Change management and innovation need to be driven into supply chain management

A key lesson acquired from the harsh economic climate is that supply chains need to be modernised, and logistics policies need to be innovated to address current challenges.

This is never easy and can be burdensome to many companies that still need to place change management on their radar. An article on the Harvard Business Review (HBR) website discusses this in significant detail. Below are insights from the first half of the article.

Building a unified view of demand

The HBR article points out that the journey starts by rethinking the demand-planning process. Traditional approaches employ consensus forecasting, in which each function—operations, finance, sales, and trade (which is responsible for marketing, promotions, discounts, and so on)—uses standard statistical techniques, historical sales data, and some external data to generate its own forecast. Then all the functions get together and hash out a compromise uniform forecast.

That process has two drawbacks. First, it takes a long time—typically four to five weeks—to generate the various forecasts and reach a consensus that satisfies all business requirements. By that time the sales data used is old. Second, rather than agreeing on the data and having the analytics produce a single forecast, the people involved typically focus on finding a balance between conflicting forecasts and rely on gut feelings about what drives sales, revenue, and margins.

A much better way to generate a unified view of demand is to start with the sets of data that all participants agree will yield the most accurate picture. The CPG manufacturer, for example, chose four kinds:

• internal data on shipments to retailers, prices, discounts, promotions, and various product characteristics
• data on consumer demand, which can be accessed through retailers’ point-of-sale technology or provided by companies such as IRI and Nielsen
• macroeconomic information—including quarterly GDP, the Purchasing Managers’ Index, the Consumer Price Index, and unemployment and inflation rates—that helps explain consumer behaviour, seasonality, and trends
• external data on other factors that can indicate or affect demand, such as web searches, social media mentions of products, average temperature, precipitation, holidays, and competitors’ prices

The article adds that using such data and advanced analytics, firms can set up an automated five-step circular process that generates supply, financial, and trade plans for the next 50 to 80 weeks—the planning horizon for most companies. Here’s what that process looks like at the CPG manufacturer:

• First, trade-planning information—about future promotions, discounts, and marketing investments—is combined with consumer, macroeconomic, and external data to generate a market demand forecast by SKU and retailer for each week of the entire horizon. From what we’ve observed, most CPG companies have never tried to predict demand at such a granular level.

• Second, the demand forecast for each retailer is combined with historical data on the company’s shipments to that retailer to generate a weekly forecast of the retailer’s orders of each SKU for the horizon.

• Third, the company aggregates all the order forecasts and converts them into a feasible supply plan. The plan considers available resources, including inventories of raw materials and finished goods, manufacturing capacity constraints, and market targets (say, for increased sales of a product category at a given retailer-region combination). It also aims to achieve certain performance goals. The CPG firm focused on minimising total supply chain costs, but the chosen objective will vary from firm to firm. At some companies, for instance, it may be to maximise revenue or the amount of supplies produced.

• The fourth step is to use the weekly SKU supply plan for all retailers to generate revenue and gross margin forecasts at the brand level for every month of the planning horizon.

• The fifth step is to compare that financial forecast with the firm’s business objectives. A gap between the two may trigger a change in the trade plan—for example, the addition of more-aggressive discounts or increased investments in marketing to stimulate sales.

Key questions

When they were considering the adoption of this new process, the CPG firm’s managers raised a number of questions—which are representative of the kinds of concerns most executives express about our approach. Let’s examine them one by one.

What degree of forecast accuracy can the process achieve? Research has proved that variability in customer demand is significantly lower than variability in retail orders—a reality that underlies the well-known bullwhip effect in supply chains. This implies that predicting consumption should be easier than predicting retail orders, and indeed, the accuracy of the CPG firm’s forecast for market demand is quite high. At any moment, the demand forecasts at the SKU, week, and retailer level for five to eight weeks out have proved to be 85% accurate.

Combining the more exact consumption forecast with historical retail orders allowed the CPG company to improve its forecast of retailers’ future orders. The accuracy of the weekly order forecasts has been 15 to 20 percentage points higher than that of the standard, consensus-based forecasts the company previously used. And more-accurate order, or shipment, forecasts clearly translate into a more effective supply plan, which reduces lost sales—therefore boosting revenue—and improves service levels and the customer experience.

Finally, because the inputs into it are more accurate, so is the financial plan. In multiple implementations of this approach at several CPG companies, the accuracy of the financial forecast made at the beginning of a given month for the next month rose to 95% to 97%.

Will we be able to understand what drives the behavioural and other changes the plans predict? This question is probably the most critical. Indeed, in our experience, virtually all executives are reluctant to blindly follow the recommendations of a black box developed by data scientists. They rightly want to be able to interpret and explain the output of the demand-forecasting process.

For instance, is an increase or decrease due to competitors’ behaviour, cannibalisation across products, promotions and discounts, or merely a special event or holiday? The good news is that the analytic technology today is mature enough to allow a single SKU weekly forecast to be decomposed into its basic components. This is done by explicitly modelling the data as a combination of key variables (competitors’ behaviour and so on) and estimating the contribution of each one to the forecast.

Executives also want to know the reasons why, say, the forecast generated last week is different from the one generated this week. This, too, is information that today’s analytic technology can provide by comparing the input data used to generate each of the forecasts.

Last, executives want to understand why forecasts and actual sales sometimes deviate. At the CPG firm, the answer is that sales are affected by the way pricing, promotion, discounts, and inventory decisions are executed by retailers—a dimension that the manufacturer’s planning team can’t see. For instance, the forecast might be off when a retailer experiences operational challenges in moving inventory onto the shelf or in implementing promotions or discounts according to plan. Information about the retailer’s inventory and prices paid by consumers at the cash register can reveal these problems, but in our experience, most retailers don’t provide it to their CPG suppliers. Thus, at the CPG firm, any significant gap between forecasts and actual sales triggers an investigation into the reason for the difference.

How can we ensure that all the functions follow the new approach? The answer is establishing a forecast centre of excellence that brings together people from various functions, information technologists, and data scientists. Their role will be to agree on the data to be used and let the analytics generate the forecasts and the supply plan according to the five-step process.

How frequently should we run this process? Here, the answer depends on the market cycles of the various businesses and brands. For most businesses, the demand forecast, retailer order forecast, and supply plan should be updated weekly or biweekly, while the financial forecast and the comparison with the firm’s objectives should be done monthly. But there are clear exceptions. Some of the CPG manufacturer’s products have short life cycles of only six or seven weeks. In such cases, companies need to update the demand forecast, retailer order forecast, and supply plan twice a week. (The same is true for makers of fashion products, whose selling seasons last no more than 10 or 11 weeks.)

Strategic thinking

Consumer demand and the current consumption habits impacting retail will play a massive role in reinvigorating any supply chain management strategy. Failure to identify this shift has been a major challenge for many South African companies and has been the root cause of financial distress in many cases.

In a future article, we will unpack the challenges surrounding supply and demand as well as redefining strategic thinking and planning.