Table 1: Prevalence of Forecasting
Current State of Forecasting: Process, Data and Tools
Figure 2: Inputs considered for forecasting Source: Primary Research (1)
Point-of-sale (POS) data is considered to be the most potentially useful information of past sales information. However, in discussions it was determined that there are significant obstacles to actually leveraging POS data — for instance, lack of availability of standardized POS data formats across countries entails manual aggregation for a complete view of sales. Another significant issue is the difficulty in incorporating available data into sales and operations planning (S&OP) heuristics.
Most companies prefer to have dedicated sales personnel at the customer site to have better visibility into forecast updates and other demand-side dynamics. "We keep an eye on Stocking Representatives i.e. intermediate agents between us and our customers' shelves, to figure out how much the expected demand could be," says a chip producer. A mobile phone manufacturer mentioned that the company has a specialist in marketing/demand management who looks beyond the service provider to the retailer and works with them directly.
Suppliers to original equipment manufacturers (OEMs) tend to believe that OEMs over-forecast to avoid market risks and so consider validating the OEMs' forecast outputs with information on market potential and industry trends provided by secondary research and industry journals. For example, a contract manufacturer mentions: "OEMs themselves have no visibility. We simply cannot rely on the forecast data received from the OEMs. Rather, we look at the demand pattern of our customers' customers, market potential and emerging trends, what applications are emerging in the market, and the way our customers would react to these trends."
Figure 3: Prevalence of Forecasting Tools Source: Primary Research (1)
Most companies [around 68 percent (1)] still rely on basic tools for forecasting. Although a few respondents indicated that they deployed tools like Manugistics, i2 Demand Planner or specific forecasting modules of leading enterprise resource planning (ERP) applications, tools like MS-Excel or homegrown applications continue to be a favored choice to help arrive at forecasts. On probing, two primary reasons were identified:
- Raw data is simply not adequate for sophisticated tools to provide a robust quantitative estimate. To remedy this, human judgment plays a significant role, and manipulations based on rough estimates can be done easily with basic tools
- Ease of use and wide acceptability of simpler applications and homegrown tools (particularly relevant in a collaborative supply chain context).
Forecast Performance: Key challenges and recommendations
The research indicated that around 70 percent of the respondents do not consider their performance in forecasting as satisfactory.
Figure 4: Forecast Performance Source: Primary Research (1)
The underlying issue, however, is not the lack of tools deployed for supporting the forecasting function. The issues primarily stem from being unable to collect "desired" data and also being unable to institutionalize a formal process of collating and utilizing input data/information to arrive at forecasts. Key challenges identified:
- Poor quality of data. Associated issues: Lack of synchronization of nomenclature for a product or part number followed by the company and its customers/suppliers. Mismatch in granularity of the expected demand data provided by customers and that is required by the company. Over-forecasting by partners. For global operations, POS data is not available across all countries.
- Lack of formal processes. Associated issues: Lack of formal processes for measuring forecast performance and feedback generation on previous period's forecasts into the current period's estimates. From the survey, 76 percent of respondents indicated an absence of a formal process measuring forecast accuracy. Eighty percent didn't have any mechanisms to capture previous period's forecast accuracy feedback into the current period's forecast. Inability to filter and apply qualitative information (e.g. market conditions, supplier dislocations, intensified competitor activity) to supplement quantitative inputs for historical performance. Forecasting is primarily only short-term focused.
- Forecasting tools are not leveraged fully. Associated issues: Poor integration of forecasting tools to transaction data source. Data inputs are not considered adequate for sophisticated tools. Data exchange is manual. Fax, phone and e-mail are still a popular method to exchange information. A resistance by users to switch from simpler applications and homegrown tools to sophisticated forecasting tools.