We must show the pain before we can propose the cure
JeanFrancoisPuget 2700028FGP Comments (7) Visits (12924)
Part of my job is to inject optimization in IBM Anaytics solutions. During one of the discussions with solution teams we argued about a fairly general issue that can prevent prescriptive analytics adoption. I think it is worth sharing.
Specifically, one colleague presented the following analytics classification.
I said that we should rather use the one below (I discussed it in Pres
Why does it matters? It matters because the value of prescriptive analytics is to recommend the best action, especially if optimization is used to compute that action. Well, that's what I'm convinced of at least.
Then, another colleague asked me: "are you a such a specialist of that domain that you can tell seasoned managers and domain experts what they should be doing?"
Of course I am not a specialist. Yet, I am convinced optimization can be used to compute better decisions than seasoned managers and experts. I am convinced given the very consistent succ
Back to the above pictures, the first one basically says that all we could provide are tools and applications that help managers and experts explore the various possibilities and let them select the best one manually. Well, the point of mathematical optimization is to sift through all possibilities in an automated manner, and in a very efficient way. Bottom line is that mathematical optimization can really tell people what they should do. How can we convince people of that?
How can we convince seasoned managers and domain experts that optimization can tell them what they should do?
A remarkable answer is provided by the team that won the 2011 Edelman Award for revenue management work at Intercontinental Hotels Group (IHG). I will summarize it here but interested readers should read their paper:
Dev Koushik, Jon A. Higbie, Craig Eister, (2012) Retail Price Optimization at InterContinental Hotels Group. Interfaces
I can summarize their approach as follows:
We must show the pain before we can propose the cure.
Executives must be convinced that they are not making optimal decisions before we can even try optimization on their problem. Let's see how the revenue analytics and IHG team approached the problem in practice. The problem this team solved was to compute prices that maximize revenue. One of the key issue the team faced was that hotel managers did not see the need for optimized prices. These executives truly believed they were already using optimal pricing strategies.
In order to change that perception, the team created a game where managers competed with optimization. Here is the description from their paper:
To help us in gaining executive approval, we decided to build a simple interactive simulation model in the form of a game (see Figure 3) in which the audience would try to guess what the optimal price should be. [...} For each turn of the game, competitor rates, demand, and capacity varied; the object was to guess the rate that would optimize revenue. The game was fun, but also communicated the challenges revenue managers faced in determining the best rate for a single date. It reminded the audience that revenue managers had to handle multiple-rate products for 350 future arrival dates [...]. If senior executives could not guess the right price in this simple game, how much more challenging was the problem that the hotel revenue managers faced? The pricing game was key in securing the funding we needed.
What a great way to prove the value of optimization!
Update on Feb 18, 2015 The interesting comments below led me to write a sequel post o