Managing the "fuzzy front end"

The fuzzy front end, an essential yet ignored phase in product development.
Yesterday, thanks to that article (Approaches to the "fuzzy front end" of innovation), I have learned a new expression: The fuzzy front end.
It stands for the early development phases of a new product. They are also often called “pre-development”, “pre-project activities” or “pre-phase 0”. The fuzzy front end ranges from the generation of an idea to either its approval for development or its termination (Murphy and Kumar 1997)

According to an extensive empirical study from Cooper and Kleinschmidt, “the greatest differences between winners and losers were found in the quality of execution of pre-development activities”. Two factors were identified to play a major role in product success: the quality of execution of pre-development activities, and a well defined product and project prior to the development phase. Yet, they noted that that pre-development activities received the least amount of attention (only 6 % of dollars and 16 % of man-days of the total) compared to product development and commercialization stages.

The fuzzy front end must be the hardest phase to manage in the product development process. In fact, this phase features very high level of uncertainty and managers are afraid to cope with the conflict between creativity and systematization.

So how should the fuzzy front end be managed? Should it be structured?
There are already lots of different theory on how the whole innovation process should be structured. One of the most cited is the “stage gate process” (FIG 1) The innovation process is divided into five phases from the preliminary assessment of an idea to its commercialization. After every stage there is a gate deciding on continuing or terminating the project. The stage-gate-model integrates the market and technological perspective. Activities are performed in parallel and decisions at the gates are made within cross-functional teams.

The main criticism about this approach is the lack of flexibility. So to overcome that Crawford came up with a model (FIG 2) where the five tasks overlap instead of a linear, sequential path.

Going back to the fuzzy front end, Khurana and Rosenthal, proposed a process model for the pre-development phase (FIG 3). The process model includes activities like product and portfolio strategy formulation which are typically assigned to strategic management. Khurana and Rosenthal emphasize the meaning of foundation elements, e. g. the formulation and communication of a strategic vision, a well planned portfolio of new products, cross-functional sharing of responsibilities, and an information system. Like the other process model, it is a good tool to visualize and structure front end activities, reduce the fuzziness, and ease communication but it lacks flexibility.

This lack is critical for the early phases because the level of uncertainty can vary greatly between the different innovation types. It’s only in later phases (the development phases) that the uncertainties have been reduced to the same level and therefore can be managed in the same way. Innovation strategies must be adapted to the respective uncertainties. We can define four big innovation categories depending on how much technology and market uncertainties they imply (FIG 4).

Now, let’s review the strategies to manage uncertainty in each of the four quadrants.

Incremental Innovation: For innovations that use a mature technology in known markets, the focus should be on the innovation process and accurate quantitative analysis. Successful incremental innovations use external market forecasting techniques, such as customer interviews or customer surveys

Market innovation and technical innovation: Under the condition of low technology and high market uncertainty, a learning-based strategy should be applied. This is also valid for the other conditions with at least one high uncertainty. For this condition, the focus should naturally be on reducing the market uncertainty, for the condition of high technology but low market uncertainty, it is important to reduce the technological risk.

Radical innovation: The final product might not be known, or its ultimate features, costs or technical feasibility. Therefore, it is difficult to determine the potential market. An empirically study by Song and Montoya-Weiss (1998) suggests that for these kind of innovations, thorough strategic planning is a key success factor. Several empirical studies confirm, that a learning-based approach is especially adequate for these kinds of innovations (Lynn and Akgun 1998, Lynn and Green 1998, Rice, O’Connor, Peters and Morone 1998). All areas and functions have to go through extensive learning-processes and sometimes years of trial-and-errors. The emphasis is on gaining maximum information and not on “getting it right” the first time.

Using e-auctions to reduce uncertainties in new product launch

Launching a radically new product brings some undeniable strategic advantages but comes at the price of high risks.

Radical innovators will enjoy the first mover-advantage. They will be alone on the market for some time, especially if their new product is difficult to copy. Even after competition has developed, they will still be remembered has the pioneers and that will help them keep their brand strong on the market.

However radical innovation strategies are, by nature, risky. When a company decides to invest in the development of a radically new product, it generally doesn’t have accurate information on the new market at hand. So it is making a very uncertain bet.

Usually the way to reduce these uncertainties on the size of the market and the price people will be willing to pay for that product is to conduct some market testing. Recently, through three articles (An Exploratory Study on the New Product Demand Curve Estimation Using Online Auction Data, Radically New Product Introduction Using On-line Auctions ), I have discovered an interesting tool to quickly and efficiently run those market tests: e-auctions.

How does e-auctions work?
The Company sets up its own auction platform. Before the auction takes place the company produces a prototype of the new produce to give customers an experience of what it will be like to use this product. It can even be a virtual model. Then these potential consumers are invited to participate into an auction online. They will bid real money and the winner will get the finish product when it is launched (if the company decides not to launch the product, he gets his money back). The auction has to be a sealed-bid auction so that all the bidders will bid independently and only the auctioneer (the company which launches the new product) will have access to all the bids. Each bid represents a "Willingness To Pay" (WTP) for the product.

With these data the company can draw the demand-price curve for the bidders and then extrapolate this curve for the whole market. However it has only assessed the WTP in an e-auction context.

In fact, the WTP will change depending on which channel is used to distribute the product. Some people are uncomfortable using the Internet. An auction might seems more complicated and more risky than just buying the product in a store. For all these reasons the WTP price using an e-auction channel will generally be lower than in a e-shop or in a regular retail store. So the Company has to determine the Internet skill level and sensitivity to the time it takes to make the transaction of the bidders. Then it can extrapolate the demand-price curve for different distribution channels.

Why use e-auctions?
• Tested consumers are putting their money where there mouth is, so the value they give can be trusted.
• E-auctions are cheaper than a regular group test.
• They are fast.
• Using Internet enables to tap in a wide spectrum of the population, geographically spread.

In conclusion, although e-auction won't give perfectly accurate results, especially has the extrapolation phases will introduce some "white noise", its seems a promising way to get more accurate data on demand and price, cheaper and quicker than with conventional group tests.