Forecasting Over the Product Lifecycle
Demand forecasting is tough, and getting it right over the entire product lifecycle during the course of launch, maturation, and end-of-life is tougher still. This is true for any business with sizable turnover in the product portfolio and especially in sectors like fashion and apparel, electronics, and aftermarket parts.
Forecasting new product introductions are difficult because you don’t have history, and you don’t have a good idea of how quickly the product will take off. Early sales data can be notoriously deceptive due to channel packing—you ship a bunch of new product to DCs or stores to fill bins and shelves, creating an initial surge that is not necessarily a good indicator of sustainable end-user demand.
Substitution and replacement part demand can be somewhat easier to predict, but even they can be tricky if channel packing creates unexpected demand spikes or if it is hard to determine which previous histories are most applicable to how demand for the new product or part will behave.
End-of-life planning and campaign planning require balancing the need to satisfy demand with not getting stuck with inventory that has to be marked down or obsoleted. The goal is to deliver high in-season customer service levels, yet with a graceful exit at end of life or end of season.
Many firms manage this demand variability with manually intensive processes. They often take a different approach for each lifecycle phase, with distinct algorithms for each one. They have to figure out at what point to switch from one phase to another, and make this decision according to the plan, not actual demand. This jolts the supply chain, sending a signal through the chain that is not based on anything that happened to product demand, but rather on the planning algorithm.
The reality is that the product lifecycle is fluid; product doesn’t suddenly “decide” that it is now in one phase or the other. Products typically transition smoothly, and the system must similarly smoothly transit with that demand pattern, as it reveals itself.
Sensing demand—and the subtle or larger changes in the demand pattern—can suggest whether the product is in or transitioning to maturation or an end-of-life demand profile. A demand modeling solution should monitor and sense automatically, and only alert the user on an exception basis if something is out of the ordinary. It then becomes a more hands-off view—as well as a calmer, quieter, more smoothly operating supply chain, in service of the entire product lifecycle.
Often the answer lies downstream—the further downstream you go, the purer the demand signal. Whether product introductions, midcourse sales, or end-of-life downslope, if you can get closer to the point of sale, you have an early signal of demand at each lifecycle phase.
New data sources and machine learning are also making improvements to forecasting the product lifecycle. For instance, if the new product is on your website, you can track indicators such as the number of clicks or page views, how many times someone downloads a specification page, or how much time they spend on a page with detailed product descriptions. If you’re announcing a replacement upgrade on Twitter, you can see how many tweets or retweets you get. If you publish product news on Facebook, you can count shares as early indicators of an end-of-life slowdown.
Forecasting over the entire product lifecycle is still tough – but new methods, new data and new technology are creating opportunities to make it more manageable.