Forecast with Foresight

A short while ago, Supply & Demand Chain Executive ran an article by Romit Dey and Joy Prakash Somani summarizing the results of an Electronics Supply Chain Association and Infosys Technologies Limited study on re-thinking demand management.

The study, which was designed to assess the impact of consumerization on major sub-segments within the high-tech industry, understand issues and challenges, and identify industry leading practices, found that 87% of respondents stated that consumerization had significantly impacted product proliferation and customer experience. The demand for new products at an increased product refresh frequency has put increased pressure on design and supply cycles and customer expectations on product customization and error-free operating performance have heightened considerably.

The study found that performance was still as critical as ever, but that 70% of the respondents did not consider their performance in forecasting to be satisfactory. The challenges identified included:

  • Poor Data Quality
    There is often a lack of synchronization on product numbers between manufacturers, distributors, retailers, and customers; a mismatch in the granularity of the expected demand data provided by retailers and customers; over-forecasting by optimistic partners; and POS data is not always available, especially in global distribution networks. Furthermore, raw data is often not adequate enough for many tools to provide a robust estimate.
  • Lack of Formal Processes
    There’s a lack of process for measuring forecast performance and generating feedback on current performance to future estimates.
  • Forecasting Tools are Not Fully Leveraged
    Sometimes this is because of a lack of integration into data sources, sometimes it’s because the available data is not considered adequate, sometimes it’s because data exchange is still paper-based, and sometimes it is because users resist switching to new and improved processes and tools.

As a result, the authors recommend a shift from passive/reactive demand management to a more active/predictive form of demand management that:

  • senses the demands of customers early & correctly,
  • influences the demands to favorably align to capability,
  • budgets for variability in demand during fulfillment, and
  • focusses on innovation to realize a first mover advantage.

Furthermore, they indicate that organizations should:

  • streamline information gathering and analysis,
  • formalize forecasting processes,
  • leverage demand shaping opportunities, and
  • collaborate within and beyond the organizational boundary.

The latter is a good start, but the article fails to point out that it is essentially impossible to correctly sense the demand of your target market before production begins – which is when it is most important. Nor does it provide you strategies to account for the unpredictable variability that is going to be incurred as a result of this inability to accurately sense demand early.

Why can’t you accurately sense demand early? Sure you can measure excitement about a new product announcement or highly anticipated feature, but this can change overnight when a competitor announces a new capability or rolls out a new stealth product that you had no knowledge of. As a whole, leveraging demand shaping opportunities, polling the market, and connecting with retailers to get a better sense of actual demand will greatly increase your forecasting performance across the board, and increase the chance of a big win, but, on a project basis, there is still the opportunity for a big miss, and it will still happen occasionally.

That’s why I’d recommend including the following two steps in the process checklist:

  • utilize advanced demand & price point prediction technology based on optimization and simulation (such as that employed by Rapt, which was recently acquired by Microsoft) to make sure you get a reliable demand prediction at a target price point and
  • focus on contracting capacity, not specific products.

What do I mean by capacity contracting? Your statistical chances of predicting the total number of cell phones, laptops, etc. that you will sell are much better than your chances of predicting the number of units of each specific cell phone, laptop, etc. that you will sell. If you’ve properly rationalized your supply base, you probably only have a couple of manufacturers making cell phones, and each is probably making multiple models. Instead of guaranteeing them 100,000 units of M1, 50,000 units of M2, and 50,000 units of M3, because you have a high statistical confidence that you’re going to sell at least 250,000 total units this year, guarantee them 200,000 units and allow yourself the ability to specify the actual order quantities at the latest date possible required to meet your turnaround time. Your suppliers win because they are guaranteed business. You win, because you don’t get stuck with a heap of unmovable inventory, which can happen if you incorrectly forecast which model will be your best seller. Furthermore, contract for a quarter or a year, not a month. Demand varies month by month, but is much more predictable quarter by quarter and year by year.