By Andrew K. Reese
Procter & Gamble has made supply chain management a core competency, earning the company a top three spot in AMR Research's "Top 25 Supply Chains" rankings for the past two years. But even while it wins industry accolades, P&G has continued to invest in improving its supply chain performance. Recently the company undertook an initiative to boost its short-term forecasting accuracy as a way of lowering safety stocks and improving its chances of winning at what P&G calls the "first moment of truth" — when the consumer looks for a product on the store shelf.
Mark Kremblewski, a Toronto-based global business expert working on demand planning at Procter & Gamble, says that the company's decision to tackle the short-term forecast challenge stemmed, in part, from a corporate-level realization a couple years ago that P&G needed to do additional work to reduce its inventory levels. The project also offered a quick-hit opportunity to produce a measurable, verifiable return on investment. "It made sense to focus on safety stock reduction because there is a theoretical direct correlation between accuracy benefit and safety targets," Kremblewski says. "There are all sorts of models showing that if you improve forecast accuracy by X amount, your safety stock goes down by Y amount — it's very calculable."
At Procter & Gamble, the company's demand planners focus on short-term horizons (one to six weeks) as well as medium- to long-term horizons (seven weeks to two years) in their forecasts, using the demand planning capabilities within P&G's SAP enterprise resource planning (ERP) solution. P&G also has an analytics group that develops safety stock models. The group's analysts use Excel spreadsheets and internally developed applications to work with data pulled from shipping and billing systems on a daily or weekly basis, adding in forecast accuracy data and then running the numbers through the models to produce target safety stock levels — a tedious, time- and resource-consuming process.
What P&G's demand planners lacked was an off-the-shelf software tool for short-term forecasting. In fact, until just a few years ago, such tools did not really exist. But Kremblewski's boss, Dick Clark, associate director for demand planning at Procter & Gamble, was aware of a company called Terra Technology that provides demand sensing and inventory optimization solutions targeted at "real-time forecasting" (thus the solution's moniker, Real–Time Forecasting, or RTF). Based in Norwalk, Conn., Terra Technology was founded in 2001, and it has built a client base that includes some of the largest consumer products companies in the world, such as Campbell Soup, by offering something of a unique solution to the short-term forecasting challenge. "To our knowledge Terra were the only people that were playing in this very short-term improvement area," Kremblewski says.
Piloting the Black Box
Before fully engaging with Terra Technology, Procter & Gamble ran through an initial test exercise with the solution provider. P&G provided shipment and historical forecast data from several different businesses within the company. Terra ran the data through its solution to produce benchmark models showing the potential for double-digit improvements in short-term forecast accuracy using the Terra tool. These results were convincing in themselves, but Terra Technology's solution also was attractive because it could run as a "black box," according to Kremblewski. In other words, once P&G connected the Real-Time Forecasting solution to the necessary data feeds from the appropriate enterprise systems, it would run on its own, without Procter & Gamble's demand planners having to tweak the system or do maintenance on it.
With the benchmark results in hand and ready to move forward, P&G's demand planning team surveyed the company's various business units and their respective supply chains to identify a likely candidate for a pilot project. "We looked for a business that was willing and able and had some need," Kremblewski says, "an area that was looking for some improvement around safety inventory and that fit the 'sweet spot,' with a reaction time on which they based their safety stock in the zone of 10, 15, 20 days." That business unit wound up being in Western Europe, and it went live last March on a pilot project. The objective for the project was to improve short-term forecasting accuracy sufficiently to yield a 10 percent reduction in safety stock.
In moving into the pilot, the demand planners worked closely with the supply side of the business to ensure that as safety stocks came down, the reductions produced no disruptions in product availability. Over time, both demand and supply gained increased confidence in the solution's results, allowing for further reductions in safety stocks on an incremental basis. The demand planning group was able to facilitate that process by "baking" Procter & Gamble's own inventory models directly into the RTF solution. Over the course of several quarters, P&G was, in fact, able to document a decrease in its short-term forecast error of more than 30 percent within this business unit, which translated into safety stock reductions of greater than 10 percent, exceeding the goals for the pilot.
Based on the results of the pilot, Procter & Gamble has started to roll the Real-Time Forecasting solution out to other business units, initially in Western Europe and North America. The individual businesses have to pay for the solution out of their separate budgets, but P&G isn't mandating the solution for all its supply chains, simply because it is not appropriate for all of them. "If you've got a four-month supply chain from China, your forecast accuracy over the next two weeks or three weeks or six weeks is meaningless," Kremblewski says. Rather, the solution makes sense only where short-term accuracy can have a real impact, and those are the business units that Procter & Gamble is targeting.
In terms of lessons learned during the pilot, Kremblewski says that companies considering such a project should not underestimate the challenge of getting the right data feeds established to supply the Real-Time Forecasting solution with the information necessary to make the forecasts. "It's not because there's anything wrong with the RTF tool, it's because we do some very strange things in our own systems," Kremblewski says, referring to the various idiosyncratic ways that P&G, like any large company, handles its enterprise data. In addition, he recommends working closely with supply chain partners to reassure them that the changes brought about by the improved forecast accuracy will not have a negative impact. As with the company's own staff, these partners must also gain confidence in the solution's results, Kremblewski explains.
Procter & Gamble has estimated that deploying Terra Technology's solution ultimately will yield more than $100 million in increased cash flow globally for P&G, according to Clark. But the greatest benefit from adopting this type of solution may come over the long term, as Procter & Gamble is able to shift its demand planners away from the inevitable firefighting and scrambling inherent in focusing on short-term forecasts. "Our demand planners have more than enough things to do without focusing on the next one, two or three days, or the next week," Kremblewski says. "The ‘black box' takes care of all that, which allows our demand planners to focus on the longer term, on the trends, on the business intelligence regarding the marketplace, what our customers are doing, etc. And that's where their real value is added, not watching orders or watching short-term trends. If the box can do it, let the box do it."