Rapt up in Revenue
When I was in sunny California last, I had a chance to sit down with Rapt and talk about their rather unique solutions that revolve around pricing strategy, decision analytics, and price optimization that, when combined, can help a company maximize their revenue opportunities.
Rapt's sophisticated software platform, that integrates more statistical, analytical, and optimization algorithms than you can shake a stick at, was designed to uncover the many complex supply, demand and price relationships that, when harnessed, predictably improve profit and market share. Unlike simpler modeling tools and platform, Rapt can break down products, or SKUS, into features and analyze the impact of each feature on demand. This is one of the reasons why their solution is becoming popular in high-tech.
Let's say you have three laptops, the Pinta, the Nina, and the Santa Maria, and each are selling quite well. However, like all electronics today, their life-cylce is limited and you need to design your next generation laptop. Each has a different processor, CPU, hard drive, display, and battery life. How do you determine the best configuration for your new laptop? Rapt's forecasting engine can integrate your historical sales data with marketplace data, analyze the sales patterns and trends at the feature level, determine which features (CPU, hard drive, etc.) are the most popular, determine how much each feature influences the overall sale, and tell you which combination of features would sell the best in a laptop. You can then use it's Price Director solution to determine the optimal price-point for your product. This product contains advanced algorithms that work on order, inventory, and market data to extract the elastic and cross-elastic effects among products, their attributes, and consumer demands which it can use to determine the optimal price points for revenue or market-share optimization.
However, one of the most interesting facets of our discussion centered around the fact that the largest uptake in their rather unique solution offering was not in consumer goods industries, but in media, and new media in particular. MSN, Yahoo!, CNET Networks, NBC Universal, The Weather Channel, and MTV Networks, among others, all use Rapt's solution to determine how to price their advertising, which is defined by high variability in demand, uncertain availability of supply, and the rapid innovation and evolution of medium capabilities. If they can tackle one of the most challenging pricing problems out there, surely they can be helpful in more traditional industries. But then again, many companies in these traditional industries most likely have not yet adopted decision optimization in their award process, should-cost modeling in their product design process, or advanced spend visibility solutions in their strategic sourcing process. All I can say is that ... the technology's finally here, let's start to use it!
Categories: Technology, Miscellaneous, Decision Optimization
Unfortunately, past history teaches us that pretty much every quantitative approach to predicting buying patterns (some of them championed by the biggest advertising firms in the world) has ultimately failed. The reason is not that these guys were bad at mathematics or statistics -- they weren't. The reason, I'm told, is that it is simply not useful to (a) ask people why they bought something, (b) try to figure out why they bought it based on its design parameters, or (c) ask them what they would buy if they had their choice -- because none of these answers, however honest or accurate, are believable.
This is well known by companies who do focus groups. For example, the focus group for "Bounce" fabric softener found the concept very unappealing. Questions like, "Won't it burn up?" and "Why would I want to put a piece of paper in the dryer?" were commonplace. If the go-ahead decision had been made based on the focus groups, this fabulously successful product would never have launched.
On the other hand, the focus group for Salvo, a tablet detergent product introduced some 30 years ago, was highly enthusiastic. Salvo, however, was a complete disaster.
Does this kind of experience stop people from doing focus groups? No -- but people who run them understand that the results are basically uninterpretable.
In consumer marketing, buying choices are generally made not according to some formula like optimizing the characteristics of an item (for sneakers, the size of the sole, the color pattern, the type of lacing, and so on). Instead, buy decisions tend to be based on random factors that are largely unanticipated (some basketball player or rock star or TV actor or fashion model with 15 minutes of fame wears a particular brand, or whatever, and it takes off unexpectedly).
So an analysis can be perfect and its results incontestable -- yet the conclusions can still be absolutely worthless.
Just a cautionary note!
Eric: Maybe that's why they are having such success in new media. Nike proved that you can never fully rely on automatic forecasting with it's i2 fiasco (and that not following solution provider advice is not a good idea either!). However, with advertising, especially web-based advertising, you're not producing a product, so there is no equivalent to an inventory overrun. In the worst case, you sell nothing ... but you produce nothing either. There isn't the same downside to a bad pricing prediction. Revenue drops, but it's not the same double whammy as a physical production overrun where revenue drops and costs skyrocket. Furthermore, I would argue that no one really knows how to price a piece of virtual real-estate anyway. There exists a page of ads which provides essentially zero information that earned its creator almost one million dollars. Then there exists websites where single pages contain content more valuable than many books I paid serious dollar for when I was a full time software architect which only generate five cents IF a reader clicks on a link. And although I do agree with you in that forecasting, as I pointed out in this post always requires human oversight and good judgement, I still think an organization would be better off using the best technology available to support them in the creation of those forecasts. The generated forecasts will not be right 100% of the time, and some of the time they will not even be close, but I'm inclined to predict that, on average, they will do better than the vast majority of individuals asked to perform the same function.