper than in-house development. Retail Solutions' service, which combines repository, daily data integration, and analytics software, runs $30,000 to $70,000 per year, per retailer, depending on the complexity and volume of the data. Major manufacturers could get into seven-figure annual contracts as they track data from 20 or more retailers.
However, if demand-signal analysis can help manufacturers gain anywhere close to 2% in sales--the low end of AMR's estimate--then the costs described above would amount to a rounding error compared with the potential benefit. Goodyear and Kimberly-Clark, for example, each report roughly $19 billion in annual revenue, so a 2% boost would add $380 million to their top lines. Both companies are in the early stages of their initiatives and declined to discuss actual returns, but both Wright of Goodyear and Pike of Kimberly-Clark say they plan to go deeper with demand-signal analysis.
Start With A Reality Check
Once you've established that demand data is available, start with simple reporting and alerting. It isn't hard to detect low stock levels and send out alerts, but some of the specialized apps get into sophisticated statistical and predictive analytics. Before retailers or manufacturers get carried away, though, they must make sure they know what problems they're trying to solve--and whether there are corrective actions they can take.
If the problem is with the retailer, a supplier has to know whether it can change the business process or get the retailer to act differently. "If you can't address the root cause of the problem, you might as well not bother with the analysis," says Cedric Guyot, VP of marketing at Retail Solutions.
Executing better on promotions and avoiding stock-outs are the big opportunities, because they increase revenue without hiking the fixed costs of manufacturing, distributing, and advertising the goods. But that's where the science of analysis and the art of experience come in. With promotions, the challenge is recognizing when retailers aren't executing on plan, and with stock-outs, the trick is avoiding false alerts that can compound stocking errors (see "5 Dos And Don'ts For Using Demand Data" ).
Demand-signal analysis may provide earlier and more granular insights, but the problems manufacturers and retailers are trying to solve have been around a long time. It's essential to have people steeped in industry knowledge leading this effort. "The companies that are in the best shape have someone from the line of business who owns the data and who's working on how to transition to using point-of-sale information," says AMR's Cecere.
Phrases like "real-time data" and "analytic insight" have been kicking around in lots of industries in recent years, but they're just buzzwords until somebody figures out how to use emerging technology trends to solve a real business problem.
In this economy, the hazards of getting demand wrong multiply. The risk is clear in a consumer who clips coupons and waits for sales, only to give the store brand a try when the name-brand manufacturer's out of stock, possibly switching to the cheaper brand for good.
But that kind of scenario can play out in any number of industries, making demand-based management a prescription for these times. With consumption patterns shifting, sometimes wildly, the predictive value of historical information is suspect.
Health care companies need to manage their bed space not just based on admissions, but on the conditions patients have as they come in and their progress as they're treated. Banks can train staff on new products in close line with demand, rather than en masse. There's a serious risk of procuring, staffing, and training out of sync with the real market--whether for raw steel, pickups, railcars, energy, hospital beds, airline seats, or loans. And it's highest for those still using rear-view-mirror methods to anticipate demand.
Software Giants Join Specialists In Demand Management
5 Dos And Don'ts For Using Demand Data