Forecasting in any field is a risky undertaking: simply put, it is tantamount to predicting outcomes. Be it interest rates in finance, or price of securities in stock market, or price of agri-products in the commodity market, forecasting is formidable to say the least. To forecast trends over large stretches of time—geopolitical, demographical, technological—is even harder not just because of dearth of data but making sense out of it, even if it were available. Researchers do not tire of telling us that so-called experts are no better than chimps.
Leave aside other fields, why should forecasting in supply chains be so difficult, especially with IT tools at our disposal, and data analytics and big data touted as a panacea of sorts. Why do supply chain managers often get it wrong? The fundamental challenge facing all supply chain specialists across chains at various stages remains the same—matching supply with demand. Inability to do so can be perilous: If the real demand is more, then the seller has a stock-out and it results in an irate and lost customer, maybe even permanently; if the real demand is less than anticipated, then the seller is stuck with idle stocks, incurring inventory carrying costs and write-downs.
Just about a decade and a half ago, Cisco’s writing off its unusable inventory worth $2.2 billion in 2001 , one of the largest inventory write-down in the history, shocked the business world and ,even now, the companies whose balance sheets take a hit due to inventory write-offs make frequent headlines, when the quarterly results of the unfortunate ones are reported.
Matching supply with demand would not be that difficult but for the uncertainty of demand; hence, the need for forecasting future demand, the most important aspect of inventory management. Supply chains are dynamic systems, embedded in broader socio-economic structure, affected by changes in society, economy and polity.
It is the consumer demand forecast at the apex of the supply chain that travels downstream at various stages from which different facilities derive their own demand for raw materials and production, although distorted. Accordingly, marketing devises brand and sales promotion; finance decides about budget and investment; human resource focuses on workforce, hiring and layoffs.
A convenient starting point in case of mature products is close examination of past sales history by what is called as time-series analysis. But many a times the past data may not be significant in estimating a future due to demographic, technological or even social changes. In fashion items and electronics, the lifecycle of products as such is very short and sufficient data may not be available to detect a trend, parse out the seasonal and random components of demand. Cell phones and garments, for instance, have a limited time-interval for sales, after which they are passé; the opportunity to correct a forecast error does not even exist.
However, if a company is about to launch a new product, the time-series methods are of no use. Though counter-intuitive, a linear relationship does not exist between many factors driving demand and the demand itself. If lucky, a new item may display ‘The Tipping Point’ behaviour, as popularised by Malcom Gladwell: After a critical mass of people own the item, instead of a gradual increase the sales surge suddenly as a large number of people want to own it—the spread of the product is like a social epidemic. Later the sales plunge all of a sudden, just as they rose.
Pre-ordering, whether a new book or a technological gadget, to know customer demand well in advance has its limits. Enter other techniques that are highly subjective or judgemental. One option is market survey by conducting face-to-face or telephonic interviews and requesting customers to fill a questionnaire. Another is market testing of products to check the customer response and estimating the overall future demand based on sample size.
Subjective as well as extrinsic factors affecting forecasts are innumerable. The state of the economy is relatively easy to handle in forecasting compared to the dynamic factors affecting the market in which a business operates. A competitor may come out with a new pricing strategy, discount offer, promotion campaign and a new look to the product, resetting the market share. All this can be extremely difficult to predict. This affects even the mature products that have a relatively stable demand, say biscuits, toothpastes, and cosmetics in the FMCG sector.
Since no single technique, either different variations of time-series analysis, or qualitative or causal techniques, holds the key, the prudent forecasters combine the various techniques, avoiding mismatch between demand and supply. In any case the managers would go insane if they were to analyse the past demand patterns and forecast, separately for each item. Moreover, the cost of culling such information would be high; hence, aggregation of demand of similar type of items is the only way out.
Despite controlling the variations by smoothing demands, the random component of demand, when asserting itself, is enough to send supply chains in a tizzy. Demands are fickle like the Roman goddess Fortuna, though algorithms may claim to predict consumer behaviour. As Yogi Berra put it, “It's tough to make predictions, especially about the future.”
P.S. Now you know why, I will not hazard any predictions for 2016.
Prashant K Singh is a logistics and supply chain management professional with the Indian Air Force. The views are personal.
He tells how supply chains & logistics affect everything around us on his blog, Unshackled, a part of Business Standard's platform, Punditry.
He tweets as @ZenPK
