Illustration by Mitch Blunt
Consumer packaged goods (CPG) companies have a big problem: They have almost no idea which of their new products will end up being popular with consumers. Despite big data, despite a decade of heavy investment in innovation, despite chief innovation officers and efficient R&D, failure rates for new products have hovered at 60 percent for years. Two-thirds of new product concepts don’t even launch.
One reason is that the retail environment has become far more complex. E-commerce continues to upend long-established business models, and consumers are shopping less at supermarkets and hypermarkets and more in convenience stores, at discounters, and online.
What’s more, although CPG companies are extremely good at the early stages of innovation—identifying promising areas of growth and creating new product ideas in those areas—and at the later stages of testing concepts and commercializing them, there’s a conspicuous hole in the middle of the process. They don’t have a clear grasp of which combinations of features, packaging, price, and even labeling will persuade consumers to make a purchase. They’re like triathletes who are world-class at swimming and running, but terrible at cycling.
There’s a way to fill that hole, but it won’t be easy. based on our experience, we think it will require progress in three key (and intertwined) areas. None of the three will work without the other two, and all will compel CPG executives to rethink aspects of their traditional business model.
First, companies need to adopt dynamic modeling to gauge various combinations of features. When companies test a product concept today, they’re limited by the relative primitiveness of the tools available to them, such as consumer concept testing and market structure analysis. Testing a preset combination of options (for example, the cinnamon-flavored cookies, in six-ounce individual packages, at 79 cents per pack) produces a basic thumbs-up or thumbs-down assessment as to whether the product will be financially viable. However, the results apply only to that combination. If you change one element, the test results become much less useful. Worse, the testing is expensive and time-consuming, with turnaround times that are measured in months, which makes testing every single combination impossible.
Ideally, companies should be able to test various combinations more dynamically, adjusting the flavor profile, pack size, price, labeling, distribution channel, and any other aspect of the value proposition—even the brand name. Developing a simulation model that can evaluate a wide range of scenarios by altering the various elements and seeing how each factor affects the outcome while the product is still in the development stage is an effective way of doing so.
How much more would consumers pay for low-calorie cinnamon cookies? Would they prefer eight-ounce packs? And should the cookies be sold at a convenience store, a big-box retailer, a warehouse store, or online (or all of the above)? The right model would break such product propositions into their component parts, reassemble them in novel ways, and estimate demand for the new combinations. This in turn would require detailed data on which features consumers value, how much they’re willing to pay for those features, and wher they’re willing to make trade-offs.
In addition, simulation models need to deliver more actionable results. Rather than providing just a basic yes or no, the results must break down revenue, volume, and margin contribution. If a new product is going to take market share from another player, the model should let the company know wher that share will be coming from, at what price, and through which channels. importantly, the model should also indicate how much volume is incremental and how much is simply cannibalizing the company’s other offerings in the same category.
